Communication method and apparatus, device, chip and storage medium

Abstract

The application provides a communication method, device, equipment, chip and storage medium. The method comprises the following steps: a first capability is sent to a network device, the first capability is used to represent a range of parameter configurations corresponding to one or more AI models supported by a terminal device; a parameter configuration is received from the network device, and communication is performed with the network device based on the parameter configuration.

Description

Communication method, device, equipment, chip and storage medium Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a communication method, apparatus, device, chip and storage medium. Background Art

[0002] In the related art, the terminal device can use the artificial intelligence (AI) model for uplink transmission (such as uplink channel coding) and / or downlink transmission (such as downlink channel decoding). In the process of the terminal device using the AI ​​model for uplink and downlink transmission, it may be necessary to switch the AI ​​model based on the parameter configuration indicated by the network device. However, the switching of the AI ​​model may exceed the capabilities of the terminal device, and may also affect the uplink and downlink transmission performance based on the AI ​​model due to frequent switching of the AI ​​model.

[0003] Summary of the Invention

[0004] The present application provides a communication method, apparatus, device, chip and storage medium.

[0005] In a first aspect, the communication method provided by the present application is applied to a terminal device, and the method includes:

[0006] Sending a first capability to the network device, where the first capability is used to represent a range of parameter configurations corresponding to one or more AI models supported by the terminal device;

[0007] Receive parameter configuration from the network device and communicate with the network device based on the parameter configuration.

[0008] In a second aspect, the communication method provided by the present application is applied to a network device, the method comprising:

[0009] Receive a first capability from a terminal device, where the first capability is used to represent a range of parameter configurations corresponding to one or more AI models supported by the terminal device;

[0010] Based on the first capability, a parameter configuration is sent to the terminal device; wherein the parameter configuration is used for communication between the terminal device and the network device.

[0011] In a third aspect, the present application provides a terminal device, which includes:

[0012] A first communication unit is configured to send a first capability to the network device, where the first capability is used to represent a range of parameter configurations corresponding to one or more AI models supported by the terminal device;

[0013] The first communication unit is further configured to receive parameter configuration from the network device and communicate with the network device based on the parameter configuration.

[0014] In a fourth aspect, the present application provides a network device, the device comprising:

[0015] A second communication unit is configured to receive a first capability from a terminal device, where the first capability is used to represent a range of parameter configurations corresponding to one or more AI models supported by the terminal device;

[0016] The second communication unit is further configured to send parameter configuration to the terminal device based on the first capability; wherein the parameter configuration is used for communication between the terminal device and the network device.

[0017] In a fifth aspect, the present application provides a terminal device comprising a memory, a processor, and a transceiver. The memory is used to store computer programs; the processor is connected to the memory and is used to retrieve and execute the computer programs from the memory to implement the method of the first aspect described above; and the transceiver is used to receive and send information during the process of transmitting and receiving information with other external devices.

[0018] In a sixth aspect, the present application provides a network device comprising a memory, a processor, and a transceiver. The memory is used to store computer programs; the processor is connected to the memory and is used to retrieve and execute the computer programs from the memory to implement the method of the second aspect described above; and the transceiver is used to receive and send information during the process of sending and receiving information with other external devices.

[0019] In a seventh aspect, the present application provides a chip comprising a memory, a processor, and a transceiver. The memory is used to store computer programs; the processor is connected to the memory and is used to retrieve and execute the computer programs from the memory, causing a device equipped with the chip to perform the method of the first or second aspect described above; and the transceiver is used to receive and send information during the process of transmitting and receiving information to and from the device or chip.

[0020] In an eighth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by at least one processor, it implements the method of the first aspect or the second aspect mentioned above.

[0021] In the ninth aspect, the present application provides a computer program product, which includes a computer storage medium, which stores a computer program, and the computer program includes instructions that can be executed by at least one processor, and when the instructions are executed by at least one processor, the method of the first aspect or the second aspect mentioned above is implemented.

[0022] In a tenth aspect, the computer program provided in the present application, when run on a computer, enables the computer to execute the method of the first aspect or the second aspect mentioned above.

[0023] The embodiments of the present application provide a communication method, apparatus, device, chip and storage medium. The terminal device can send a first capability to the network device, and the first capability is used to characterize the range of parameter configurations corresponding to one or more AI models supported by the terminal device; the terminal device can receive the parameter configuration from the network device and communicate with the network device based on the parameter configuration. In this way, the terminal device sends the first capability to the network device so that the network device can reasonably perform parameter configuration, avoid the terminal device switching the AI ​​model based on the parameter configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and avoid frequent AI model switching caused by parameter configuration as much as possible, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0025] FIG1 is a schematic diagram of a communication architecture;

[0026] Figure 2 is a schematic diagram of a neuron structure;

[0027] FIG3 is a schematic diagram of a neural network structure;

[0028] FIG4 is a flow chart of a communication method provided in an embodiment of the present application;

[0029] FIG5 is a schematic diagram of the structure of a terminal device 500 provided in an embodiment of the present application;

[0030] FIG6 is a schematic diagram of the structure of a network device 600 provided in an embodiment of the present application;

[0031] FIG7 is a schematic diagram of the structure of a communication device provided in an embodiment of the present application;

[0032] FIG8 is a schematic diagram of the structure of a chip provided in an embodiment of the present application;

[0033] FIG9 is a schematic block diagram of a communication system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] The following will describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0035] FIG1 is a schematic diagram of an application scenario of an embodiment of the present application.

[0036] As shown in Figure 1, a communication system 100 may include a terminal device 110 and a network device 120. The network device 120 may communicate with the terminal device 110 via an air interface. The terminal device 110 and the network device 120 support multi-service transmission.

[0037] It should be understood that the embodiments of the present application are only illustrative of the communication system 100, but the embodiments of the present application are not limited thereto. That is, the technical solutions of the embodiments of the present application can be applied to various communication systems, such as: Long Term Evolution (LTE) system, LTE Time Division Duplex (TDD), Universal Mobile Telecommunication System (UMTS), Internet of Things (IoT) system, Narrow Band Internet of Things (NB-IoT) system, enhanced Machine-Type Communications (eMTC) system, 5G communication system (also known as New Radio (NR) communication system), or future communication systems.

[0038] In the communication system 100 shown in Figure 1, the network device 120 may be an access network device that communicates with the terminal device 110. The access network device may provide communication coverage for a specific geographical area and may communicate with the terminal device 110 located within the coverage area.

[0039] The network device 120 may be an evolved Node B (eNB or eNodeB) in an LTE system, or a Next Generation Radio Access Network (NG RAN) device, or a base station (gNB) in an NR system, or a wireless controller in a Cloud Radio Access Network (CRAN), or the network device 120 may be a relay station, an access point, a vehicle-mounted device, a wearable device, a hub, a switch, a bridge, a router, or a network device in a future evolved Public Land Mobile Network (PLMN), etc.

[0040] The terminal device 110 may be any terminal device, including but not limited to a terminal device connected to the network device 120 or other terminal devices by wire or wireless connection.

[0041] For example, the terminal device 110 may refer to an access terminal, user equipment (UE), a subscriber unit, a subscriber station, a mobile station, a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, a terminal, a wireless communication device, a user agent, or a user apparatus. The access terminal may be a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, an IoT device, a satellite handheld terminal, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA), a handheld device with wireless communication capabilities, a computing device or other processing device connected to a wireless modem, an in-vehicle device, a wearable device, a terminal device in a 5G network, or a terminal device in a future evolution network, etc.

[0042] The terminal device 110 can be used for device-to-device (D2D) communication.

[0043] FIG1 exemplarily shows a network device and two terminal devices. Optionally, the communication system 100 may include multiple network devices and each network device may include another number of terminal devices within its coverage area, which is not limited in this embodiment of the present application.

[0044] It should also be understood that the "predefined" or "predefined rules" mentioned in the embodiments of the present application can be implemented by pre-saving corresponding codes, tables or other methods that can be used to indicate relevant information in devices (for example, including terminal devices and network devices). This application does not limit its specific implementation method.

[0045] To facilitate understanding of the technical solutions of the embodiments of the present application, the relevant technologies of the embodiments of the present application are described below. The following relevant technologies can be arbitrarily combined with the technical solutions of the embodiments of the present application as optional solutions, and they all fall within the protection scope of the embodiments of the present application.

[0046] AI models are capable of handling a variety of tasks. They possess the ability to self-learn and adapt, dynamically adjusting and making decisions based on environmental changes. AI models can also be referred to as machine learning (ML) models; the two are equivalent or interchangeable.

[0047] In practical applications, AI models can be constructed using neural networks. A neural network is a computational model consisting of multiple interconnected neuron nodes. The connections between nodes represent weighted values ​​from input signals to output signals, called weights. Each node performs a weighted summation of different input signals and outputs the result through a specific activation function. Refer to the neuron structure diagram shown in Figure 2, where a1, a2, ..., an, and 1 are neuron inputs, w1, w2, ..., wn, and b represent weights, Sum represents the summation function, f represents the activation function, and t represents the output result.

[0048] A simple neural network, shown in Figure 3, consists of an input layer, hidden layers, and an output layer. By varying the connections, weights, and activation functions of multiple neurons, different outputs can be generated, thereby fitting the mapping from input to output. Each node in the previous level is connected to all nodes in the next level. This fully connected model is also called a deep neural network (DNN).

[0049] An AI model can be trained and obtained through the process of constructing a dataset, training, validating, and testing it. When a neural network is used for wireless communication between terminal devices and network devices, training can be divided into offline training and online training. Network devices can obtain a static training result through offline training of a dataset, which is referred to as offline training. During the use of the AI ​​model by network devices or terminal devices, as the terminal devices further measure and / or report, the network devices can continue to collect more data and conduct real-time online training to optimize the AI ​​model parameters and achieve better inference and prediction results. After obtaining the AI ​​model, the corresponding model output can be inferred by inputting the current information into the AI ​​model.

[0050] When AI models are used in wireless communications, they can be categorized as single-ended or dual-ended AI models. Single-ended AI models are deployed on one side of a terminal device or network device, allowing for immediate use and training on that side. Dual-ended AI models require pairing on both the terminal device and the network device, with both models trained together. This means the AI ​​models deployed on both sides correspond to each other and cannot be used or updated independently.

[0051] To achieve different functions, different AI models can be introduced and corresponding inputs and outputs can be defined. For example, when using an AI model for channel state information (CSI) feedback, the channel information (such as eigenvectors, beam information, delay information, etc.) obtained based on reference signal measurements can be used as the input of the AI ​​model to infer the corresponding CSI quantization bits. On the network device side, there will be a corresponding AI model that uses the CSI quantization bits as input to infer the corresponding channel information.

[0052] Exemplarily, when the AI ​​model is used for beam management, the terminal device can use the measured reference signal receiving power (RSRP) corresponding to multiple beams (CSI-RS resources) in the second beam (Channel State Information-Reference Signal (CSI-RS) resource) set as input, and infer the optimal beam (CSI-RS resource index) and corresponding RSRP in the first beam set based on the first AI model, and report the inference result to the network device. In addition, the AI ​​model can also be used for other processes such as positioning, channel coding, data channel decoding, modulation and demodulation, and channel estimation.

[0053] The generalization ability of an AI model is one of the most important concepts in machine learning. It determines the practical value of an AI model and has become a key metric for evaluating the quality of an AI model. The generalization ability of an AI model refers to the ability of the AI ​​model to apply the knowledge learned from the training set to the test set or a new dataset. In other words, AI models with better generalization ability can be applied to different datasets (such as test sets or new datasets), while AI models with poor generalization ability are prone to overfitting, that is, performing well on the training set but poorly on the test set or a new dataset.

[0054] In embodiments of the present application, the generalization capability of the AI ​​model used by a terminal device can be characterized by the range of parameter configurations corresponding to the AI ​​model. If the terminal device can use the same AI model to process different parameter configuration values ​​indicated by a network device, i.e., the same AI model can support different parameter configuration values, then this AI model has excellent generalization capability. If the terminal device needs to switch between different AI models, i.e., use different AI models to support different parameter configuration values, then this AI model has poor generalization capability.

[0055] For example, when the terminal device has trained the same AI model for different values ​​of parameter configuration, the terminal device can use the same AI model for the parameter configuration indicated by the network device. In this case, the generalization capability of the AI ​​model is better.

[0056] For example, when the terminal device has trained different AI models for different values ​​of parameter configuration, the terminal device may need to use a different AI model for the parameter configuration indicated by the network device. In this case, the generalization capability of the AI ​​model is poor.

[0057] In the related art, the terminal device can use the AI ​​model for uplink transmission (such as uplink channel coding, uplink precoding, and uplink modulation) and / or downlink transmission (such as downlink channel decoding, downlink channel estimation, and downlink data detection). In the process of the terminal device using the AI ​​model for uplink and downlink transmission, the generalization capabilities of the AI ​​models used by different terminal devices are different. Since the network device does not know the generalization capabilities of the AI ​​model used by the terminal device, it is impossible to determine whether the indicated parameter configuration will cause the switching of the AI ​​model. In the case where the generalization capability of the AI ​​model is poor (such as the terminal device has trained different AI models for different values ​​of the parameter configuration), the terminal device may need to switch the AI ​​model based on the parameter configuration indicated by the network device. However, the switching of the AI ​​model may exceed the capabilities of the terminal device, and may also affect the uplink and downlink transmission performance based on the AI ​​model due to frequent switching of the AI ​​model.

[0058] Based on this, an embodiment of the present application provides a communication method, in which a terminal device can send a first capability to a network device, and the first capability is used to characterize the range of parameter configurations corresponding to one or more AI models supported by the terminal device; the terminal device can receive the parameter configuration from the network device and communicate with the network device based on the parameter configuration. In this way, the terminal device sends the first capability to the network device, so that the network device can reasonably configure the parameters, avoid the terminal device switching the AI ​​model based on the parameter configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and avoid frequent AI model switching caused by parameter configuration as much as possible, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0059] To facilitate understanding of the technical solutions of the embodiments of the present application, the technical solutions of the present application are described in detail below through specific embodiments. The above related technologies can be combined arbitrarily with the technical solutions of the embodiments of the present application as optional solutions, and all of them fall within the scope of protection of the embodiments of the present application. The embodiments of the present application include at least part of the following contents.

[0060] FIG4 is a flow chart of a communication method provided in an embodiment of the present application. As shown in FIG4 , the method may include the following steps.

[0061] S410. The terminal device sends a first capability to the network device. The first capability is used to represent the range of parameter configurations corresponding to one or more AI models supported by the terminal device.

[0062] Correspondingly, the network device receives the first capability reported by the terminal device.

[0063] It should be noted that the terminal device sends the first capability to the network device, which can be understood as the terminal device sending the generalization capability of one or more AI models to the network device. The generalization capability of the one or more AI models can be represented by the range of parameter configurations corresponding to one or more AI models supported by the terminal device.

[0064] Accordingly, the network device can receive the generalization capabilities of one or more AI models from the terminal device.

[0065] It should be noted that the range of parameter configuration corresponding to one or more AI models can be understood as the range of parameter configuration corresponding to one AI model when there is one AI model; or, when there are multiple AI models, the range of parameter configuration corresponding to each of the multiple AI models.

[0066] In some embodiments, the parameter configuration may include the configuration of one or more of the following parameters: demodulation reference signal (DMRS), CSI-RS, CSI reporting content, modulation and coding scheme (MCS), transport block size, number of transmission layers, transmission bandwidth, number of antenna ports, carrier frequency, subcarrier spacing, transmission configuration indicator (TCI) status, cell identity (ID), and multiple-input multiple-output (MIMO) transmission scheme.

[0067] In some embodiments, the AI ​​models corresponding to different parameter configuration ranges may be independent or interrelated, and this is not limited in the embodiments of the present application.

[0068] It should be noted that the AI ​​models corresponding to different parameter configuration ranges can be independent, that is, different parameter configuration ranges can correspond to different AI models. For example, AI model 1, AI model 2, and AI model 3 are three different AI models. The DMRS configuration range can correspond to AI model 1, the CSI-RS configuration range can correspond to AI model 2, and the CSI reporting content configuration range can correspond to AI model 3.

[0069] It should also be noted that the AI ​​models corresponding to different parameter configuration ranges can also be interrelated, that is, different parameter configuration ranges can correspond to the same AI model. For example, the range of transport block size configuration, the range of transport layer configuration, the range of transmission bandwidth configuration, and the range of antenna port configuration can all correspond to AI model 1; for another example, the range of carrier frequency configuration, the range of subcarrier spacing configuration, and the range of TCI state configuration can all correspond to AI model 2.

[0070] In some embodiments, the range of parameter configurations corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the parameter configurations.

[0071] That is to say, the range of parameter configuration corresponding to one or more AI models may include: when the terminal device supports only one AI model, the terminal device can use the one AI model for different values ​​of the parameter configuration; when the terminal device supports multiple AI models, the terminal device can use different AI models for different values ​​of the parameter configuration.

[0072] It should be noted that the terminal device can use one bit to indicate to the network device whether the AI ​​model deployed on the terminal device side supports different values ​​of parameter configuration.

[0073] For example, assume that the parameter configuration has P values, which are respectively recorded as value 1, value 2, ..., value P. When the number of AI models supported by the terminal device is one, for value 1, value 2, ..., value P, the terminal device can use the one AI model.

[0074] For example, when the parameter configuration is DMRS configuration, assuming that the AI ​​model supported by the terminal device is AI model 1, for the three values ​​of the DMRS base sequence: sequence ID1, sequence ID2 and sequence ID3, the terminal device can use the AI ​​model 1.

[0075] For another example, when the parameter configuration is CSI-RS configuration, assuming that the AI ​​model supported by the terminal device is AI model 2, for the three values ​​of the CSI-RS symbol number: symbol number 1, symbol number 2 and symbol number 3, the terminal device can use the AI ​​model 2.

[0076] For another example, when the parameter configuration is CSI reporting content configuration, assuming that the AI ​​model supported by the terminal device is AI model 3, the terminal device can use the AI ​​model 3 for RSRP reporting and CSI reporting.

[0077] For another example, when the parameter configuration is MCS configuration, assuming that the AI ​​model supported by the terminal device is AI model 4, for MCS indexes 0-12 and 12-27, the terminal device can use AI model 4.

[0078] For example, it is assumed that the parameter configuration has Q values, which are respectively recorded as value 1, value 2, ..., value Q. When the terminal device supports multiple AI models, the terminal device can use different AI models for value 1, value 2, ..., value Q.

[0079] For example, when the parameter configuration is DMRS configuration, assuming that the AI ​​models supported by the terminal device are AI model 1 and AI model 2, for the two values ​​of the DMRS base sequence: sequence ID1 and sequence ID2, the terminal device can use AI model 1 and AI model 2 respectively.

[0080] For another example, when the parameter configuration is CSI-RS configuration, assuming that the AI ​​models supported by the terminal device are AI model 3 and AI model 4, for the two values ​​of the CSI-RS symbol number: symbol number 1 and symbol number 2, the terminal device can use AI model 3 and AI model 4 respectively.

[0081] For another example, when the parameter configuration is CSI reporting content configuration, assuming that the AI ​​models supported by the terminal device are AI model 5 and AI model 6, the terminal device can use AI model 5 and AI model 6 for RSRP reporting and CSI reporting, respectively.

[0082] For another example, when the parameter configuration is MCS configuration, assuming that the AI ​​models supported by the terminal device are AI model 7 and AI model 8, for MCS indexes 0-12 and 12-27, the terminal device can use AI model 7 and AI model 8.

[0083] The following describes in detail whether terminal devices use the same AI model when different parameter configuration values ​​are used, combining several possible implementation methods.

[0084] In one possible implementation, the range of DMRS configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the DMRS configuration.

[0085] Since the training of the AI ​​model relies on the received data / signal of DMRS as input, different DMRS configurations may correspond to different AI models, and different DMRS inputs will also result in different AI model outputs.

[0086] The terminal device can train on an AI model for all DMRS configurations that may be indicated by the network device (such as DMRS base sequence, DMRS code division multiplexing (CDM) group, DMRS port multiplexing mode, number of DMRS ports, DMRS port set, etc.), so that the trained AI model has better generalization capabilities. The terminal device does not need to switch the AI ​​model for any DMRS configuration indicated by the network device. However, this training method requires a large data / signal set, the model parameter adjustment is complex, and the obtained channel estimation accuracy or detection performance is also difficult to guarantee.

[0087] The terminal device can also use different AI models for different DMRS configurations (such as different DMRS base sequences, different DMRS CDM groups, and different DMRS port multiplexing methods) for separate training. Each AI model targets part of the configuration, which not only reduces the complexity of training, but also improves channel estimation performance or detection performance. However, the generalization ability of the AI ​​model obtained based on this solution is poor. When the network device adopts different DMRS configurations, the terminal device may need to switch the corresponding AI model so that the AI ​​model corresponding to the current DMRS configuration can be used for channel estimation or detection.

[0088] For example, assuming that the DMRS base sequence has two values, denoted as value 1 and value 2, the range of the DMRS base sequence corresponding to one or more AI models may include whether the terminal device uses the same AI model for the two values ​​of the DMRS base sequence. For example, for values ​​1 and 2, the terminal device uses the same AI model; for another example, for values ​​1 and 2, the terminal device uses different AI models.

[0089] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated DMRS configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the DMRS configuration (such as adjusting the DMRS base sequence, or adjusting the DMRS CDM group) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0090] In another possible implementation, the range of CSI-RS configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the CSI-RS configuration.

[0091] Since CSI measurement is based on CSI-RS, AI model training relies on CSI-RS received data / signals as input. Different CSI-RS configurations correspond to different AI models, and different CSI-RS data / signal inputs also result in different AI model outputs.

[0092] The terminal device can train on a single AI model for all CSI-RS configurations that the network device may indicate (such as the CSI-RS base sequence, CSI-RS port number set, CSI-RS port multiplexing method, CSI-RS symbol number, CSI-RS transmission bandwidth, etc.). This allows the trained AI model to have superior generalization capabilities, and the terminal device can avoid switching AI models for any CSI-RS configuration indicated by the network device. However, this training method requires a large data / signal set, the model parameter adjustment is complex, and the accuracy of the resulting CSI feedback is difficult to guarantee.

[0093] The terminal device can also use different AI models for different CSI-RS configurations (such as different CSI-RS base sequences, different sets of CSI-RS port numbers, and different CSI-RS port multiplexing methods) for separate training. Each AI model targets part of the configuration, which not only reduces the complexity of training but also improves CSI feedback performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When the network device adopts a different CSI-RS configuration, the terminal device may need to switch the corresponding AI model so that the AI ​​model corresponding to the current CSI-RS configuration can be used for CSI measurement.

[0094] For example, assuming the CSI-RS base sequence has two values, denoted as value 1 and value 2, the range of the CSI-RS base sequence corresponding to one or more AI models may include whether the terminal device uses the same AI model for the two values ​​of the CSI-RS base sequence. For example, the terminal device uses the same AI model for values ​​1 and 2; for another example, the terminal device uses different AI models for values ​​1 and 2.

[0095] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated CSI-RS configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the CSI-RS configuration (such as adjusting the CSI-RS base sequence, or adjusting the CSI-RS CDM group) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0096] In another possible implementation, the range of the CSI reporting content configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the CSI reporting content configuration.

[0097] When the CSI reporting content configuration output by the AI ​​model is different, the AI ​​model used by the terminal device may also be different. The training of the AI ​​model depends on the CSI reporting content configuration as output.

[0098] The terminal device can train a single AI model for all CSI reporting configurations that the network device may indicate (such as RSRP reporting and CSI reporting, CSI compression and CSI prediction, CSI compression, CSI prediction, and CSI prediction and compression, etc.). This results in a trained AI model with superior generalization capabilities, allowing the terminal device to avoid switching AI models for any CSI reporting configuration indicated by the network device. However, this training method requires a large data / signal set, complicates model parameter adjustment, and makes it difficult to guarantee the accuracy of the resulting CSI feedback.

[0099] The terminal device can also use different AI models for training according to different CSI reporting content configurations (such as different AI models for RSRP reporting and CSI reporting, different AI models for CSI compression and CSI prediction, and different AI models for CSI compression, CSI prediction, and CSI prediction and compression). Each AI model is configured for a CSI reporting content. This not only reduces the complexity of training but also improves CSI feedback performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When the network device uses different CSI reporting content configurations, the terminal device may need to switch the corresponding AI model so that the AI ​​model corresponding to the current CSI reporting content configuration can be used for CSI feedback.

[0100] For example, the scope of CSI reporting content configuration corresponding to one or more AI models may include whether the terminal device uses the same AI model for RSRP reporting and CSI reporting. For example, the terminal device uses the same AI model for RSRP reporting and CSI reporting; for another example, the terminal device uses different AI models for RSRP reporting and CSI reporting.

[0101] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated CSI reporting content configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the CSI reporting content configuration (such as adjusting RSRP reporting and CSI reporting, or adjusting CSI compression and CSI prediction) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0102] In another possible implementation, the range of the MCS configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the MCS configuration.

[0103] AI model training depends on the MCS configuration of uplink and downlink data / signals. The number of antenna ports and transmission layers affects the input of AI model training data / signals. In other words, terminal devices may train different AI models for different MCS configurations.

[0104] Terminal devices can train a single AI model for all MCS configurations supported by the protocol (such as MCS index range, modulation mode, modulation mode set, bit rate range, etc.). This results in a highly generalizable AI model, allowing the device to adapt to any MCS configuration indicated by the network device without switching the AI ​​model. However, this training method requires a large data / signal set, complicates model parameter adjustment, and makes it difficult to guarantee data / signal transmission performance.

[0105] Terminal devices can also use different AI models for training for different MCS configurations (such as different MCS index ranges, different modulation methods, and different bit rate ranges). Each AI model is trained for part of the configuration. This not only reduces the complexity of training, but also improves data / signal transmission performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When the network device adopts a different MCS configuration, the terminal device may need to switch the corresponding AI model so that the AI ​​model corresponding to the current MCS configuration can be used for data / signal transmission.

[0106] For example, assuming there are two modulation modes, designated as modulation mode 1 and modulation mode 2, the range of modulation modes corresponding to one or more AI models may include whether the terminal device uses the same AI model for modulation mode 1 and modulation mode 2. For example, the terminal device uses the same AI model for modulation mode 1 and modulation mode 2; for another example, the terminal device uses different AI models for modulation mode 1 and modulation mode 2.

[0107] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated MCS configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the MCS configuration (such as adjusting the MCS index range, adjusting the modulation mode, or adjusting the bit rate range) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0108] In another possible implementation, the range of transmission block size configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the transmission block size configuration.

[0109] AI model training depends on the transmission block size configuration of uplink and downlink data / signals. The number of antenna ports and transmission layers affects the input of AI model training data / signals. In other words, terminal devices may train different AI models for different transmission block size configurations.

[0110] Terminal devices can train a single AI model for all transport block size configurations supported by the protocol (e.g., transport block size ranges, etc.). This results in a highly generalizable AI model, allowing the device to adapt to any transport block size configuration specified by the network device without switching the AI ​​model. However, this training approach requires a large data / signal set, complicates model parameter adjustment, and makes it difficult to guarantee the transmission performance of the corresponding data / signal.

[0111] Terminal devices can also use different AI models for different transmission block size configurations (such as different transmission block size ranges) for separate training, with each AI model targeting a portion of the configuration. This not only reduces the complexity of training but also improves data / signal transmission performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When network devices use different transmission block size configurations, terminal devices may need to switch the corresponding AI model so that the AI ​​model corresponding to the current transmission block size configuration can be used for data / signal transmission.

[0112] For example, assuming that the transport block size has two values, denoted as value 1 and value 2, the range of transport block size configurations corresponding to one or more AI models may include whether the terminal device uses the same AI model for the two transport block size values. For example, for values ​​1 and 2, the terminal device uses the same AI model; for another example, for values ​​1 and 2, the terminal device uses different AI models.

[0113] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated transmission block size configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the transmission block size configuration (such as adjusting the transmission block size range) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0114] In another possible implementation, the range of transmission layer configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the transmission layer configuration.

[0115] Because data / signal transmission and reception can be performed at the transport layer, AI model training depends on the number of transport layers configured, which in turn affects the input data / signals used for AI model training. This means that under different transport layer configurations, terminal devices may train different AI models for data / signal transmission or reception.

[0116] Terminal devices can train a single AI model for all transmission layer configurations (e.g., transmission layer ranges, etc.). This results in a trained AI model with superior generalization capabilities, allowing the terminal device to remain consistent with any transmission layer configuration specified by the network device without switching AI models. However, this training method requires a large data / signal set, complicates model parameter adjustment, and can make it difficult to guarantee the transmission performance of the corresponding data / signal.

[0117] Terminal devices can also use different AI models for different transmission layer configurations (such as different transmission layer ranges) for separate training. Each AI model targets a portion of the configuration. This not only reduces the complexity of training but also improves data / signal transmission performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When network devices use different transmission layer configurations, terminal devices may need to switch the corresponding AI model so that the AI ​​model corresponding to the current transmission layer configuration can be used for data / signal transmission.

[0118] For example, assuming the number of transmission layers has two possible values, denoted as 1 and 2, the range of transmission layer configurations corresponding to one or more AI models may include whether the terminal device uses the same AI model for the two possible values ​​of the transmission layer number. For example, the terminal device uses the same AI model for both values ​​1 and 2; for another example, the terminal device uses different AI models for both values ​​1 and 2.

[0119] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated transmission layer configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the transmission layer configuration (such as adjusting the transmission layer range) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0120] In another possible implementation, the range of transmission bandwidth configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the transmission bandwidth configuration.

[0121] Because data / signal transmission and reception can be based on transmission bandwidth, AI model training depends on the transmission bandwidth configuration value, which affects the input of AI model training data / signals. In other words, under different transmission bandwidth configurations, terminal devices may train different AI models for data / signal transmission or reception.

[0122] Terminal devices can train a single AI model for all bandwidth configurations (such as bandwidth ranges), resulting in a highly generalizable AI model. The AI ​​model can be trained without switching models for any bandwidth configuration specified by the network device. However, this training method requires a large data / signal set, complicates model parameter adjustment, and can be difficult to guarantee data / signal transmission performance.

[0123] Terminal devices can also use different AI models for different transmission bandwidth configurations (such as different transmission bandwidth ranges) for separate training. Each AI model targets a portion of the configuration. This not only reduces the complexity of training but also improves data / signal transmission performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When network devices use different transmission bandwidth configurations, terminal devices may need to switch the corresponding AI model so that the AI ​​model corresponding to the current transmission bandwidth configuration can be used for data / signal transmission.

[0124] For example, assuming that the transmission bandwidth has two values, denoted as value 1 and value 2, the range of the transmission bandwidth configuration corresponding to one or more AI models may include whether the terminal device uses the same AI model for the two transmission bandwidth values. For example, for values ​​1 and 2, the terminal device uses the same AI model; for another example, for values ​​1 and 2, the terminal device uses different AI models.

[0125] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated transmission bandwidth configuration will cause the switching of the AI ​​model. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the transmission bandwidth configuration (such as adjusting the transmission bandwidth range) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0126] In another possible implementation, the range of antenna port number configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the antenna port number configuration.

[0127] Since data / signal transmission and reception can be performed based on antenna ports, AI model training depends on the antenna port configuration value, which affects the input of AI model training data / signals. In other words, under different antenna port configurations, the terminal device may train different AI models for data / signal transmission or reception.

[0128] Terminal devices can train a single AI model for all antenna port configurations (e.g., antenna port ranges, etc.). This results in a highly generalizable AI model, allowing the device to operate without switching AI models for any antenna port configuration specified by the network device. However, this training method requires a large data / signal set, complicates model parameter adjustment, and makes it difficult to guarantee data / signal transmission performance.

[0129] Terminal devices can also use different AI models for different antenna port configurations (such as different antenna port ranges) for separate training. Each AI model targets a portion of the configuration. This not only reduces the complexity of training but also improves data / signal transmission performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When network devices use different antenna port configurations, terminal devices may need to switch the corresponding AI model so that the AI ​​model corresponding to the current antenna port configuration can be used for data / signal transmission.

[0130] For example, assuming the number of antenna ports has two possible values, denoted as 1 and 2, the range of antenna port number configurations corresponding to one or more AI models may include whether the terminal device uses the same AI model for the two possible values ​​of the number of antenna ports. For example, the terminal device uses the same AI model for both values ​​1 and 2; for another example, the terminal device uses different AI models for both values ​​1 and 2.

[0131] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated antenna port number configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the antenna port number configuration (such as adjusting the antenna port number range) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0132] In another possible implementation, the range of carrier frequency configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the carrier frequency configuration.

[0133] Since data / signal transmission and reception can be based on carrier frequency, AI model training depends on the carrier frequency configuration value, which affects the input of AI model training data / signals. In other words, under different carrier frequency configurations, terminal devices may train different AI models for data / signal transmission or reception.

[0134] Terminal devices can train a single AI model for all carrier frequency configurations (e.g., carrier frequency ranges, etc.). This results in a highly generalizable AI model, allowing the device to adapt to any carrier frequency configuration specified by the network device without switching the AI ​​model. However, this training method requires a large data / signal set, complicates model parameter adjustment, and makes data / signal transmission performance difficult to guarantee.

[0135] Terminal devices can also use different AI models for different carrier frequency configurations (such as different carrier frequency ranges) for separate training, with each AI model targeting a portion of the configuration. This not only reduces the complexity of training but also improves data / signal transmission performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When network devices use different carrier frequency configurations, terminal devices may need to switch the corresponding AI model so that the AI ​​model corresponding to the current carrier frequency configuration can be used for data / signal transmission.

[0136] For example, assuming that the carrier frequency has two values, denoted as value 1 and value 2, the range of the carrier frequency configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for the two values ​​of the carrier frequency. For example, for values ​​1 and 2, the terminal device uses the same AI model; for another example, for values ​​1 and 2, the terminal device uses different AI models.

[0137] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated carrier frequency configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the carrier frequency configuration (such as adjusting the carrier frequency range) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0138] In another possible implementation, the range of subcarrier spacing configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the subcarrier spacing configuration.

[0139] Since data / signal transmission and reception can be based on subcarrier spacing, AI model training depends on the subcarrier spacing configuration value, which affects the input of AI model training data / signals. In other words, under different subcarrier spacing configurations, terminal devices may train different AI models for data / signal transmission or reception.

[0140] Terminal devices can train a single AI model for all subcarrier spacing configurations (e.g., subcarrier spacing ranges, etc.). This results in a highly generalizable AI model, allowing the device to adapt to any subcarrier spacing configuration specified by the network without switching the AI ​​model. However, this training approach requires a large data / signal set, complicates model parameter adjustment, and can be difficult to guarantee data / signal transmission performance.

[0141] Terminal devices can also use different AI models for different subcarrier spacing configurations (such as different subcarrier spacing ranges) for separate training, with each AI model targeting part of the configuration. This not only reduces the complexity of training but also improves data / signal transmission performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When network devices use different subcarrier spacing configurations, terminal devices may need to switch the corresponding AI model so that the AI ​​model corresponding to the current subcarrier spacing configuration can be used for data / signal transmission.

[0142] For example, assuming that the subcarrier spacing has two values, denoted as value 1 and value 2, the range of subcarrier spacing configurations corresponding to one or more AI models may include: whether the terminal device uses the same AI model for the two subcarrier spacing values. For example, for values ​​1 and 2, the terminal device uses the same AI model; for another example, for values ​​1 and 2, the terminal device uses different AI models.

[0143] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated subcarrier spacing configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the subcarrier spacing configuration (such as adjusting the subcarrier spacing range) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0144] In another possible implementation, the range of TCI state configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the TCI state configuration.

[0145] Because data / signal transmission and reception can be based on the TCI state, AI model training depends on the TCI state configuration value, and the TCI state affects the input of AI model training data / signals. In other words, under different TCI state configurations, the terminal device may train different AI models for data / signal transmission or reception.

[0146] Terminal devices can train a single AI model for all TCI state configurations. This results in a highly generalizable AI model, allowing the device to remain consistent with any TCI state configuration indicated by the network device without switching AI models. However, this training method requires a large data / signal set, complicates model parameter adjustment, and can be difficult to guarantee data / signal transmission performance.

[0147] Terminal devices can also use different AI models for different TCI state configurations (such as single TCI state and multiple TCI states) for separate training. Each AI model targets a portion of the configuration. This not only reduces the complexity of training but also improves data / signal transmission performance. However, the generalization capability of the AI ​​model obtained based on this solution is poor. When network devices use different TCI state configurations, terminal devices may need to switch the corresponding AI model so that the AI ​​model corresponding to the current TCI state configuration can be used for data / signal transmission.

[0148] For example, the scope of the TCI state configuration corresponding to one or more AI models may include whether the terminal device uses the same AI model for both the single TCI state and the multi-TCI state. For example, the terminal device uses the same AI model for both the single TCI state and the multi-TCI state; for another example, the terminal device uses different AI models for both the single TCI state and the multi-TCI state.

[0149] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated TCI state configuration will cause the switching of the AI ​​model. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the TCI state configuration will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0150] In another possible implementation, the range of cell ID configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different values ​​of the cell ID configuration.

[0151] Since terminal devices need to access a cell before sending or receiving data / signals, data / signal transmission occurs between the terminal device and the serving cell. Therefore, AI model training can be performed based on the cell. The AI ​​model training depends on the value of the cell ID configuration, which affects the input of AI model training data / signals. In other words, when the serving cell changes, for example, when the terminal device accesses a cell with a different cell ID, the AI ​​model used may also be different.

[0152] Terminal devices can train a single AI model for all cell ID configurations (e.g., cell ID ranges, etc.). This results in a highly generalizable AI model, allowing the device to remain consistent with any cell ID configuration specified by the network device without switching AI models. However, this training approach requires a large data / signal set, complicates model parameter adjustment, and makes data / signal transmission performance difficult to guarantee.

[0153] Terminal devices can also use different AI models for different cell ID configurations (such as different cell ID ranges) for separate training, with each AI model targeting a portion of the configuration. This not only reduces the complexity of training but also improves data / signal transmission performance. However, the AI ​​model obtained based on this solution has poor generalization capabilities. When network devices use different cell ID configurations, terminal devices may need to switch the corresponding AI model so that the AI ​​model corresponding to the current cell ID configuration can be used for data / signal transmission.

[0154] For example, assuming there are two cell IDs, ID1 and ID2, the range of cell ID configurations corresponding to one or more AI models can include whether the terminal device uses the same AI model for ID1 and ID2. For example, the terminal device uses the same AI model for ID1 and ID2; for another example, the terminal device uses different AI models for ID1 and ID2.

[0155] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated cell ID configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the cell ID configuration (such as adjusting the cell ID range) will cause the terminal device side to switch the AI ​​model based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0156] In another possible implementation, the range of MIMO transmission scheme configuration corresponding to one or more AI models may include: whether the terminal device uses the same AI model for different MIMO transmission scheme configurations.

[0157] Since terminal devices transmit and receive data / signals, or measure CSI, all based on certain MIMO transmission schemes, AI models used for these functions must be trained based on certain MIMO transmission scheme assumptions. The MIMO transmission scheme configuration affects the input data / signals used for AI model training and the CSI measurement results. In other words, under different MIMO transmission scheme configurations, terminal devices may train different AI models for data / signal transmission or reception, or CSI measurement.

[0158] Terminal devices can train a single AI model for all MIMO transmission scheme configurations. This results in a highly generalizable AI model, allowing the device to adapt to any MIMO transmission scheme specified by the network device without switching AI models. However, this training method requires a large data / signal set, complicates model parameter adjustment, and can be difficult to guarantee for data / signal transmission performance and CSI measurement accuracy.

[0159] Terminal devices can also use different AI models for training for different MIMO transmission scheme configurations (such as single transmission receiving point (Transmit / Receive Point, TRP) transmission schemes and multi-TRP collaborative schemes, such as open-loop transmission schemes and closed-loop transmission schemes). Each AI model is configured for part of them. This not only reduces the complexity of training, but also improves data / signal transmission performance or CSI measurement accuracy. However, the generalization ability of the AI ​​model obtained based on this scheme is poor. When the network device adopts different MIMO transmission scheme configurations, the terminal device may need to switch the corresponding AI model so that the AI ​​model corresponding to the current MIMO transmission scheme configuration can be used for data / signal transmission or CSI measurement.

[0160] For example, the range of TCI state configurations corresponding to one or more AI models may include: whether the terminal device uses the same AI model for a single TRP transmission scheme and a multi-TRP collaboration scheme. For example, the terminal device uses the same AI model for a single TRP transmission scheme and a multi-TRP collaboration scheme; for another example, the terminal device uses different AI models for a single TRP transmission scheme and a multi-TRP collaboration scheme.

[0161] Through this method, different terminal devices can adopt different AI model training methods, so that AI models with different generalization capabilities can be obtained. Since the network device does not know the generalization capabilities of one or more AI models used by the terminal device, it is impossible to determine whether the indicated MIMO transmission scheme configuration will cause the AI ​​model to switch. Therefore, the terminal device can report the generalization capabilities of the one or more AI models to the network device (that is, the terminal device can send the first capability to the network device), so that the network device can determine whether adjusting the MIMO transmission scheme configuration will cause the AI ​​model to switch on the terminal device side based on the generalization capabilities of the AI ​​model reported by the terminal device.

[0162] S420. The network device sends parameter configuration to the terminal device based on the first capability.

[0163] Accordingly, the terminal device can receive parameter configuration from the network device.

[0164] Among them, parameter configuration is used for communication between terminal devices and network devices.

[0165] Exemplarily, the parameter configuration may be a DMRS configuration, and the DMRS configuration may indicate an ID of a DMRS base sequence.

[0166] Exemplarily, the parameter configuration may be a CSI-RS configuration, and the CSI-RS configuration may indicate the number of CSI-RS ports.

[0167] Exemplarily, the parameter configuration may be an MCS configuration and a CSI reporting content configuration, the MCS configuration may indicate an MCS index, and the CSI reporting content configuration may indicate RSRP reporting or CSI reporting.

[0168] In some embodiments, the first AI model corresponding to the value of the parameter configuration is used for communication between the terminal device and the network device.

[0169] Among them, the first AI model belongs to one or more AI models.

[0170] Based on different types of parameter configurations, the network device may determine the parameter configuration based on the first capability, including one or more of the following possible implementations.

[0171] In one possible implementation method, the network device can determine the DMRS configuration based on the range of DMRS configuration corresponding to one or more AI models reported by the terminal device.

[0172] When the terminal device reports that one or more AI models can be used for any DMRS configuration, the network device can determine that the adjustment of the DMRS configuration does not need to consider the switching of the AI ​​model.

[0173] When the terminal device reports that one or more AI models can be used for a certain DMRS configuration (such as a DMRS sequence ID), the network device does not need to consider the switching of the AI ​​model when determining the DMRS configuration.

[0174] When a terminal device reports that one or more AI models are trained for different values ​​of a DMRS configuration (such as the number of DMRS symbols), the network device needs to consider the possible switching of the AI ​​model on the terminal device side when indicating the DMRS configuration. For example, the network device tries to avoid dynamically adjusting the DMRS configuration to avoid dynamic switching of the AI ​​model on the terminal device side, which exceeds the capabilities of the terminal device side (by default, terminal devices do not support dynamic switching of AI models because the deployment of AI models takes a certain amount of time).

[0175] In another possible implementation, the network device may determine the CSI-RS configuration based on the range of the CSI-RS configuration corresponding to one or more AI models reported by the terminal device.

[0176] When the terminal device reports that one or more AI models can be used for any CSI-RS configuration, the network device can determine that the adjustment of the CSI-RS configuration does not need to consider the switching of the AI ​​model.

[0177] When the terminal device reports that one or more AI models can be used for a certain CSI-RS configuration (such as the number of CSI-RS ports), the network device does not need to consider the switching of AI models when determining the CSI-RS configuration.

[0178] When a terminal device reports that one or more AI models are trained for different values ​​of a CSI-RS configuration (such as the number of CSI-RS symbols), the network device needs to consider the possible switching of the AI ​​model on the terminal device side when indicating the CSI-RS configuration. For example, the network device should try to avoid dynamically adjusting the CSI-RS configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0179] In another possible implementation, the network device may determine the CSI reporting content configuration based on the range of the CSI reporting content configuration corresponding to one or more AI models reported by the terminal device.

[0180] When one or more AI models reported by the terminal device can be used for any CSI reporting content configuration, the network device can determine that the adjustment of the CSI reporting content configuration does not need to consider the switching of the AI ​​model.

[0181] When the terminal device reports that one or more AI models can be used for a certain CSI reporting content configuration (such as CSI prediction or CSI compression), the network device does not need to consider the switching of AI models when determining the CSI reporting content configuration.

[0182] When a terminal device reports one or more AI models trained for different values ​​of a CSI reporting content configuration (such as CSI prediction or CSI compression), the network device needs to consider the possible switching of AI models on the terminal device side when indicating the CSI reporting content configuration. For example, the network device should try to avoid dynamically adjusting the CSI reporting content configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0183] In another possible implementation, the network device may determine the MCS configuration based on the range of the MCS configuration corresponding to one or more AI models reported by the terminal device.

[0184] When the terminal device reports that one or more AI models can be used for any MCS configuration, the network device can determine that the adjustment of the MCS configuration does not need to consider the switching of the AI ​​model.

[0185] When the terminal device reports that one or more AI models can be used for a certain MCS configuration (such as modulation mode or bit rate range), the network device does not need to consider the switching of AI models when determining the MCS configuration.

[0186] When a terminal device reports that one or more AI models are trained for different values ​​of a certain MCS configuration (such as modulation mode or bit rate range), the network device needs to consider the possible switching of AI models on the terminal device side when indicating the MCS configuration. For example, the network device should try to avoid dynamically adjusting the MCS configuration to avoid dynamic AI model switching on the terminal device side, which would exceed the capabilities of the terminal device side.

[0187] In another possible implementation, the network device may determine the transmission block size configuration based on the range of transmission block size configurations corresponding to one or more AI models reported by the terminal device.

[0188] When the terminal device reports that one or more AI models can be used for any transmission block size configuration, the network device can determine that the adjustment of the transmission block size configuration does not need to consider the switching of the AI ​​model.

[0189] When the terminal device reports that one or more AI models can be used for a certain transmission block size configuration (such as a transmission block size range), the network device does not need to consider the switching of the AI ​​model when determining the transmission block size configuration.

[0190] When a terminal device reports that one or more AI models are trained for different values ​​of a certain transport block size configuration (such as a transport block size range), the network device needs to consider the possible switching of the AI ​​model on the terminal device side when indicating the transport block size configuration. For example, the network device tries to avoid dynamically adjusting the transport block size configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0191] Another possible implementation method is that the network device can determine the transmission layer configuration based on the range of transmission layer configurations corresponding to one or more AI models reported by the terminal device.

[0192] When the terminal device reports that one or more AI models can be used for any transmission layer configuration, the network device can determine that the adjustment of the transmission layer configuration does not need to consider the switching of the AI ​​model.

[0193] When the terminal device reports that one or more AI models can be used for a certain transmission layer configuration (such as a range of transmission layers), the network device does not need to consider the switching of AI models when determining the transmission layer configuration.

[0194] When a terminal device reports that one or more AI models are trained for different values ​​of a certain transmission layer configuration (such as a transmission layer range), the network device needs to consider the possible switching of AI models on the terminal device side when indicating the transmission layer configuration. For example, the network device should try to avoid dynamically adjusting the transmission layer configuration to avoid dynamic AI model switching on the terminal device side, which would exceed the capabilities of the terminal device side.

[0195] In another possible implementation, the network device may determine the transmission bandwidth configuration based on the range of the transmission bandwidth configuration corresponding to one or more AI models reported by the terminal device.

[0196] When the terminal device reports that one or more AI models can be used for any transmission bandwidth configuration, the network device can determine that the adjustment of the transmission bandwidth configuration does not need to consider the switching of the AI ​​model.

[0197] When the terminal device reports that one or more AI models can be used for a certain transmission bandwidth configuration (such as a transmission bandwidth range), the network device does not need to consider the switching of the AI ​​model when determining the transmission bandwidth configuration.

[0198] When a terminal device reports that one or more AI models are trained for different values ​​of a certain transmission bandwidth configuration (such as a transmission bandwidth range), the network device needs to consider the possible switching of AI models on the terminal device side when indicating the transmission bandwidth configuration. For example, the network device should try to avoid dynamically adjusting the transmission bandwidth configuration to avoid dynamic AI model switching on the terminal device side, which would exceed the capabilities of the terminal device side.

[0199] In another possible implementation, the network device may determine the antenna port number configuration based on the range of antenna port number configurations corresponding to one or more AI models reported by the terminal device.

[0200] When the terminal device reports that one or more AI models can be used for any antenna port number configuration, the network device can determine that the adjustment of the antenna port number configuration does not need to consider the switching of the AI ​​model.

[0201] When the terminal device reports that one or more AI models can be used for a certain antenna port number configuration (such as an antenna port number range), the network device does not need to consider the switching of the AI ​​model when determining the antenna port number configuration.

[0202] When a terminal device reports that one or more AI models are trained for different values ​​of a certain antenna port number configuration (such as an antenna port number range), the network device needs to consider the possible switching of AI models on the terminal device side when indicating the antenna port number configuration. For example, the network device tries to avoid dynamically adjusting the antenna port number configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0203] In another possible implementation, the network device may determine the carrier frequency configuration based on the range of carrier frequency configurations corresponding to one or more AI models reported by the terminal device.

[0204] When the terminal device reports that one or more AI models can be used for any carrier frequency configuration, the network device can determine that the adjustment of the carrier frequency configuration does not need to consider the switching of the AI ​​model.

[0205] When the terminal device reports that one or more AI models can be used for a certain carrier frequency configuration (such as a carrier frequency range), the network device does not need to consider the switching of the AI ​​model when determining the carrier frequency configuration.

[0206] When a terminal device reports that one or more AI models are trained for different values ​​of a certain carrier frequency configuration (such as a carrier frequency range), the network device needs to consider the possible switching of the AI ​​model on the terminal device side when indicating the carrier frequency configuration. For example, the network device tries to avoid dynamically adjusting the carrier frequency configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0207] For example, when the AI ​​model is trained for frequency range 1 (FR1) and FR2 respectively, the network device should try to avoid dynamic switching between FR1 and FR2 to avoid dynamic switching of the AI ​​model on the terminal device side, which exceeds the capability of the terminal device side.

[0208] In another possible implementation, the network device may determine the subcarrier spacing configuration based on the range of subcarrier spacing configuration corresponding to one or more AI models reported by the terminal device.

[0209] When the terminal device reports that one or more AI models can be used for any subcarrier spacing configuration, the network device can determine that the adjustment of the subcarrier spacing configuration does not need to consider the switching of the AI ​​model.

[0210] When the terminal device reports that one or more AI models can be used for a certain subcarrier spacing configuration (such as a subcarrier spacing range), the network device does not need to consider the switching of the AI ​​model when determining the subcarrier spacing configuration.

[0211] When a terminal device reports that one or more AI models are trained for different values ​​of a subcarrier spacing configuration (such as a subcarrier spacing range), the network device needs to consider the possible switching of the AI ​​model on the terminal device side when indicating the subcarrier spacing configuration. For example, the network device tries to avoid dynamically adjusting the subcarrier spacing configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0212] Another possible implementation method is that the network device can determine the TCI state configuration based on the range of TCI state configuration corresponding to one or more AI models reported by the terminal device.

[0213] When the terminal device reports that one or more AI models can be used for any TCI state configuration, the network device can determine that the adjustment of the TCI state configuration does not need to consider the switching of the AI ​​model.

[0214] When the terminal device reports that one or more AI models can be used for a certain TCI state configuration, the network device does not need to consider the switching of the AI ​​model when determining the TCI state configuration.

[0215] When a terminal device reports that one or more AI models are trained for different values ​​of a TCI state configuration (such as a single TCI state and multiple TCI states), the network device needs to consider the possible switching of AI models on the terminal device side when indicating the TCI state configuration. For example, the network device should try to avoid dynamically adjusting the TCI state configuration to avoid dynamic AI model switching on the terminal device side, which would exceed the capabilities of the terminal device side.

[0216] In another possible implementation, the network device may determine the cell ID configuration based on the range of the cell ID configuration corresponding to one or more AI models reported by the terminal device.

[0217] When the terminal device reports that one or more AI models can be used for any cell ID configuration, the network device can determine that the adjustment of the cell ID configuration does not need to consider the switching of the AI ​​model. Therefore, when performing cell ID handover, there is no need to reserve time for the terminal device side to switch the AI ​​model.

[0218] When the terminal device reports that one or more AI models can be used for a certain cell ID configuration (such as a cell ID range), the network device does not need to consider the switching of the AI ​​model when determining the cell ID configuration. Therefore, when switching the cell ID within the cell ID configuration, there is no need to reserve time for switching the AI ​​model on the terminal device side.

[0219] When a terminal device reports that one or more AI models are trained for different values ​​of a cell ID configuration (such as a cell ID range), the network device needs to consider the possible switching of the AI ​​model on the terminal device side when indicating the cell ID configuration. For example, the network device tries to avoid dynamically adjusting the cell ID configuration to avoid dynamic switching of the AI ​​model on the terminal device side, which exceeds the capabilities of the terminal device side, and when switching the cell ID, it is necessary to reserve time for the terminal device side to switch the AI ​​model.

[0220] Another possible implementation method is that the network device can determine the MIMO transmission scheme configuration based on the range of MIMO transmission scheme configuration corresponding to one or more AI models reported by the terminal device.

[0221] When the terminal device reports that one or more AI models can be used for any MIMO transmission scheme configuration, the network device can determine that the adjustment of the MIMO transmission scheme configuration does not need to consider the switching of the AI ​​model.

[0222] When the terminal device reports that one or more AI models can be used for a certain MIMO transmission scheme configuration, the network device does not need to consider the switching of the AI ​​model when determining the MIMO transmission scheme configuration.

[0223] When a terminal device reports that one or more AI models are trained separately for different values ​​of a MIMO transmission scheme configuration (such as a single TRP transmission scheme and a multi-TRP transmission scheme), the network device needs to consider the possible switching of the AI ​​model on the terminal device side when indicating the MIMO transmission scheme configuration. For example, the network device tries to avoid dynamically adjusting the MIMO transmission scheme configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0224] Through this method, when the network device determines the parameter configuration based on the first capability, it can consider the switching of the AI ​​model of the terminal device due to the adjustment of the parameter configuration, thereby reasonably determining the signaling and range used for the adjustment of the parameter configuration. In other words, the network device can determine the appropriate parameter configuration based on the first capability reported by the terminal device, avoiding the terminal device side from switching the AI ​​model that exceeds its own capability (such as dynamic AI model switching), or the performance loss caused by frequent AI model switching, thereby ensuring the performance of uplink or downlink transmission based on the AI ​​model.

[0225] In some embodiments, the method may further include: the terminal device may send a second capability to the network device, where the second capability is used to indicate whether the terminal device supports one or more of the following:

[0226] Switching between AI models based on downlink control information (DCI) signaling;

[0227] Switching between AI models based on Media Access Control (MAC) signaling;

[0228] Switching between AI models is performed based on Radio Resource Control (RRC) signaling.

[0229] Accordingly, the network device may receive the second capability from the terminal device.

[0230] It should be noted that the AI ​​model here generally refers to the AI ​​model used on the terminal device side, which can be one or more of the aforementioned AI models, or other AI models, and the embodiments of the present application do not limit this.

[0231] Exemplarily, the terminal device may support switching between AI models based on RRC signaling, but does not support switching between AI models based on DCI signaling and MAC signaling.

[0232] Exemplarily, the terminal device may support switching between AI models based on RRC signaling and MAC signaling, but does not support switching between AI models based on DCI signaling.

[0233] Exemplarily, the terminal device may support switching between AI models based on DCI signaling, RRC signaling, and MAC signaling.

[0234] Through this method, the terminal device can send the second capability to the network device, and the network device can know whether the terminal device can support the switching of the AI ​​model based on a certain signaling (such as DCI signaling), thereby avoiding the switching of the AI ​​model exceeding the capability of the terminal device and ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0235] In some embodiments, based on the first capability, the network device sends the parameter configuration to the terminal device, which may include: based on the first capability and the second capability, the network device sends the parameter configuration to the terminal device.

[0236] Among them, when the first AI model corresponding to the parameter configuration value and the second AI model corresponding to the first value are different AI models, the switching between the first AI model and the second AI model does not exceed the second capability; the first value is the parameter value before the network device sends the parameter configuration.

[0237] That is to say, the switching of the AI ​​model caused by the parameter configuration does not exceed the second capability sent by the terminal device.

[0238] It should be noted that when the terminal device cannot support AI model switching based on a certain signaling (that is, AI model switching based on the signaling exceeds the second capability), if the network device uses the signaling to configure parameters, the value of the parameter configuration cannot cause the AI ​​model to be switched on the terminal device side.

[0239] For example, if the terminal device does not support switching of AI models based on DCI signaling and the parameter configuration is indicated by DCI signaling, the parameter configuration cannot cause the terminal device to switch the AI ​​model. For example, if the terminal device does not support switching of AI models based on DCI signaling and the CSI reporting content configuration is indicated by DCI signaling, the network device cannot adjust the CSI reporting content configuration through DCI signaling, thereby causing the terminal device to switch the AI ​​model.

[0240] Exemplarily, when the terminal device supports switching of AI models based on RRC signaling and the parameter configuration is indicated by RRC signaling, the parameter configuration may cause the terminal device to switch the AI ​​model (for example, from the second AI model to the first AI model). For example, different CSI-RS configurations correspond to different AI models. When the terminal device supports switching of AI models based on RRC signaling and the CSI-RS configuration is indicated by RRC signaling, the terminal device can switch the second AI model to the first AI model based on the CSI-RS configuration.

[0241] Through this method, the network device can send a parameter configuration to the terminal device, and the parameter configuration causes the AI ​​model switching to not exceed the second capability. After receiving the parameter configuration, the terminal device can avoid the AI ​​model switching exceeding its own capability, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0242] In some embodiments, the method may further include: when the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value are different AI models, the terminal device does not expect the switching between the first AI model and the second AI model to exceed the second capability; the first value is the parameter value before the terminal device receives the parameter configuration.

[0243] That is, the terminal device does not expect the switching of the AI ​​model caused by the parameter configuration to exceed the second capability sent by the terminal device.

[0244] That is to say, the terminal device expects that the switching of the AI ​​model caused by the parameter configuration is within the second capability sent by the terminal device.

[0245] Exemplarily, the terminal device does not expect the switching of the AI ​​model caused by the transport block size configuration to exceed the second capability sent by the terminal device. For example, when the terminal device does not support the switching of the AI ​​model based on DCI signaling, and the transport block size configuration is indicated by DCI signaling, the transport block size configuration cannot cause the terminal device to switch the AI ​​model, and can only switch within the range of transport block sizes corresponding to the same AI model; for another example, when the terminal device does not support the switching of the AI ​​model based on DCI signaling and MAC signaling, and the transport block size configuration is indicated by DCI signaling and MAC signaling, the transport block size configuration cannot cause the terminal device to switch the AI ​​model, and can only switch within the range of transport block sizes corresponding to the same AI model.

[0246] Exemplarily, the terminal device does not expect the switching of the AI ​​model caused by the configuration of the number of transmission layers or the number of antenna ports to exceed the second capability sent by the terminal device. For example, when the terminal device does not support the switching of the AI ​​model based on DCI signaling, and the configuration of the number of transmission layers or the number of antenna ports is indicated by DCI signaling, the configuration of the number of transmission layers or the number of antenna ports cannot cause the terminal device to switch the AI ​​model, and can only switch within the range of the number of transmission layers or the number of antenna ports corresponding to the same AI model; for another example, when the terminal device does not support the switching of the AI ​​model based on DCI signaling and MAC signaling, and the configuration of the number of transmission layers or the number of antenna ports is indicated by DCI signaling and MAC signaling, the configuration of the number of transmission layers or the number of antenna ports cannot cause the terminal device to switch the AI ​​model, and can only switch within the range of the number of transmission layers or the number of antenna ports corresponding to the same AI model.

[0247] Exemplarily, the terminal device does not expect the switching of the AI ​​model caused by the transmission bandwidth configuration to exceed the second capability sent by the terminal device. For example, when the terminal device does not support the switching of the AI ​​model based on DCI signaling, and the transmission bandwidth configuration is indicated by DCI signaling, the transmission bandwidth configuration cannot cause the terminal device to switch the AI ​​model, and can only switch within the transmission bandwidth corresponding to the same AI model; for another example, when the terminal device does not support the switching of the AI ​​model based on DCI signaling and MAC signaling, and the transmission bandwidth configuration is indicated by DCI signaling and MAC signaling, the transmission bandwidth configuration cannot cause the terminal device to switch the AI ​​model, and can only switch within the transmission bandwidth corresponding to the same AI model.

[0248] Exemplarily, the terminal device does not expect the switching of the AI ​​model caused by the MIMO transmission scheme configuration to exceed the second capability sent by the terminal device. For example, when the terminal device does not support the switching of the AI ​​model based on DCI signaling, and the MIMO transmission scheme configuration is indicated by DCI signaling, the MIMO transmission scheme configuration cannot cause the terminal device to switch the AI ​​model, and can only switch within the scope of the MIMO transmission scheme corresponding to the same AI model; for another example, when the terminal device does not support the switching of the AI ​​model based on DCI signaling and MAC signaling, and the MIMO transmission scheme configuration is indicated by DCI signaling and MAC signaling, the MIMO transmission scheme configuration cannot cause the terminal device to switch the AI ​​model, and can only switch within the scope of the MIMO transmission scheme corresponding to the same AI model.

[0249] In the above method, if the switching of the AI ​​model caused by the parameter configuration exceeds the second capability sent by the terminal device, the terminal device can use the previous AI model for communication with the network device without switching the AI ​​model; or, the terminal device can communicate with the network device without using the AI ​​model, but instead use traditional non-AI methods to communicate; or, the terminal device can regard this configuration as an error configuration (Error Configuration) and thus not communicate with the network device.

[0250] Through this method, the terminal device does not expect the switching of the AI ​​model caused by the parameter configuration to exceed the second capability sent by the terminal device. Therefore, when the terminal device switches the AI ​​model based on the parameter configuration, it can avoid the switching of the AI ​​model exceeding its own capability and ensure the uplink and downlink transmission performance based on the AI ​​model.

[0251] In some embodiments, when the parameter configuration is indicated through DCI signaling and / or MAC signaling, the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value are the same AI model.

[0252] The first value is the parameter value before the network device sends the parameter configuration.

[0253] It should be noted that when parameter configuration is indicated via DCI signaling and / or MAC signaling (such as dynamic indication), the parameter configuration cannot cause the terminal device to switch the AI ​​model based on DCI signaling and / or MAC signaling. In other words, the parameter value indicated by the parameter configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model.

[0254] Among them, the network equipment can correspond to different AI models through parameter configuration indicated by RRC signaling.

[0255] Exemplarily, when the DMRS configuration is indicated by DCI signaling and / or MAC signaling, the DMRS configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. That is, the parameter value indicated by the DMRS configuration and the original parameter value need to correspond to the same AI model. For example, when the terminal device reports different AI models corresponding to different DMRS port sets, the DMRS port indicated by the network device through DCI signaling and / or MAC signaling needs to correspond to the port in the port set of the same AI model, and the port sets corresponding to different AI models can be switched through high-level signaling (such as RRC signaling).

[0256] Exemplarily, when the CSI-RS configuration is indicated by DCI signaling and / or MAC signaling (for example, non-periodic CSI reporting is triggered by DCI, different CSI reports correspond to different CSI-RS resource configurations, and the CSI-RS resource configuration is determined based on DCI), the CSI-RS configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. In other words, the parameter value indicated by the CSI-RS configuration and the original parameter value need to correspond to the same AI model. For example, when the terminal device reports different AI models corresponding to different CSI-RS base sequences, the CSI-RS base sequence indicated by the network device through DCI signaling and / or MAC signaling needs to correspond to the same AI model, and the CSI-RS base sequences corresponding to different AI models can be switched through high-layer signaling (such as RRC signaling).

[0257] Exemplarily, when the CSI reporting content configuration is indicated by DCI signaling and / or MAC signaling (for example, non-periodic CSI reporting is triggered by DCI, different CSI reports correspond to different CSI reporting contents, and the CSI reporting content is determined based on DCI), the CSI reporting content configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. In other words, the parameter value indicated by the CSI reporting content configuration and the original parameter value need to correspond to the same AI model. For example, when the terminal device reports different AI models for RSRP reporting and CSI reporting respectively, the CSI reporting content configuration indicated by the network device through DCI signaling and / or MAC signaling needs to correspond to the same AI model, and the CSI reporting content configurations corresponding to different AI models can be switched through high-layer signaling (such as RRC signaling).

[0258] Exemplarily, when the MCS configuration or transport block size configuration is indicated through DCI signaling and / or MAC signaling, the MCS configuration or transport block size configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. That is, the parameter value indicated by the MCS configuration or transport block size configuration and the original parameter value need to correspond to the same AI model. For example, when the terminal device reports different AI models for different modulation modes, the modulation mode indicated by the network device through DCI signaling and / or MAC signaling needs to correspond to the same AI model, and the modulation modes corresponding to different AI models can be switched through high-layer signaling (such as RRC signaling).

[0259] Exemplarily, when the number of transmission layers configuration or the number of antenna ports configuration is indicated by DCI signaling and / or MAC signaling, the number of transmission layers configuration or the number of antenna ports configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. That is, the parameter value indicated by the number of transmission layers configuration or the number of antenna ports configuration and the original parameter value need to correspond to the same AI model. For example, when the terminal device reports different AI models for different numbers of antenna ports, the number of antenna ports indicated by the network device through DCI signaling and / or MAC signaling needs to correspond to the same AI model, and the number of antenna ports corresponding to different AI models can be switched through high-layer signaling (such as RRC signaling).

[0260] Exemplarily, when the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration is indicated by DCI signaling and / or MAC signaling, the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. That is, the parameter value indicated by the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration and the original parameter value need to correspond to the same AI model. For example, when the terminal device reports different AI models for different transmission bandwidths, the transmission bandwidth indicated by the network device through DCI signaling and / or MAC signaling needs to correspond to the same AI model, and the transmission bandwidth corresponding to different AI models can be switched through high-level signaling (such as RRC signaling).

[0261] Exemplarily, when the MIMO transmission scheme configuration is indicated by DCI signaling and / or MAC signaling, the MIMO transmission scheme configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. That is, the parameter value indicated by the MIMO transmission scheme configuration and the original parameter value need to correspond to the same AI model. For example, when the terminal device reports different AI models for a single TRP transmission scheme and a multi-TRP collaboration scheme, the MIMO transmission scheme configuration indicated by the network device through DCI signaling and / or MAC signaling needs to correspond to the same AI model, and the MIMO transmission scheme configurations corresponding to different AI models can be switched through high-layer signaling (such as RRC signaling).

[0262] Through this method, when the parameter configuration is indicated through DCI signaling and / or MAC signaling, the first AI model and the second AI model corresponding to the first value are the same AI model, which can prevent the AI ​​model from switching beyond its own capabilities and ensure the uplink and downlink transmission performance based on the AI ​​model.

[0263] In some embodiments, the method may further include: when the parameter configuration is indicated through DCI signaling and / or MAC signaling, the terminal device does not expect the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value to be different AI models.

[0264] That is to say, when the parameter configuration is indicated through DCI signaling and / or MAC signaling, the terminal device expects that the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value are the same AI model.

[0265] Among them, the first value is the parameter value before the terminal device receives the parameter configuration.

[0266] For example, when the DMRS configuration is indicated by DCI signaling and / or MAC signaling, the terminal device does not expect the parameter value indicated by the DMRS configuration to correspond to a different AI model than the original parameter value (i.e., the first value). In other words, the DMRS configuration cannot cause the terminal device to switch the AI ​​model based on DCI signaling and / or MAC signaling. The DMRS configuration indicated by the network device through RRC signaling can correspond to different AI models.

[0267] For example, when the terminal device reports different DMRS port sets corresponding to different AI models, if the network device indicates the DMRS configuration through DCI signaling, the terminal device can dynamically switch between DMRSs under the same DMRS port set, but cannot dynamically switch between DMRSs under different DMRS port sets.

[0268] For example, when the CSI-RS configuration is indicated by DCI signaling and / or MAC signaling, the terminal device does not expect the parameter value indicated by the CSI-RS configuration to correspond to a different AI model than the original parameter value (i.e., the first value). In other words, the CSI-RS configuration cannot cause the terminal device to switch the AI ​​model based on DCI signaling and / or MAC signaling. Among them, the CSI-RS configuration indicated by the network device through RRC signaling can correspond to different AI models.

[0269] For example, when the terminal reports different AI models for different numbers of CSI-RS ports, if the network device triggers non-periodic CSI reporting through DCI, the number of CSI-RS ports used for CSI reporting cannot be dynamically adjusted (for example, multiple CSI reports triggered at the same time use different numbers of CSI-RS ports).

[0270] For example, when the CSI reporting content configuration is indicated by DCI signaling and / or MAC signaling, the terminal device does not expect the parameter value indicated by the CSI reporting content configuration to correspond to a different AI model than the original parameter value (i.e., the first value). In other words, the CSI reporting content configuration cannot cause the terminal device to switch the AI ​​model based on DCI signaling and / or MAC signaling. Among them, the CSI reporting content configuration indicated by the network device through RRC signaling can correspond to different AI models.

[0271] For example, when the terminal device uses different AI models to report different CSI reporting contents, if the network device triggers non-periodic CSI reporting through DCI signaling, the CSI reporting content of the CSI report cannot be dynamically adjusted (for example, multiple CSI reports triggered at the same time use different CSI reporting contents).

[0272] Exemplarily, when the MCS configuration or transport block size configuration is indicated by DCI signaling and / or MAC signaling, the terminal device does not expect the parameter value indicated by the MCS configuration or transport block size configuration to correspond to a different AI model than the original parameter value (i.e., the first value). In other words, the MCS configuration or transport block size configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. Among them, the MCS configuration or transport block size configuration indicated by the network device through RRC signaling can correspond to different AI models.

[0273] For example, when the terminal device reports different modulation modes and uses different AI models, if the network device indicates the MCS through DCI signaling, the terminal device can dynamically switch between MCSs under the same modulation mode, but cannot dynamically switch between MCSs of different modulation modes.

[0274] For example, when the antenna port number configuration or the transmission layer number configuration is indicated by DCI signaling and / or MAC signaling, the terminal device does not expect the parameter value indicated by the antenna port number configuration or the transmission layer number configuration to correspond to a different AI model than the original parameter value (i.e., the first value). In other words, the antenna port number configuration or the transmission layer number configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. Among them, the antenna port number configuration or the transmission layer number configuration indicated by the network device through RRC signaling can correspond to different AI models.

[0275] For example, when a terminal device reports different numbers of antenna ports and uses different AI models, if the network device indicates the number of antenna ports through DCI signaling, the terminal device cannot dynamically switch between different numbers of antenna ports.

[0276] Exemplarily, when the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration is indicated by DCI signaling and / or MAC signaling, the terminal device does not expect the parameter value indicated by the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration to correspond to a different AI model than the original parameter value (i.e., the first value). In other words, the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration cannot cause the terminal device side to switch the AI ​​model based on DCI signaling and / or MAC signaling. Among them, the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration indicated by the network device through RRC signaling can correspond to different AI models.

[0277] For example, when the terminal device reports different AI models for different subcarrier spacings, if the network device indicates the switching of the partial bandwidth (Bandwidth Part, BWP) through DCI signaling (different BWPs may use different subcarrier spacings), the terminal device cannot dynamically switch between different subcarrier spacings, that is, the indicated BWP and the original BWP need to use the same subcarrier spacing.

[0278] For example, when the MIMO transmission scheme configuration is indicated via DCI signaling and / or MAC signaling, the terminal device does not expect the MIMO transmission scheme indicated by the parameter configuration to correspond to a different AI model than the original MIMO transmission scheme. In other words, the MIMO transmission scheme configuration cannot cause the terminal device to switch the AI ​​model based on DCI signaling and / or MAC signaling. The MIMO transmission scheme configuration indicated by the network device via RRC signaling can correspond to different AI models.

[0279] For example, when the terminal device reports that open-loop transmission and closed-loop transmission use different AI models, if the network device indicates the MIMO transmission scheme through DCI signaling, dynamic switching between open-loop transmission and closed-loop transmission cannot be performed.

[0280] In some embodiments, the terminal device can determine whether to perform a cell handover and whether the cell handover will cause an AI model switch based on the cell ID configured by the network device. If the cell handover may cause an AI model switch, the terminal device needs to reserve time for the AI ​​model switch; otherwise, the terminal device can still use the original AI model for data transmission or reception.

[0281] In the above method, if the first AI model and the second AI model corresponding to the first value are different AI models, the terminal device can use the previous second AI model to communicate with the network device without switching the AI ​​model; or, the terminal device can communicate with the network device without using the AI ​​model, but instead use a traditional non-AI method to communicate; or, the terminal device can regard this configuration as an incorrect configuration and thus not communicate with the network device.

[0282] Through this method, when the parameter configuration is indicated through DCI signaling and / or MAC signaling, the terminal device does not expect that the first AI model and the second AI model corresponding to the first value are different AI models, thereby avoiding the switching of the AI ​​model exceeding its own capabilities and ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0283] S430: The terminal device communicates with the network device based on the parameter configuration.

[0284] In some embodiments, the terminal device communicates with the network device based on the parameter configuration, which may include: the terminal device communicates with the network device based on the first AI model corresponding to the value of the parameter configuration.

[0285] Among them, the first AI model belongs to one or more AI models.

[0286] In some embodiments, the number of first AI models corresponding to the value of the parameter configuration is one.

[0287] For example, the parameter values ​​indicated by the DMRS configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for channel estimation or data monitoring.

[0288] Exemplarily, the parameter values ​​indicated by the CSI-RS configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for CSI feedback.

[0289] For example, the CSI reporting content configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for CSI feedback.

[0290] For example, the MCS configuration or the transport block size configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for data detection.

[0291] For example, the configuration of the number of transmission layers or the number of antenna ports cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for data detection.

[0292] For example, the transmission bandwidth configuration, subcarrier spacing configuration or carrier frequency configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for data detection.

[0293] For example, the TCI state configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for data detection.

[0294] For example, the cell ID configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for data detection.

[0295] For example, the network device cannot instruct two signals to be sent or received at the same time, and the MIMO transmission schemes used by the two signals correspond to different AI models, otherwise the terminal device cannot use two AI models to send or receive signals at the same time.

[0296] Through this method, when the number of first AI models is one, the terminal device can know which AI model to use to perform the corresponding function (such as performing CSI feedback), thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0297] In some embodiments, the terminal device does not expect the first AI model corresponding to the value of the parameter configuration to include at least two different AI models.

[0298] That is to say, the terminal device expects that the number of first AI models corresponding to the value of the parameter configuration is one.

[0299] For example, at least for terminal devices that do not have multi-model capabilities, the terminal device does not expect the parameter values ​​indicated by the DMRS configuration to correspond to two different AI models at the same time. Otherwise, the terminal device cannot determine which AI model to use for channel estimation or data detection. For example, when different DMRS port sets correspond to different AI models, the network device cannot indicate ports in two DMRS port sets at the same time.

[0300] For example, at least for terminal devices that do not have multi-model capabilities, the terminal device does not expect the parameter values ​​indicated by the CSI-RS configuration to correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for CSI feedback. For example, when different AI models are used for different numbers of CSI-RS ports, when the network device activates semi-continuous CSI reporting through MAC signaling, the multiple CSI-RS resources used for the same CSI reporting measurement cannot correspond to different AI models (such as the bandwidth of one CSI-RS resource is greater than a first threshold, and the bandwidth of another CSI-RS resource is less than the first threshold. The first threshold can be a predefined parameter value or a parameter value set in other ways. This is not limited in the embodiments of the present application).

[0301] For example, at least for terminal devices that do not have multi-model capabilities, the terminal device does not expect the CSI reporting content indicated by a CSI reporting content configuration to correspond to two different AI models at the same time. Otherwise, the terminal device cannot determine which AI model to use for CSI feedback. For example, when the terminal device reports different CSI reporting contents using different AI models, when the network device activates semi-persistent CSI reporting through MAC signaling, multiple CSI reporting contents used for the same CSI reporting measurement cannot correspond to different AI models.

[0302] For example, at least for terminal devices that do not have multi-model capabilities, the terminal device does not expect the parameter values ​​indicated by an MCS configuration or transport block size configuration to correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for data detection. For example, when the terminal device reports different code rate ranges (for example, greater than 1 / 3 and less than 1 / 3) using different AI models, when the network device indicates the MCS of two codewords through DCI, the MCS of the two codewords cannot indicate the code rate ranges corresponding to different AI models (for example, the code rate of one AI model is greater than 1 / 3 and the code rate of the other AI model is less than 1 / 3).

[0303] For example, at least for terminal devices that do not have multi-model capabilities, the terminal device does not expect the parameter values ​​indicated by an antenna port number configuration or a transmission layer number configuration to correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for data detection. For example, in the case where different AI models are used for different numbers of antenna ports, if multiple DCIs simultaneously schedule the terminal device for uplink or downlink transmission, the multiple DCIs cannot simultaneously indicate two different numbers of antenna ports; in the case where different AI models are used for different numbers of transmission layers, if multiple DCIs simultaneously schedule the terminal device for uplink or downlink transmission, the multiple DCIs cannot simultaneously indicate two different numbers of transmission layers.

[0304] For example, at least for terminal devices that do not have multi-model capabilities, the terminal device does not expect the parameter values ​​indicated by the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration to correspond to two different AI models at the same time. Otherwise, the terminal device cannot determine which AI model to use for data detection. For example, when different transmission bandwidth ranges correspond to different AI models, the network device cannot indicate two transmission bandwidth ranges at the same time.

[0305] For example, at least for terminal devices that do not have multi-model capabilities, the terminal device does not expect the parameter values ​​indicated by the TCI state configuration to correspond to two different AI models at the same time. Otherwise, the terminal device cannot determine which AI model to use for data detection. For example, when different AI models are used for single TCI state and multi-TCI state respectively, the network device cannot indicate single TCI state and multi-TCI state at the same time.

[0306] For example, at least for terminal devices that do not have multi-model capabilities, the terminal device does not expect the parameter values ​​indicated by the cell ID configuration to correspond to two different AI models at the same time. Otherwise, the terminal device cannot determine which AI model to use for data detection. For example, if different cell ID ranges correspond to different AI models, the network device cannot indicate two cell ID ranges at the same time.

[0307] For example, at least for terminal devices that do not have multi-model capabilities, the terminal device does not expect the network device to instruct two signals to be sent or received at the same time, and the MIMO transmission schemes used by the two signals correspond to different AI models, otherwise the terminal device cannot determine which AI model to use for sending or receiving these signals. For example, the network device cannot instruct the terminal device to receive two physical downlink shared channels (PDSCH) at the same time. These two PDSCHs use a single TRP transmission scheme and a multi-TRP transmission scheme, respectively, and they correspond to different AI models.

[0308] In the above method, if the parameter configuration of the network device corresponds to two different AI models, the terminal device may not use the AI ​​model to communicate with the network device, but instead use traditional non-AI methods to communicate with the network device; or, the terminal device may regard this configuration as an incorrect configuration and thus not communicate with the network device.

[0309] Through this method, the terminal device does not expect the first AI model to include at least two different AI models, so that the terminal device can know which AI model to use to perform the corresponding function (such as performing CSI feedback), thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0310] An embodiment of the present application provides a communication method, in which a terminal device can send a first capability to a network device, and the first capability is used to characterize the range of parameter configurations corresponding to one or more AI models; the terminal device can receive the parameter configuration from the network device, and communicate with the network device based on the first AI model corresponding to the parameter configuration; wherein the first AI model belongs to one or more AI models. In this way, the terminal device sends the first capability to the network device, so that the network device can reasonably configure the parameters, avoid the terminal device switching the AI ​​model based on the parameter configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and avoid frequent AI model switching caused by parameter configuration as much as possible, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0311] Based on S410, when the functions of one or more AI models are different, the representation methods of the first capability may include one or more of the following.

[0312] Method #A: When one or more AI models are used for channel estimation or data reception, the first capability can represent the range of DMRS configuration corresponding to the one or more AI models.

[0313] In some embodiments, one or more AI models may be used for downlink channel estimation or for downlink data reception. For example, when one or more AI models are used for downlink channel estimation, the one or more AI models input a received signal and / or a pilot sequence and output a channel estimation result. When one or more AI models are used for downlink data reception, the one or more AI models input a received signal and / or a pilot sequence and output detected data bits.

[0314] In some embodiments, the scope of the DMRS configuration corresponding to one or more AI models may include one or more of the following:

[0315] One AI model corresponds to all DMRS configurations;

[0316] Different AI models correspond to different DMRS base sequences;

[0317] Different AI models correspond to different DMRS CDM groups;

[0318] Different AI models correspond to different DMRS port multiplexing methods;

[0319] Different AI models correspond to different numbers of DMRS symbols;

[0320] Different AI models correspond to different DMRS port sets.

[0321] It should be noted that, when the number of the one or more AI models is one, the one AI model can correspond to all DMRS configurations.

[0322] Exemplarily, the one AI model may correspond to all DMRS base sequences, all DMRS CDM groups, all DMRS port multiplexing modes, all DMRS symbol numbers, all DMRS port sets, and the like.

[0323] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0324] It should be noted that, when the number of the one or more AI models is multiple, the multiple AI models may be different AI models, and each of the multiple AI models may correspond to a partial DMRS configuration.

[0325] Exemplarily, different AI models may correspond to different DMRS base sequences. For example, different AI models may correspond to different DMRS sequence IDs.

[0326] For example, different AI models may correspond to different DMRS CDM groups. For example, a CDM multiplexing mode is used within a DMRS CDM group, and a frequency division multiplexing (FDM) / time division multiplexing (TDM) multiplexing mode is used between different DMRS CDM groups.

[0327] For example, different AI models may correspond to different DMRS port multiplexing modes, which may be CDM / FDM / TDM.

[0328] For example, different AI models may correspond to different DMRS port sets. For example, DMRS ports may be divided into several DMRS port sets, and different AI models may correspond to different DMRS port sets. Ports within a DMRS port set use the same AI model. For another example, each AI model may correspond to a DMRS port, and different AI models may correspond to different DMRS ports.

[0329] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0330] It should be noted that, when the first capability represents the range of DMRS configurations corresponding to one or more AI models, the first capability (for example, 2 bits) can indicate one or more of per UE (that is, all DMRS configurations share one AI model), per Sequence (that is, one AI model corresponds to one DMRS base sequence, and different DMRS base sequences use different AI models), and per Port (that is, one AI model corresponds to one DMRS port, and different DMRS ports use different AI models); or, the first capability (for example, 1 bit) can indicate whether all DMRS configurations share one AI model.

[0331] It should be understood that the terminal device can report the range of DMRS configuration corresponding to one or more AI models to the network device through a bitmap. Each bit reported by the terminal device represents different values ​​for the DMRS configuration, whether the terminal device uses the same AI model, that is, whether the adjustment of the DMRS configuration requires the terminal device to switch the AI ​​model; one bit corresponds to one DMRS configuration.

[0332] Exemplarily, the terminal device can report 3 bits of information, where the first bit indicates whether the terminal device uses the same AI model for different DMRS base sequences, the second bit indicates whether the terminal device uses the same AI model for different DMRS symbol numbers, and the third bit indicates whether the terminal device uses the same AI model for different DMRS port sets. 0 indicates using different AI models, and 1 indicates using the same AI model; or, 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 3-bit information reported by the terminal device, the network device can determine whether the adjustment of a DMRS configuration (such as DMRS Sequence ID or DMRS symbol number) will cause the terminal device side to switch the AI ​​model.

[0333] Through this method, when the first capability represents the range of DMRS configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the DMRS configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the DMRS configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0334] Method #B: When one or more AI models are used for CSI measurement, the first capability can represent the range of CSI-RS configuration corresponding to the one or more AI models, and / or the range of CSI reporting content configuration corresponding to the one or more AI models.

[0335] In some embodiments, one or more AI models can be used for CSI measurement. For example, one or more AI models input a downlink received signal and / or a CSI-RS sequence and output a CSI measurement result, such as RSRP or Precoding Matrix Indicator (PMI) bits.

[0336] In some embodiments, the range of CSI-RS configurations corresponding to one or more AI models may include one or more of the following:

[0337] One AI model corresponds to all CSI-RS configurations;

[0338] Different AI models correspond to different CSI-RS base sequences;

[0339] Different AI models correspond to different sets of CSI-RS port numbers;

[0340] Different AI models correspond to different CSI-RS port multiplexing methods;

[0341] Different AI models correspond to different numbers of CSI-RS symbols;

[0342] Different AI models correspond to different CSI-RS transmission bandwidths.

[0343] It should be noted that, when the number of the one or more AI models is one, the one AI model can correspond to all CSI-RS configurations.

[0344] For example, this AI model can correspond to all CSI-RS base sequences, all CSI-RS port number sets, all CSI-RS port multiplexing modes, all CSI-RS symbol numbers, all CSI-RS transmission bandwidths, and the like.

[0345] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0346] It should be noted that, when there are multiple one or more AI models, the multiple AI models may be different AI models, and each of the multiple AI models may correspond to a partial CSI-RS configuration.

[0347] Exemplarily, different AI models may correspond to different CSI-RS base sequences. For example, different AI models may correspond to different CSI-RS sequence IDs.

[0348] For example, different AI models may correspond to different sets of CSI-RS port numbers. For example, one AI model may correspond to {2, 4, 8} CSI-RS ports, another AI model may correspond to {12, 16, 20, 24} CSI-RS ports, and another AI model may correspond to {32, 64, 128} CSI-RS ports. For another example, each AI model may correspond to a specific number of CSI-RS ports, and different AI models may correspond to different numbers of CSI-RS ports.

[0349] For example, different AI models may correspond to different CSI-RS port multiplexing modes, which may be CDM / FDM / TDM.

[0350] For example, different AI models may correspond to different numbers of CSI-RS symbols, which may be configured through RRC signaling.

[0351] Exemplarily, different AI models may correspond to different CSI-RS transmission bandwidths. For example, one AI model may correspond when the CSI-RS transmission bandwidth is less than a second threshold, and another AI model may correspond when the CSI-RS transmission bandwidth is greater than or equal to the second threshold.

[0352] The second threshold may be a predefined parameter value or a parameter value set in other ways, and this embodiment of the present application does not limit this.

[0353] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0354] It should be noted that, when the first capability represents the configuration range of the CSI-RS corresponding to one or more AI models, the first capability (e.g., 2 bits) can indicate one or more of per UE (i.e., all CSI-RS configurations share one AI model), per Sequence (i.e., one AI model corresponds to one CSI-RS base sequence, and different CSI-RS base sequences use different AI models), and per Port Number (i.e., one AI model corresponds to one CSI-RS port number, and different CSI-RS port numbers use different AI models); alternatively, the first capability (e.g., 1 bit) can indicate whether all CSI-RS configurations share one AI model.

[0355] It should be understood that the terminal device can report the range of CSI-RS configurations corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents different values ​​for the CSI-RS configuration, whether the terminal device uses the same AI model, that is, whether the adjustment of the CSI-RS configuration requires the terminal device to switch the AI ​​model; one bit corresponds to one CSI-RS configuration.

[0356] Exemplarily, the terminal device can report 3 bits of information, where the first bit indicates whether the terminal device uses the same AI model for different CSI-RS base sequences, the second bit indicates whether the terminal device uses the same AI model for different numbers of CSI-RS ports, and the third bit indicates whether the terminal device uses the same AI model for different CSI-RS transmission bandwidths. 0 indicates using different AI models, and 1 indicates using the same AI model; or, 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 3-bit information reported by the terminal device, the network device can determine whether the adjustment of a certain CSI-RS configuration (such as the number of CSI-RS ports or the CSI-RS transmission bandwidth) will cause the terminal device side to switch the AI ​​model.

[0357] Through this method, when the first capability represents the range of CSI-RS configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the CSI-RS configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the CSI-RS configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0358] In some embodiments, the scope of the CSI reporting content configuration corresponding to one or more AI models may include one or more of the following:

[0359] An AI model is used for RSRP reporting and CSI reporting;

[0360] Different AI models are used for RSRP reporting and CSI reporting respectively;

[0361] An AI model for CSI compression and CSI prediction;

[0362] Different AI models are used for CSI compression and CSI prediction respectively;

[0363] An AI model is used for CSI compression, CSI prediction, and CSI prediction and compression;

[0364] Different AI models are used for CSI compression, CSI prediction, and CSI prediction and compression respectively.

[0365] It should be noted that, when the number of the one or more AI models is one, the one AI model can be used for RSRP reporting and CSI reporting; and / or, the one AI model can be used for CSI compression, CSI prediction, and CSI prediction and compression.

[0366] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0367] It should be noted that, when there are multiple one or more AI models, the multiple AI models may be different AI models, and each of the multiple AI models may correspond to part of the CSI reporting content.

[0368] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0369] It should be noted that, when the first capability represents the range of CSI reporting content configuration corresponding to one or more AI models, the first capability (e.g., 1 bit) can indicate whether CSI compression and CSI prediction in CSI reporting use the same AI model; and / or, the first capability (e.g., 1 bit) can indicate whether RSRP reporting and CSI reporting use the same AI model; and / or, the first capability (e.g., 1 bit) can indicate whether CSI compression, CSI prediction, and CSI prediction and compression use the same AI model.

[0370] It should be understood that the terminal device can report the range of the CSI reporting content configuration corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents different values ​​for the CSI reporting content configuration, whether the terminal device uses the same AI model, that is, whether the adjustment of the CSI reporting content configuration requires the terminal device to switch the AI ​​model; one bit corresponds to one CSI reporting content configuration.

[0371] For example, the terminal device can report 2 bits of information, where the first bit indicates whether different AI models are used for RSRP reporting and CSI reporting, respectively, and the second bit indicates whether different AI models are used for CSI compression and CSI prediction, respectively. 0 indicates using different AI models, and 1 indicates using the same AI model; or, 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 2-bit information reported by the terminal device, the network device can determine whether an adjustment to a certain CSI reporting content configuration will cause the terminal device to switch the AI ​​model.

[0372] Through this method, when the first capability represents the range of CSI reporting content configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the CSI reporting content configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the CSI reporting content configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0373] Method #C: When one or more AI models are used for data transmission or data reception, the first capability may represent one or more of the following:

[0374] The scope of the MCS configuration corresponding to one or more AI models;

[0375] The range of transport block size configurations corresponding to one or more AI models;

[0376] The range of transmission layer configurations corresponding to one or more AI models;

[0377] The transmission bandwidth configuration range corresponding to one or more AI models;

[0378] The range of antenna port configurations corresponding to one or more AI models;

[0379] The range of carrier frequency configurations corresponding to one or more AI models;

[0380] The range of subcarrier spacing configuration corresponding to one or more AI models;

[0381] The scope of the TCI state configuration corresponding to one or more AI models;

[0382] The range of cell ID configurations corresponding to one or more AI models.

[0383] In some embodiments, one or more AI models can be used for data transmission or data reception. Data reception includes channel decoding and / or demodulation, while data transmission includes channel coding and / or modulation. For example, for downlink transmission, one or more AI models input the downlink received signal and output data detection results, such as the source bitstream; for uplink transmission, one or more AI models input the data to be transmitted and output the modulated and coded transmit signal.

[0384] In some embodiments, the scope of the MCS configuration corresponding to one or more AI models includes one or more of the following:

[0385] One AI model is used for all MCSs;

[0386] Different AI models are used for different MCSs;

[0387] MCS index range applicable to different AI models;

[0388] One AI model is used for all modulation methods;

[0389] Different AI models are used for different modulation methods;

[0390] A set of modulation methods suitable for different AI models;

[0391] One AI model is used for all bitrate ranges;

[0392] The applicable bitrate range for different AI models.

[0393] It should be noted that when the number of the one or more AI models is one, the one AI model can be used for all MCSs; and / or, the one AI model can be used for all modulation modes; and / or, the one AI model can be used for all bit rate ranges.

[0394] For example, one AI model can be used for all MCSs. For example, the same AI model is used on the terminal device side for all MCSs supported by the protocol. Each MCS corresponds to a modulation mode and a bit rate.

[0395] For example, one AI model can be used for all modulation schemes, for example, one AI model can be used for Quadrature Phase Shift Keying (QPSK), 16-bit Quadrature Amplitude Modulation (QAM), and 64QAM.

[0396] For example, one AI model can be used for all bitrate ranges. For example, the same AI model is used on the terminal device side for all bitrate ranges supported by the protocol.

[0397] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0398] It should be noted that, when there are multiple one or more AI models, the multiple AI models may be different AI models, and each of the multiple AI models may correspond to a partial MCS configuration.

[0399] For example, different AI models can be used for different MCSs. For example, different AI models are used on the terminal device side for different MCSs supported by the protocol.

[0400] For example, different AI models may have applicable MCS index ranges. For example, an MCS index range of 0-12 may correspond to one AI model, and an MCS index range of 12-27 may correspond to another AI model.

[0401] For example, different AI models can be used for different modulation schemes, such as QPSK, 16QAM, 64QAM, and 256QAM.

[0402] For example, different AI models are applicable to different modulation schemes. For example, one model supports QPSK and 16QAM data transmission and reception, while another AI model supports 64QAM and 256QAM data transmission and reception.

[0403] For example, different AI models are applicable to bitrate ranges. For example, one AI model is applicable to a bitrate less than a third threshold (such as 1 / 3), while another AI model is applicable to a bitrate greater than the third threshold.

[0404] Among them, the third threshold can be a predefined parameter value or a parameter value set in other ways, and the embodiment of the present application does not limit this.

[0405] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0406] It should be noted that, when the first capability represents the range of MCS configurations corresponding to one or more AI models, the first capability (for example, 2 bits) can indicate one or more of per UE (ie, all MCS configurations share one AI model), per Modulation Order (ie, one AI model corresponds to one modulation mode, different AI models correspond to different modulation modes), and per MCS (ie, one AI model corresponds to one MCS, and different AI models correspond to different MCSs).

[0407] Exemplarily, the first capability may be indicated in one or more of the following ways:

[0408] 1-bit information, used to indicate whether different MCSs use the same AI model;

[0409] 1-bit information, used to indicate whether different modulation modes use the same AI model;

[0410] 1-bit information, used to indicate whether different bit rates use the same AI model;

[0411] N1 bit information is used to indicate the MCS index range supported by different AI models;

[0412] N2 bits of information are used to indicate the modulation modes supported by different AI models;

[0413] N3 bits of information are used to indicate the bit rate ranges supported by different AI models.

[0414] It should be understood that the terminal device can report the range of MCS configurations corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents different values ​​for the MCS configuration and whether the terminal device uses the same AI model, that is, whether the adjustment of the MCS configuration requires the terminal device to switch the AI ​​model; one bit corresponds to one MCS configuration.

[0415] Exemplarily, the terminal device can report 2 bits of information, where the first bit indicates whether the terminal device uses the same AI model for different modulation modes, and the second bit indicates whether the terminal device uses the same AI model for different bit rate ranges. 0 indicates using different AI models, and 1 indicates using the same AI model; or, 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 2-bit information reported by the terminal device, the network device can determine whether the adjustment of a certain MCS configuration (such as modulation mode or bit rate range) will cause the terminal device side to switch the AI ​​model.

[0416] Through this method, when the first capability represents the range of MCS configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the MCS configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the MCS configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and reducing the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0417] In some embodiments, the range of transport block size configurations corresponding to one or more AI models may include one or more of the following:

[0418] One AI model is used for all transport block sizes;

[0419] Different AI models are used for different transmission block sizes;

[0420] The applicable transmission block size range for different AI models.

[0421] It should be noted that, when the number of the one or more AI models is one, the one AI model can be used for all transmission block sizes.

[0422] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0423] It should be noted that, if there are multiple AI models, the multiple AI models may be different AI models, and the different AI models may be applicable to different transport block size ranges. For example, one AI model may be applicable to a transport block size greater than the fourth threshold, while another AI model may be applicable to a transport block size less than the fourth threshold.

[0424] Among them, the fourth threshold can be a predefined parameter value or a parameter value set in other ways, and the embodiment of the present application does not limit this.

[0425] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0426] It should be noted that, when the first capability represents the range of transport block size configurations corresponding to one or more AI models, the first capability may indicate whether different transport block sizes use the same AI model; and / or, the first capability may indicate the range of transport block sizes applicable to different AI models.

[0427] It should be understood that the terminal device can report the range of transmission block size configurations corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents a different value for the transmission block size configuration, and whether the terminal device uses the same AI model, that is, whether the adjustment of the transmission block size configuration requires the terminal device to switch the AI ​​model.

[0428] For example, a terminal device may report 1 bit of information, which may indicate whether the terminal device uses the same AI model for different transport block sizes. A value of 0 indicates using different AI models, and a value of 1 indicates using the same AI model; or, alternatively, a value of 0 indicates using the same AI model, and a value of 1 indicates using different AI models. Based on the 1 bit of information reported by the terminal device, the network device can determine whether an adjustment to the transport block size configuration will result in an AI model switch on the terminal device side.

[0429] Through this method, when the first capability represents the range of transmission block size configurations corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the transmission block size configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the transmission block size configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and reducing the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0430] In some embodiments, one or more AI models can be used for data transmission or data reception. For downlink transmission, one or more AI models input the downlink received signal and output data detection results, such as the source bitstream. For uplink transmission, one or more AI models input the data to be transmitted and output the transmit signal for each antenna port.

[0431] In some embodiments, the range of antenna port number configurations corresponding to one or more AI models includes one or more of the following:

[0432] One AI model is used for all antenna ports;

[0433] Different AI models are used for different numbers of antenna ports;

[0434] The range of antenna port numbers applicable to different AI models.

[0435] It should be noted that when the number of the one or more AI models is one, the one AI model can be used for all antenna ports.

[0436] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0437] It should be noted that, when there are multiple AI models, the multiple AI models may be different AI models, and the different AI models may be used for different numbers of antenna ports. For example, the number of antenna ports applicable to one AI model is greater than the fifth threshold, and the number of antenna ports applicable to another AI model is less than the fifth threshold.

[0438] Among them, the fifth threshold can be a predefined parameter value or a parameter value set in other ways, and the embodiment of the present application does not limit this.

[0439] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0440] It should be noted that when the first capability represents the range of antenna port number configurations corresponding to one or more AI models, the first capability may indicate whether different numbers of antenna ports use the same AI model; and / or, the first capability may indicate the range of antenna port numbers applicable to different AI models.

[0441] For example, the terminal device can use a bitmap to indicate the number of antenna ports applicable to the supported AI model, where each bit corresponds to one antenna port number. For example, the number of antenna ports can be 2, 4, 8, 12, and so on.

[0442] It should be understood that the terminal device can report the range of antenna port number configurations corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents different values ​​for the antenna port number configuration, and whether the terminal device uses the same AI model, that is, whether the adjustment of the antenna port number configuration requires the terminal device to switch the AI ​​model.

[0443] For example, the terminal device can report 1 bit of information, which can indicate whether the terminal device uses the same AI model for different numbers of antenna ports. 0 indicates using different AI models, and 1 indicates using the same AI model; or 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 1 bit of information reported by the terminal device, the network device can determine whether the adjustment of the antenna port number configuration will cause the terminal device to switch the AI ​​model.

[0444] Through this method, when the first capability represents the range of antenna port number configurations corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the antenna port number configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the antenna port number configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0445] In some embodiments, the range of transmission layer configurations corresponding to one or more AI models may include one or more of the following:

[0446] One AI model is used for all transmission layers;

[0447] Different AI models are used for different numbers of transmission layers;

[0448] The range of transmission layers applicable to different AI models.

[0449] It should be noted that when the number of the one or more AI models is one, the one AI model can be used for all transmission layers.

[0450] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0451] It should be noted that, if there are multiple AI models, the multiple AI models may be different AI models, and the different AI models may be used for different numbers of transmission layers. For example, one AI model may be suitable for a number of transmission layers greater than the sixth threshold, while another AI model may be suitable for a number of transmission layers less than the sixth threshold.

[0452] Among them, the sixth threshold can be a predefined parameter value or a parameter value set in other ways, and the embodiment of the present application does not limit this.

[0453] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0454] It should be noted that, when the first capability represents the range of transmission layer configurations corresponding to one or more AI models, the first capability may indicate whether different transmission layers use the same AI model; and / or, the first capability may indicate the range of transmission layer numbers applicable to different AI models.

[0455] For example, the terminal device can use a bitmap to indicate the number of transmission layers applicable to the supported AI model, where each bit corresponds to a transmission layer number. For example, the range of transmission layer numbers can be 1, 2, and so on.

[0456] It should be understood that the terminal device can report the range of the transmission layer configuration corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents a different value for the transmission layer configuration, and whether the terminal device uses the same AI model, that is, whether the adjustment of the transmission layer configuration requires the terminal device to switch the AI ​​model.

[0457] For example, the terminal device can report 1 bit of information, which can indicate whether the terminal device uses the same AI model for different numbers of transmission layers. 0 indicates using different AI models, and 1 indicates using the same AI model; or 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 1 bit of information reported by the terminal device, the network device can determine whether the adjustment of the transmission layer configuration will cause the terminal device to switch the AI ​​model.

[0458] Through this method, when the first capability represents the range of the transmission layer configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the transmission layer configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the transmission layer configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0459] In some embodiments, one or more AI models can be used for data transmission or data reception. For example, for downlink transmission, one or more AI models input downlink received signals and output data detection results or channel measurement results; for uplink transmission, one or more AI models input to-be-transmitted data or signals and output a transmit signal processed by the transmitter.

[0460] In some embodiments, the range of transmission bandwidth configuration corresponding to one or more AI models may include one or more of the following:

[0461] One AI model is used for all transmission bandwidth;

[0462] Different AI models are used for different transmission bandwidths;

[0463] The transmission bandwidth range applicable to different AI models.

[0464] It should be noted that, when the number of the one or more AI models is one, the one AI model can be used for all transmission bandwidths.

[0465] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0466] It should be noted that, when there are multiple one or more AI models, the multiple AI models may be different AI models, and the different AI models may be used for different transmission bandwidths.

[0467] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0468] It should be noted that, when the first capability represents the range of transmission bandwidth configuration corresponding to one or more AI models, the first capability may indicate whether different transmission bandwidths use the same AI model; and / or, the first capability may indicate the transmission bandwidth range applicable to different AI models.

[0469] For example, the terminal device can use a bitmap to indicate the transmission bandwidth applicable to the supported AI model, where each bit corresponds to a transmission bandwidth. For example, the transmission bandwidth can be 5MHz, 10MHz, 20MHz, 100MHz, etc.

[0470] It should be understood that the terminal device can report the range of transmission bandwidth configuration corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents a different value for the transmission bandwidth configuration, and whether the terminal device uses the same AI model, that is, whether the adjustment of the transmission bandwidth configuration requires the terminal device to switch the AI ​​model.

[0471] For example, the terminal device can report 1 bit of information, which can indicate whether the terminal device uses the same AI model for different transmission bandwidths. 0 indicates using different AI models, and 1 indicates using the same AI model; or 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 1 bit of information reported by the terminal device, the network device can determine whether the adjustment of the transmission bandwidth configuration will cause the terminal device to switch the AI ​​model.

[0472] Through this method, when the first capability represents the range of the transmission bandwidth configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the transmission bandwidth configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the transmission bandwidth configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0473] In some embodiments, the range of carrier frequency configurations corresponding to one or more AI models may include one or more of the following:

[0474] One AI model is used for all carrier frequencies;

[0475] Different AI models are used for different carrier frequencies;

[0476] Carrier frequency range applicable to different AI models.

[0477] It should be noted that, when the number of the one or more AI models is one, the one AI model can be used for all carrier frequencies.

[0478] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0479] It should be noted that, when there are multiple one or more AI models, the multiple AI models may be different AI models, and the different AI models may be used for different carrier frequencies.

[0480] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0481] It should be noted that when the first capability represents the range of carrier frequency configuration corresponding to one or more AI models, the first capability may indicate whether different carrier frequencies use the same AI model; and / or, the first capability may indicate the carrier frequency range to which different AI models are applicable.

[0482] For example, the terminal device can use a bitmap to indicate the carrier frequency applicable to the supported AI model, where each bit corresponds to a carrier frequency. The carrier frequency can be predefined FR1, FR2, FR3, etc.

[0483] It should be understood that the terminal device can report the range of carrier frequency configuration corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents different values ​​for the carrier frequency configuration, and whether the terminal device uses the same AI model, that is, whether the adjustment of the carrier frequency configuration requires the terminal device to switch the AI ​​model.

[0484] For example, the terminal device can report 1 bit of information, which can indicate whether the terminal device uses the same AI model for different carrier frequencies. 0 indicates using different AI models, and 1 indicates using the same AI model; or 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 1 bit of information reported by the terminal device, the network device can determine whether the adjustment of the carrier frequency configuration will cause the terminal device to switch the AI ​​model.

[0485] Through this method, when the first capability represents the range of carrier frequency configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the carrier frequency configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the carrier frequency configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0486] In some embodiments, the range of subcarrier spacing configuration corresponding to one or more AI models may include one or more of the following:

[0487] One AI model is used for all subcarrier spacings;

[0488] Different AI models are used for different subcarrier spacings;

[0489] The subcarrier spacing range applicable to different AI models.

[0490] It should be noted that when the number of the one or more AI models is one, the one AI model can be used for all subcarrier spacings.

[0491] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0492] It should be noted that, when there are multiple one or more AI models, the multiple AI models may be different AI models, and the different AI models may be used for different subcarrier spacings.

[0493] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0494] It should be noted that when the first capability represents the range of subcarrier spacing configuration corresponding to one or more AI models, the first capability may indicate whether different subcarrier spacings use the same AI model; and / or, the first capability may indicate the subcarrier spacing range applicable to different AI models.

[0495] For example, the terminal device can use a bitmap to indicate the subcarrier spacing applicable to the supported AI model, where each bit corresponds to a subcarrier spacing. The subcarrier spacing can be predefined 15kHz, 30kHz, 120kHz, etc.

[0496] It should be understood that the terminal device can report the range of subcarrier spacing configuration corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents different values ​​for the subcarrier spacing configuration, and whether the terminal device uses the same AI model, that is, whether the adjustment of the subcarrier spacing configuration requires the terminal device to switch the AI ​​model.

[0497] For example, the terminal device can report 1 bit of information, which can indicate whether the terminal device uses the same AI model for different subcarrier spacing. 0 indicates using different AI models, and 1 indicates using the same AI model; or 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 1 bit of information reported by the terminal device, the network device can determine whether the adjustment of the subcarrier spacing configuration will cause the terminal device to switch the AI ​​model.

[0498] Through this method, when the first capability represents the range of subcarrier spacing configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the subcarrier spacing configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the subcarrier spacing configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0499] In some embodiments, the scope of the TCI state configuration corresponding to one or more AI models includes one or more of the following:

[0500] One AI model is used for all TCI states;

[0501] Different AI models are used for different TCI states;

[0502] One AI model is used for single TCI state and multi-TCI state;

[0503] Different AI models are used for single TCI state and multi-TCI state respectively.

[0504] It should be noted that, when the number of the one or more AI models is one, the one AI model can be used for all TCI states; and / or, the one AI model can be used for a single TCI state and multiple TCI states.

[0505] For example, one AI model can be used for all TCI states. When network devices are configured with different TCI states, the terminal device does not need to switch models, and the TCI state can be indicated through DCI signaling.

[0506] For example, an AI model can be used for both single and multiple TCI states. When a network device configures a single TCI state for an uplink or downlink signal, and multiple TCI states for an uplink or downlink signal, the terminal device can use the same AI model for both signal transmission and signal reception.

[0507] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0508] It should be noted that, when there are multiple one or more AI models, the multiple AI models may be different AI models, and each of the multiple AI models may correspond to a partial TCI state.

[0509] For example, different AI models are used for different TCI states. When the network device is configured with different TCI states, the terminal device needs to switch the AI ​​model. In order to avoid dynamic AI model switching, the TCI state can be indicated through RRC signaling or MAC signaling.

[0510] For example, different AI models are used for single TCI states and multiple TCI states, respectively. When a network device configures a single TCI state for an uplink or downlink signal, the terminal device uses one AI model for sending or receiving that signal; when a network device configures multiple TCI states for an uplink or downlink signal, the terminal device uses another AI model for sending or receiving that signal.

[0511] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0512] It should be noted that, when the first capability represents the range of TCI state configurations corresponding to one or more AI models, the first capability (e.g., 1 bit) can indicate whether different TCI states use the same AI model; and / or, the first capability (e.g., 1 bit) can indicate whether the single TCI state and the multi-TCI state use the same AI model.

[0513] It should be understood that the terminal device can report the range of TCI state configurations corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents different values ​​for the TCI state configuration, whether the terminal device uses the same AI model, that is, whether the adjustment of the TCI state configuration requires the terminal device to switch the AI ​​model; one bit corresponds to one TCI state configuration.

[0514] For example, the terminal device can report 2 bits of information, where the first bit indicates whether different AI models are used for different TCI states, and the second bit indicates whether different AI models are used for single TCI state and multi-TCI state, respectively. 0 indicates using different AI models, and 1 indicates using the same AI model; or 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 2 bits of information reported by the terminal device, the network device can determine whether the adjustment of a certain TCI state configuration will cause the terminal device to switch the AI ​​model.

[0515] Through this method, when the first capability represents the range of TCI state configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the TCI state configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model based on the TCI state configuration exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0516] In some embodiments, the range of cell ID configurations corresponding to one or more AI models may include one or more of the following:

[0517] One AI model is used for all cell IDs;

[0518] Different AI models are used for different cell IDs;

[0519] The range of cell IDs applicable to different AI models.

[0520] Exemplarily, the cell ID configuration may be carried by a synchronization signal of the target cell.

[0521] It should be noted that, when the number of the one or more AI models is one, the one AI model can be used for all cell IDs.

[0522] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0523] It should be noted that, when there are multiple one or more AI models, the multiple AI models may be different AI models, and the different AI models may be used for different cell IDs.

[0524] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0525] It should be noted that, when the first capability represents the range of cell ID configurations corresponding to one or more AI models, the first capability can use 1-bit information to indicate whether the terminal device uses the same AI model for data reception or data transmission when the terminal device accesses cells with different cell IDs; and / or, the first capability can indicate the cell ID ranges to which different AI models supported by the terminal device are respectively applicable, that is, each AI model can be used in cells corresponding to a group of cell IDs.

[0526] It should be understood that the terminal device can report the range of cell ID configuration corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents a different value for the cell ID configuration, and whether the terminal device uses the same AI model, that is, whether the adjustment of the cell ID configuration requires the terminal device to switch the AI ​​model.

[0527] For example, the terminal device can report 1 bit of information, which can indicate whether the terminal device uses the same AI model for different cell IDs. 0 indicates using different AI models, and 1 indicates using the same AI model; or 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 1 bit of information reported by the terminal device, the network device can determine whether the adjustment of the cell ID configuration will cause the terminal device to switch the AI ​​model.

[0528] Through this method, when the first capability represents the range of cell ID configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the cell ID configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model configured based on the cell ID exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0529] Method #D: When one or more AI models are used for CSI measurement, data transmission, or data reception, the first capability may represent the range of MIMO transmission scheme configuration corresponding to the one or more AI models.

[0530] In some embodiments, one or more AI models can be used for data transmission or data reception. For example, for downlink transmission, one or more AI models input the downlink received signal and output data detection results, such as the source bitstream; for uplink transmission, one or more AI models input the data to be transmitted and output the transmit signal on each antenna port; for CSI measurement, one or more AI models input the downlink received signal and output CSI measurement results, such as RSRP or PMI bits.

[0531] In some embodiments, the range of MIMO transmission scheme configurations corresponding to one or more AI models may include one or more of the following:

[0532] One AI model is used for single-TRP transmission scheme and multi-TRP collaboration scheme;

[0533] Different AI models are used for single-TRP transmission solutions and multi-TRP collaboration solutions;

[0534] One AI model is used for both open-loop and closed-loop transmission solutions;

[0535] Different AI models are used for open-loop transmission solutions and closed-loop transmission solutions respectively.

[0536] It should be noted that the MIMO transmission scheme may include a single TRP transmission scheme, a multi-TRP collaborative scheme, an open-loop transmission scheme, a closed-loop transmission scheme, and may also include other transmission schemes, which are not limited in the embodiments of the present application.

[0537] It should also be noted that the single TRP transmission scheme and / or the multi-TRP coordination scheme can be determined by the number of TCI states configured by the network device for a certain signal or channel; or, a dedicated signaling (such as RRC signaling) can be used to indicate whether to adopt the multi-TRP coordination scheme. The multi-TRP coordination scheme may include one or more of the following:

[0538] A scheme for non-coherent joint transmission (NC-JT) of multiple TRPs;

[0539] Coherent Joint Transmission (CJT) scheme for multiple TRPs;

[0540] A solution for single frequency network (SFN) diversity transmission using multiple TRPs;

[0541] A scheme in which multiple TRPs perform repeated transmission.

[0542] It should be understood that an open-loop transmission scheme may refer to a transmission scheme that does not require the terminal device to feed back precoding information (such as PMI), and a closed-loop transmission scheme may refer to a scheme that requires the terminal device to feed back precoding information.

[0543] It should be noted that when the number of the one or more AI models is one, the one AI model can be used for a single TRP transmission scheme and a multi-TRP collaboration scheme; and / or, the one AI model can be used for an open-loop transmission scheme and a closed-loop transmission scheme.

[0544] Further, when the number of the first AI model is one, the one AI model may be the first AI model.

[0545] It should be noted that when the number of the one or more AI models is multiple, the multiple AI models can be different AI models, and the different AI models can be used for a single TRP transmission scheme and a multi-TRP collaboration scheme respectively; and / or, the different AI models can be used for an open-loop transmission scheme and a closed-loop transmission scheme respectively.

[0546] Further, in a case where the first AI model includes at least two different AI models, the plurality of AI models may include the first AI model.

[0547] It should be noted that, when the first capability represents the range of MIMO transmission scheme configuration corresponding to one or more AI models, the first capability (e.g., 1 bit) can indicate whether the single TRP transmission scheme and the multi-TRP collaboration scheme use the same AI model; and / or, the first capability (e.g., 1 bit) can indicate whether the open-loop transmission scheme and the closed-loop transmission scheme use the same AI model.

[0548] It should be understood that the terminal device can report the range of MIMO transmission scheme configuration corresponding to one or more AI models to the network device through Bitmap. Each bit reported by the terminal device represents different values ​​​​for the MIMO transmission scheme configuration, whether the terminal device uses the same AI model, that is, whether the adjustment of the MIMO transmission scheme configuration requires the terminal device to switch the AI ​​model; one bit corresponds to a MIMO transmission scheme configuration.

[0549] Exemplarily, the terminal device can report 2 bits of information, where the first bit indicates whether the single TRP transmission scheme and the multi-TRP collaboration scheme use the same AI model, and the second bit indicates whether the open-loop transmission scheme and the closed-loop transmission scheme use the same AI model. 0 indicates using different AI models, and 1 indicates using the same AI model; or, 0 indicates using the same AI model, and 1 indicates using different AI models. Based on the 2-bit information reported by the terminal device, the network device can determine whether the adjustment of a certain MIMO transmission scheme configuration will cause the AI ​​model to be switched on the terminal device side.

[0550] Through this method, when the first capability represents the range of the MIMO transmission scheme configuration corresponding to one or more AI models, after the terminal device sends the first capability to the network device, the subsequent network device can indicate the MIMO transmission scheme configuration to the terminal device based on the first capability, thereby avoiding the switching of the AI ​​model configured based on the MIMO transmission scheme exceeding the capability of the terminal device (such as dynamic AI model switching), and can reduce the frequency of AI model switching, thereby ensuring the uplink and downlink transmission performance based on the AI ​​model.

[0551] In some embodiments, when the parameter configuration is indicated through DCI signaling and / or MAC signaling, and the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value are different AI models, the terminal device does not communicate with the network device based on the first AI model.

[0552] Among them, the first value is the parameter value before the terminal device receives the parameter configuration.

[0553] Exemplarily, when the DMRS configuration indicates different DMRS sequence IDs, and different DMRS sequence IDs correspond to different AI models, the terminal device does not use the AI ​​model for downlink channel estimation or downlink data detection.

[0554] Exemplarily, when the DMRS configuration indicates different numbers of DMRS symbols, and different numbers of DMRS symbols correspond to different AI models, the terminal device does not use the AI ​​model for downlink channel estimation or downlink data detection.

[0555] It should be noted that since most terminal devices can support dynamic switching between AI model-based transmission modes and non-AI model transmission modes, but cannot support dynamic switching between different AI models (model deployment takes time), when the parameter configuration requires the terminal device to switch AI models beyond its own capabilities, the terminal device can fall back from the AI ​​model-based transmission mode to the non-AI model-based transmission mode, thereby performing downlink channel estimation or downlink data detection based on the non-AI model transmission mode.

[0556] Through this method, when the parameter configuration is indicated through DCI signaling and / or MAC signaling, and the first AI model and the second AI model corresponding to the parameter configuration value are different AI models, the terminal device can communicate with the network device not based on the first AI model, thereby avoiding the AI ​​model switching exceeding the capability of the terminal device (for example, dynamic AI model switching), thereby ensuring uplink and downlink transmission performance.

[0557] In some embodiments, when the first AI model corresponding to the parameter configuration value includes at least two different AI models, the terminal device does not communicate with the network device based on the first AI model; wherein the first value is the parameter value before the terminal device receives the parameter configuration.

[0558] Through this method, when the first AI model corresponding to the value of the parameter configuration includes at least two different AI models, the terminal device can communicate with the network device not based on the first AI model, thereby avoiding the switching of the AI ​​model exceeding the capability of the terminal device (for example, dynamic AI model switching), thereby ensuring the uplink and downlink transmission performance.

[0559] In some embodiments, when the AI ​​model of the terminal device does not support all DMRS configurations, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, through DCI format 0_0 or 1_0 scheduling), and the fallback DCI does not include the DMRS configuration, the terminal device may communicate with the network device not based on the AI ​​model, for example, the terminal device does not use the AI ​​model for downlink channel estimation or downlink data detection.

[0560] In other embodiments, when the AI ​​model of the terminal device does not support all DMRS configurations, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, through DCI format 0_0 or 1_0 scheduling), and the DMRS configuration corresponding to the fallback DCI is different from the DMRS configuration corresponding to the non-fallback DCI, the terminal device may communicate with the network device not based on the AI ​​model.

[0561] For example, since most terminal devices can support dynamic switching between AI model-based transmission modes and non-AI model transmission modes, but cannot support dynamic switching between different AI models (model deployment takes time), when the AI ​​model of the terminal device does not support all DMRS configurations, the transmission scheduled by the fallback DCI uses the default DMRS configuration (such as the default number of DMRS symbols and the default DMRS port), and the default DMRS configuration is different from the DMRS configuration previously indicated by the non-fallback DCI, for the transmission scheduled by the fallback DCI, the terminal device does not use AI model-based channel estimation or detection, but uses linear filtering or a linear receiver.

[0562] In some embodiments, when different AI models are used for different numbers of transmission layers or antenna ports, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, scheduled through DCI format 0_0 or 1_0), and the fallback DCI does not include the number of transmission layers configuration or the number of antenna ports configuration, the terminal device may communicate with the network device not based on the AI ​​model, for example, the terminal device does not use the AI ​​model for data sending or data receiving.

[0563] In other embodiments, when different AI models are used for different numbers of transmission layers or antenna ports, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, through DCI format 0_0 or 1_0 scheduling), and the number of transmission layers or antenna ports configuration corresponding to the fallback DCI is different from the number of transmission layers or antenna ports configuration corresponding to the non-fallback DCI, the terminal device may communicate with the network device not based on the AI ​​model.

[0564] For example, since most terminal devices can support dynamic switching between AI model-based transmission modes and non-AI model-based transmission modes, but cannot support dynamic switching between different AI models (model deployment takes time), when different AI models are used for different transmission layers, the transmission scheduled by the fallback DCI uses the default transmission layer configuration (such as the transmission layer is 1), and the default transmission layer configuration is different from the transmission layer configuration previously indicated by the non-fallback DCI, for the transmission scheduled by the fallback DCI, the terminal device does not use data sending or data reception based on the AI ​​model, but uses data sending or data reception based on the non-AI model.

[0565] In some embodiments, when different AI models are used for different carrier frequencies or subcarrier spacings, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, scheduled through DCI format 0_0 or 1_0), and the fallback DCI does not include carrier frequency configuration or subcarrier spacing configuration, the terminal device may communicate with the network device not based on the AI ​​model, for example, the terminal device does not use the AI ​​model for data sending or data receiving.

[0566] In other embodiments, when different AI models are used for different carrier frequencies or subcarrier spacings, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, through DCI format 0_0 or 1_0 scheduling), and the carrier frequency configuration or subcarrier spacing configuration corresponding to the fallback DCI is different from the carrier frequency configuration or subcarrier spacing configuration corresponding to the non-fallback DCI, the terminal device may communicate with the network device not based on the AI ​​model.

[0567] For example, since most terminal devices can support dynamic switching between AI model-based transmission modes and non-AI model-based transmission modes, but cannot support dynamic switching between different AI models (model deployment takes time), when different carrier frequencies use different AI models, the transmission scheduled by the fallback DCI uses the default carrier frequency configuration (such as FR1), and the default carrier frequency configuration is different from the carrier frequency configuration previously indicated by the non-fallback DCI, for the transmission scheduled by the fallback DCI, the terminal device does not use AI model-based data sending or data reception, but uses non-AI model data sending or data reception.

[0568] In some embodiments, when different MIMO transmission schemes use different AI models and the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, through DCI format 0_0 or 1_0 scheduling), the terminal device may communicate with the network device not based on the AI ​​model, for example, the terminal device does not use the AI ​​model for data transmission or data reception.

[0569] In other embodiments, when different MIMO transmission schemes adopt different AI models, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, through DCI format 0_0 or 1_0 scheduling), and the MIMO transmission scheme configuration corresponding to the fallback DCI is different from the MIMO transmission scheme configuration corresponding to the non-fallback DCI, the terminal device may communicate with the network device not based on the AI ​​model.

[0570] For example, since most terminal devices can support dynamic switching between AI model-based transmission modes and non-AI model transmission modes, but cannot support dynamic switching between different AI models (model deployment takes time), when different MIMO transmission schemes use different AI models, the transmission scheduled by the fallback DCI uses the default single TRP transmission scheme, and the transmission scheduled by the non-fallback DCI is configured with a multi-TRP transmission scheme, for the transmission scheduled by the fallback DCI, the terminal device does not use data sending or data reception based on the AI ​​model, but uses data sending or data reception based on the non-AI model.

[0571] Through this method, the terminal device can communicate with the network device without relying on the AI ​​model, thereby avoiding the switching of the AI ​​model exceeding the capabilities of the terminal device (such as dynamic AI model switching), thereby ensuring the uplink and downlink transmission performance.

[0572] The communication method provided in the embodiment of the present application is described in detail below in conjunction with specific application scenarios.

[0573] The embodiment of the present application provides a method for reporting the generalization performance capability of an AI model, where the terminal device can report the range of parameter configuration corresponding to the AI ​​model used (i.e., the first capability), so that when the network device performs parameter configuration, it can consider the AI ​​model switching caused by the terminal device due to the adjustment of the parameter configuration, and reasonably determine the signaling and range used for the parameter configuration. Through this method, the network device can determine the appropriate parameter configuration based on the first capability reported by the terminal device, avoiding the terminal device from performing AI model switching that exceeds its own capability (such as dynamic AI model switching), or performance loss caused by frequent AI model switching, thereby ensuring the performance of uplink or downlink transmission based on the AI ​​model.

[0574] The following describes the technical solutions provided by the embodiments of the present application based on different types of parameter configurations in combination with Embodiments 1 to 8.

[0575] Embodiment 1: DMRS configuration.

[0576] 1. The terminal device can report the generalization capability of the AI ​​model used, where the generalization capability is represented by the range of the DMRS configuration corresponding to the AI ​​model.

[0577] In the embodiments of the present application, the AI ​​model can be used for downlink channel estimation or for downlink data reception. For example, when the AI ​​model is used for downlink channel estimation, the input of the AI ​​model is the received signal and / or pilot sequence, and the output is the result of channel estimation. When the AI ​​model is used for downlink data reception, the input of the AI ​​model is the received signal and / or pilot sequence, and the output is the detected data bits.

[0578] Since the training of the AI ​​model relies on the received signal of DMRS as input, different DMRS configurations may correspond to different AI models, and different DMRS inputs will also result in different AI model outputs.

[0579] In some embodiments, a terminal device can train a single AI model for all DMRS configurations that a network device may indicate (e.g., different DMRS sequences, different numbers of DMRS ports, etc.). This trained AI model is highly generalizable and can be used by network devices to indicate any DMRS configuration. However, this training approach requires a large dataset, makes model parameter adjustment complex, and makes it difficult to guarantee the accuracy of the resulting channel estimation or detection performance.

[0580] In other embodiments, the terminal device can use different AI models for training for different DMRS configurations (such as different DMRS sequences, different numbers of DMRS ports, different DMRS port sets, etc.), and each AI model is only for part of the configuration. This not only reduces the complexity of training, but also improves the final channel estimation and detection performance. The AI ​​model obtained by this method has poor generalization. When the network device adopts different DMRS configurations, the terminal device needs to switch the corresponding AI model so that the AI ​​model corresponding to the current DMRS configuration can be used for channel estimation or detection.

[0581] In other words, different terminal devices can use different training methods to obtain AI models with different generalization capabilities, and the network device is unaware of this generalization capability. Therefore, the terminal device can report the generalization capability to the network device, so that the network device can determine whether adjusting the DMRS configuration will cause the terminal device's AI model to switch.

[0582] The DMRS configuration scope corresponding to the AI ​​model can include one or more of the following:

[0583] One AI model corresponds to all DMRS configurations.

[0584] Different AI models correspond to different DMRS base sequences; for example, different DMRS sequence IDs.

[0585] Different AI models correspond to different DMRS CDM (code division multiplexing) groups; among them, CDM multiplexing is adopted within a CDM group, and FDM / TDM multiplexing is adopted between different CDM groups.

[0586] Different AI models correspond to different DMRS port multiplexing modes; the multiplexing mode can be CDM / FDM / TDM.

[0587] Different AI models correspond to different numbers of DMRS symbols.

[0588] Different AI models correspond to different DMRS port sets. For example, DMRS ports can be divided into several sets, and different AI models correspond to different port sets. The ports in the port set use the same AI model.

[0589] For example, the terminal device can report UE capabilities (UE capabilities include first capabilities) to the network device, and the UE capabilities (e.g., 2 bits) are used to indicate one or more of the following: the AI ​​model is per UE (i.e., all DMRS configurations share one AI model), per Sequence (i.e., one AI model corresponds to one DMRS base sequence, and different base sequences use different AI models), per Port (i.e., one AI model corresponds to one DMRS port, and different DMRS ports use different AI models). For another example, the UE capabilities (e.g., 1 bit) are used to indicate whether all DMRS configurations share one AI model. Based on the UE capabilities reported by the terminal device, the network device can determine whether the adjustment of a certain DMRS configuration (such as DMRS Sequence ID or DMRS Port Index) will cause the AI ​​model to be switched on the terminal device side.

[0590] In some embodiments, the range of the DMRS configuration corresponding to the AI ​​model can be reported in the form of a bitmap. Each bit that the terminal device can report represents a different value for the DMRS configuration, and whether the terminal device uses the same AI model, that is, whether the adjustment of the DMRS configuration requires the terminal device to switch the AI ​​model.

[0591] Exemplarily, the terminal device can report different values ​​of the DMRS configuration to the network device in the form of a Bitmap, and whether the AI ​​model used by the terminal device is the same. For example, the terminal device can report 3 bits of information, where the first bit indicates whether different DMRS sequences use the same AI model, the second bit indicates whether different DMRS symbol numbers use the same AI model, and the third bit indicates whether different DMRS port sets use the same AI model. Among them, 0 indicates using different models, and 1 indicates using the same model. Based on the information reported by the terminal device, the network device can determine whether the adjustment of a DMRS configuration (such as DMRS Sequence ID or DMRS symbol number) will cause the terminal device side to switch the AI ​​model.

[0592] 2. The network device receives the generalization capability of the AI ​​model reported by the terminal device, where the generalization capability is represented by the range of the DMRS configuration corresponding to the AI ​​model.

[0593] 3. Based on the generalization capability, the network device can indicate the DMRS configuration related to the first AI model to the terminal device.

[0594] In some embodiments, the network device may determine the DMRS configuration based on the range of the DMRS configuration corresponding to the AI ​​model reported by the terminal device.

[0595] If the AI ​​model reported by the terminal device can be used for any DMRS configuration, the network device does not need to consider the switching of the AI ​​model when determining the adjustment of the DMRS configuration.

[0596] If the AI ​​model reported by the terminal device can be used for different values ​​of a certain DMRS configuration (such as DMRS sequence ID), the network device does not need to consider the switching of the AI ​​model when determining the DMRS configuration.

[0597] If the AI ​​model reported by the terminal device is trained for different values ​​of a DMRS configuration (such as the number of DMRS symbols), the network device needs to consider the possible AI model switching on the terminal device side when configuring the DMRS configuration. For example, the network device should not dynamically adjust the DMRS configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side (by default, terminal devices do not support dynamic AI model switching because model deployment takes a certain amount of time).

[0598] At least for a certain function (such as channel estimation or data detection), the parameter value indicated by the DMRS configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use to implement the function.

[0599] In some embodiments, when the DMRS configuration is dynamically indicated through DCI signaling (for example, the DMRS port is indicated through DCI signaling), the DMRS configuration cannot cause dynamic AI model switching on the terminal device side, that is, the parameter value indicated by the DMRS configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model. For example, when the terminal device reports different AI models corresponding to different DMRS port sets, the DMRS port indicated by the network device through DCI signaling can be a port in the port set corresponding to the same AI model, and the port sets corresponding to different AI models can be switched through high-level signaling.

[0600] In other embodiments, when the DMRS configuration is indicated via DCI signaling or MAC signaling, the parameter value indicated by the DMRS configuration corresponds to the same AI model as the original parameter value (i.e., the first value). In other words, the DMRS configuration cannot cause the terminal device to switch the AI ​​model based on DCI signaling or MAC signaling. The DMRS configuration indicated by the network device via RRC signaling can correspond to different AI models.

[0601] 4. The terminal device can receive the DMRS configuration indicated by the network device and use the first AI model corresponding to the DMRS configuration to transmit with the network device.

[0602] In some embodiments, when the DMRS configuration is indicated through DCI signaling, the terminal device does not expect that the parameter value indicated by the DMRS configuration corresponds to a different AI model than the original parameter value (i.e., the first value). In other words, the DMRS configuration cannot cause dynamic AI model switching on the terminal device side.

[0603] In other embodiments, when the DMRS configuration is indicated via DCI signaling or MAC signaling, the terminal device does not expect that the parameter value indicated by the DMRS configuration corresponds to a different AI model than the original parameter value (i.e., the first value); that is, the DMRS configuration cannot cause the terminal device to switch the AI ​​model based on DCI signaling or MAC signaling. The DMRS configuration indicated by the network device via RRC signaling can correspond to different AI models.

[0604] In some embodiments, when the DMRS configuration is indicated by DCI signaling, and the parameter value indicated by the DMRS configuration corresponds to a different AI model from the original parameter value (i.e., the first value), the terminal device does not use the AI ​​model for transmission with the network device. For example, the DMRS configuration indicates different DMRS sequence IDs, and different DMRS sequences correspond to different AI models. That is, when the DMRS configuration requires the terminal device to switch to an AI model that exceeds its capabilities, the terminal device can fall back from the transmission mode based on the AI ​​model to the traditional transmission mode based on the non-AI model for downlink channel estimation or downlink data detection. This is because most terminal devices can support dynamic switching between transmission modes based on the AI ​​model and transmission modes based on the non-AI model, but cannot support dynamic switching between different AI models (model deployment takes time).

[0605] In other embodiments, when the DMRS configuration is indicated by DCI signaling or MAC signaling, and the parameter value indicated by the DMRS configuration corresponds to a different AI model than the original parameter value (i.e., the first value), the terminal device does not use the AI ​​model for transmission with the network device. For example, if the DMRS configuration indicates different numbers of DMRS symbols, and different numbers of DMRS symbols correspond to different AI models, the terminal device does not use the AI ​​model for downlink channel estimation or downlink data detection.

[0606] At least for a certain function (such as channel estimation or data detection), the parameter values ​​indicated by the DMRS configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use to implement the function. For example, if different DMRS port sets correspond to different AI models for channel estimation, the network device cannot indicate ports in two DMRS port sets at the same time.

[0607] In some embodiments, if the AI ​​model of the terminal device does not support all DMRS configurations, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, through DCI format 0_0 or 1_0 scheduling), and the fallback DCI does not include the DMRS configuration, then the terminal device does not use the AI ​​model for transmission with the network device, for example, the terminal device does not use the AI ​​model for downlink channel estimation or downlink data detection.

[0608] In other embodiments, if the AI ​​model of the terminal device does not support all DMRS configurations, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, scheduled through DCI format 0_0 or 1_0), and the DMRS configuration corresponding to the fallback DCI is different from the DMRS configuration corresponding to the non-fallback DCI, then the terminal device does not use the AI ​​model for transmission with the network device. For example, if the transmission scheduled by the fallback DCI uses the default DMRS configuration (such as the default number of DMRS symbols and the default DMRS port), and the default configuration is different from the DMRS configuration previously indicated by the non-fallback DCI, then for the transmission scheduled by the fallback DCI, the terminal device does not use channel estimation or detection based on the AI ​​model, but uses linear filtering or linear receiver based on a non-AI model. This is because most terminal devices can support dynamic switching between channel estimation based on the AI ​​model and traditional channel estimation methods, but cannot support dynamic switching between different AI models (model deployment takes time).

[0609] Embodiment 2: CSI-RS configuration.

[0610] 1. The terminal device can report the generalization capability of the AI ​​model used, where the generalization capability is represented by the range of the CSI-RS configuration corresponding to the AI ​​model.

[0611] In an embodiment of the present application, an AI model can be used for downlink CSI measurement. For example, the input of the AI ​​model is a downlink received signal and / or a CSI-RS sequence, and the output is a CSI measurement result, such as a PMI bit.

[0612] Since CSI measurement is based on CSI-RS, AI model training relies on the received CSI-RS signal as input. Different CSI-RS configurations may correspond to different AI models, and different CSI-RS signal inputs will also result in different AI model outputs.

[0613] In some embodiments, a terminal device can train a single AI model for all CSI-RS configurations that a network device may indicate (e.g., different CSI-RS base sequences, different numbers of CSI-RS symbols, and numbers of CSI-RS ports). This results in a highly generalizable AI model that can be used by network devices to indicate any CSI-RS configuration. However, this training approach requires a large dataset, complicates model parameter adjustment, and can make the accuracy of the resulting CSI feedback difficult to guarantee.

[0614] In other embodiments, the terminal device can use different AI models for training for different CSI-RS configurations (e.g., different CSI-RS sequences, different numbers of CSI-RS ports, etc.), with each AI model only targeting a portion of the configuration. This not only reduces the complexity of training but also improves the final CSI feedback performance. The AI ​​model obtained by this method has poor generalization. When the network device uses a different CSI-RS configuration, the terminal device needs to switch the AI ​​model accordingly, so that the AI ​​model corresponding to the current CSI-RS configuration is used for CSI measurement.

[0615] In other words, different terminal devices can use different training methods to obtain AI models with different generalization capabilities, and the network device is unaware of this generalization capability. Therefore, the terminal device can report the generalization capability to the network device, so that the network device can determine whether adjusting the CSI-RS configuration will cause the terminal device's AI model to switch.

[0616] The range of CSI-RS configurations corresponding to AI models can include one or more of the following:

[0617] One AI model corresponds to all CSI-RS configurations.

[0618] Different AI models correspond to different CSI-RS base sequences; for example, different CSI-RS base sequence IDs.

[0619] Different AI models correspond to different sets of CSI-RS port numbers. For example, one AI model corresponds to {2, 4, 8} CSI-RS ports, one AI model corresponds to {12, 16, 20, 24} CSI-RS ports, and one AI model corresponds to {32, 64, 128} CSI-RS ports. Alternatively, each AI model can correspond to only one of these port numbers, and different AI models can be used for training with different port numbers.

[0620] Different AI models correspond to different CSI-RS port multiplexing modes; the multiplexing mode can be CDM / FDM / TDM.

[0621] Different AI models correspond to different numbers of CSI-RS symbols; the number of symbols can be configured through RRC signaling.

[0622] Different AI models correspond to different CSI-RS transmission bandwidths. For example, when the CSI-RS transmission bandwidth is less than a certain value (i.e., a second threshold), one AI model is adopted; when the CSI-RS transmission bandwidth is greater than the certain value, another AI model is adopted.

[0623] For example, the terminal device can report UE capabilities to the network device. The UE capabilities (e.g., 2 bits) can be used to indicate one or more of the following: per UE (i.e., all CSI-RS configurations share one AI model), per Sequence (i.e., one AI model corresponds to one CSI-RS base sequence, and different base sequences use different AI models), and per Port Number (i.e., one AI model corresponds to one CSI-RS port number, and different AI models are used for different CSI-RS port numbers). Based on the information reported by the terminal device, the network device can determine whether the adjustment of a certain CSI-RS configuration (such as the CSI-RS sequence ID or the number of CSI-RS ports) will cause the AI ​​model on the terminal device side to switch.

[0624] In some embodiments, the terminal device can report different values ​​for the CSI-RS configuration to the network device in the form of a Bitmap, whether the AI ​​model used by the terminal device is the same, and one bit corresponds to one CSI-RS configuration. For example, the terminal device can report 3 bits of information, where the first bit indicates whether different CSI-RS base sequences use the same AI model, the second bit indicates whether different numbers of CSI-RS ports use the same AI model, and the third bit indicates whether different CSI-RS transmission bandwidths use the same AI model. Among them, 1 indicates the use of different AI models, and 0 indicates the use of the same AI model. Based on the information reported by the terminal device, the network device can determine whether the adjustment of a certain CSI-RS configuration will cause the AI ​​model to be switched on the terminal device side.

[0625] 2. The network device receives the generalization capability of the AI ​​model reported by the terminal device, where the generalization capability is represented by the range of the CSI-RS configuration corresponding to the AI ​​model.

[0626] 3. Based on the generalization capability, the network device can indicate the CSI-RS configuration related to the first AI model to the terminal device.

[0627] The network device can determine the CSI-RS configuration based on the range of the CSI-RS configuration corresponding to the AI ​​model reported by the terminal device.

[0628] If the AI ​​model reported by the terminal device can be used for any CSI-RS configuration, the network device can determine that the adjustment of the CSI-RS configuration does not need to consider the switching of the AI ​​model.

[0629] If the AI ​​model reported by the terminal device can be used for different values ​​of a CSI-RS configuration (such as the number of CSI-RS ports), the network device determines the CSI-RS configuration without considering the switching of the AI ​​model.

[0630] If the AI ​​model reported by the terminal device is trained for different values ​​of a CSI-RS configuration (such as the number of CSI-RS symbols), the network device needs to consider the potential AI model switching on the terminal device side when indicating the CSI-RS configuration. For example, the network device should not dynamically adjust the CSI-RS configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0631] The parameter values ​​indicated by the CSI-RS configuration cannot correspond to two different AI models for CSI feedback at the same time, otherwise the terminal device cannot determine which AI model to use.

[0632] In some embodiments, when the CSI-RS configuration is dynamically indicated through DCI signaling (for example, non-periodic CSI reporting is triggered by DCI signaling, different CSI reports correspond to different CSI-RS resource configurations, and the CSI-RS resource configuration is determined according to DCI), the CSI-RS configuration cannot cause dynamic AI model switching on the terminal device side, that is, the parameter value indicated by the CSI-RS configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model for CSI feedback. The above method can also be used when the CSI-RS configuration is indicated by MAC CE signaling.

[0633] In other embodiments, the network device may receive the AI ​​model switching capability (i.e., the second capability) reported by the terminal device, which may be used to indicate whether the terminal device supports AI model switching based on DCI signaling, or supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. If the terminal device cannot support AI model switching of a certain signaling, when the network device uses the signaling to configure CSI-RS, the value of the CSI-RS configuration cannot cause the AI ​​model switching of the terminal device. In other words, the parameter value indicated by the CSI-RS configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model.

[0634] 4. The terminal device can receive the CSI-RS configuration indicated by the network device and use the first AI model corresponding to the CSI-RS configuration to transmit with the network device.

[0635] In some embodiments, when the CSI-RS configuration is indicated by DCI signaling, the terminal device does not expect that the parameter value indicated by the CSI-RS configuration corresponds to a different AI model than the original parameter value (i.e., the first value), that is, the CSI-RS configuration cannot cause dynamic AI model switching on the terminal device side. For example, when the terminal device reports different AI models for different numbers of CSI-RS ports, if the network device triggers non-periodic CSI reporting through DCI signaling, the number of CSI-RS ports used for CSI reporting cannot be dynamically adjusted (for example, multiple CSI reports triggered at the same time use different numbers of CSI-RS ports). The above method can also be used when the CSI reporting configuration (CSI-RS configuration) is indicated by MAC CE signaling.

[0636] In some embodiments, the parameter values ​​indicated by the CSI-RS configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot determine which AI model to use for CSI feedback. For example, when the terminal device reports different AI models for different CSI-RS port numbers, if the network device activates semi-persistent CSI reporting through MAC CE signaling, the multiple CSI-RS resources used for the same CSI reporting measurement cannot correspond to different AI models (for example, the bandwidth of one CSI-RS resource is greater than the threshold value (i.e., the first threshold value), and the bandwidth of another resource is less than the threshold value).

[0637] In other embodiments, the terminal device may report the AI ​​model switching capability (i.e., the second capability) to the network device, where the capability is used to indicate whether the terminal device supports AI model switching based on DCI signaling, supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. The terminal device may report whether it supports one or more of the above AI model switching. For example, the terminal device may report that it supports AI model switching based on RRC signaling, but does not support AI model switching based on DCI signaling and MAC signaling.

[0638] The terminal device does not expect the AI ​​model switching caused by the CSI-RS configuration to exceed the AI ​​model switching capability reported by the terminal device. For example, if the terminal device does not support AI model switching based on DCI signaling, and the CSI-RS configuration is indicated by DCI signaling, the CSI-RS configuration cannot cause the terminal device to switch the AI ​​model. For another example, if the terminal device supports AI model switching based on RRC signaling but does not support AI model switching based on DCI signaling and MAC signaling, the CSI-RS configuration indicated by RRC signaling can correspond to different AI models, and the CSI-RS configuration indicated by DCI signaling and MAC signaling cannot switch the AI ​​model on the terminal device side.

[0639] Example 3: CSI reporting content configuration.

[0640] 1. The generalization capability of the AI ​​model used by the terminal device. The generalization capability is represented by the range of the CSI reporting content configuration corresponding to the AI ​​model.

[0641] In an embodiment of the present application, an AI model can be used for downlink CSI measurement. For example, the input of the AI ​​model is the downlink received signal, and the output is the CSI measurement result, such as RSRP or PMI bits.

[0642] When the CSI reporting content output by the AI ​​model is different, the AI ​​model used may also be different. The training data of the AI ​​model depends on the CSI reporting content as its output. The CSI reporting content configuration can be RSRP reporting and CSI reporting; CSI compression and CSI prediction; or CSI compression, CSI prediction, or CSI prediction and compression.

[0643] In some embodiments, a terminal device can train a single AI model for all possible CSI reporting configurations that a network device might indicate. This trained AI model is highly generalizable and can be used to indicate any CSI reporting configuration. However, this training approach requires a large dataset, complicates model parameter adjustment, and can be difficult to guarantee the accuracy of the resulting CSI feedback.

[0644] In other embodiments, the terminal device can use different AI models for training according to different CSI reporting content configurations (e.g., RSRP reporting, PMI reporting, etc.). Each AI model is trained for only one CSI reporting content configuration. This not only reduces the complexity of training but also improves the final CSI feedback performance. The AI ​​model obtained by this method has poor generalization. When the network device adopts a different CSI reporting content configuration, the terminal device needs to switch the corresponding AI model so that the AI ​​model corresponding to the CSI reporting content configuration is used for CSI feedback.

[0645] In other words, different terminal devices can use different training methods to obtain AI models with different generalization capabilities. The terminal devices can report the generalization capabilities to the network devices, so that the network devices can determine whether adjusting the CSI reporting content configuration will cause the terminal device's AI model to switch.

[0646] The CSI reporting content configuration for AI models can include one or more of the following:

[0647] One AI model is used for RSRP reporting and CSI reporting;

[0648] Different AI models are used for RSRP reporting and CSI reporting respectively;

[0649] An AI model for CSI compression and CSI prediction;

[0650] Different AI models are used for CSI compression and CSI prediction respectively;

[0651] An AI model is used for CSI compression, CSI prediction, and CSI prediction and compression;

[0652] Different AI models are used for CSI compression, CSI prediction, and CSI prediction and compression respectively.

[0653] For example, the terminal device can report UE capabilities to the network device, and the UE capabilities (e.g., 1 bit) can be used to indicate whether CSI compression and CSI prediction in CSI reporting use the same AI model.

[0654] In some embodiments, the terminal device can report different values ​​for the CSI reporting content configuration to the network device in the form of a bitmap, and whether the AI ​​model used by the terminal device is the same. One bit corresponds to whether different AI models are used for RSRP reporting and CSI reporting respectively, and the other bit corresponds to whether different AI models are used for CSI compression and CSI prediction respectively. For example, 0 indicates that the same AI model is used, and 1 indicates that different AI models are used. Based on the information reported by the terminal device, the network device can determine whether the adjustment of the CSI reporting content configuration will cause the terminal device to switch the AI ​​model.

[0655] 2. The network device receives the generalization capability of the AI ​​model reported by the terminal device. The generalization capability is represented by the range of the CSI reporting content configuration corresponding to the AI ​​model.

[0656] 3. Based on the generalization capability, the network device can instruct the terminal device on the CSI reporting content configuration related to the first AI model.

[0657] The network device can determine the CSI reporting content configuration based on the range of the CSI reporting content configuration corresponding to the AI ​​model reported by the terminal device.

[0658] When the AI ​​model used for reporting by the terminal device can be used for all CSI reporting content configurations, the network device can determine that the CSI reporting content configuration does not need to consider the switching of the AI ​​model.

[0659] When the AI ​​model used by the terminal device for reporting is trained for a specific CSI reporting configuration (such as CSI prediction or CSI compression), the network device needs to consider the potential AI model switching on the terminal device side when indicating the CSI reporting configuration. For example, the network device should not dynamically adjust the CSI reporting configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0660] The parameter values ​​indicated by the CSI reporting content configuration cannot correspond to two different AI models for CSI feedback at the same time, otherwise the terminal device will not be able to determine which AI model to use.

[0661] In some embodiments, when the CSI reporting content configuration is dynamically indicated by DCI signaling (for example, aperiodic CSI reporting is triggered by DCI signaling, different CSI reports correspond to different CSI reporting content, and the CSI reporting content is determined based on the DCI), the CSI reporting content configuration cannot cause dynamic AI model switching on the terminal device side, that is, the CSI reporting content triggered by the DCI signaling needs to correspond to the same AI model for CSI feedback. The above method can also be used when the CSI reporting content configuration is indicated by MAC CE signaling.

[0662] 4. The terminal device receives the CSI reporting content configuration indicated by the network device, and uses the first AI model corresponding to the CSI reporting content configuration to transmit between the terminal device and the network device.

[0663] In some embodiments, when the CSI reporting content configuration is indicated by DCI signaling, the terminal device does not expect that the parameter value indicated by the CSI reporting content configuration corresponds to a different AI model than the original parameter value (i.e., the first value). In other words, the CSI reporting content configuration cannot cause dynamic AI model switching on the terminal device side. For example, when the terminal device reports different CSI reporting content configurations using different AI models, if the network device triggers non-periodic CSI reporting through DCI signaling, the CSI reporting content configuration of the CSI report cannot be dynamically adjusted (for example, multiple CSI reports triggered at the same time use different CSI reporting content configurations). The above method can also be used in cases where the CSI reporting content configuration is indicated by MAC CE signaling.

[0664] In some embodiments, the parameter values ​​of a CSI reporting content configuration indication cannot simultaneously correspond to two different AI models for the same CSI feedback. For example, if a terminal device uses different AI models for different values ​​of the CSI reporting content configuration reported by the terminal device, and if the network device activates semi-persistent CSI reporting via MAC CE signaling, the parameter values ​​of the same CSI reporting content configuration indication cannot correspond to different AI models.

[0665] In other embodiments, the terminal device may report the AI ​​model switching capability (i.e., the second capability) to the network device, which may be used to indicate whether the terminal device supports AI model switching based on DCI signaling, or supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. At the same time, the terminal device does not expect the AI ​​model switching caused by the CSI reporting content configuration to exceed the AI ​​model switching capability reported by the terminal device. For example, if different values ​​of the CSI reporting content configuration correspond to different AI models, and the terminal device does not support AI model switching based on DCI signaling, the network device cannot adjust the CSI reporting content configuration for the terminal device through DCI signaling.

[0666] Embodiment 4: MCS configuration or transport block size configuration.

[0667] 1. The terminal device can report the generalization capability of the AI ​​model used, where the generalization capability can be represented by the range of the MCS configuration or transport block size configuration corresponding to the AI ​​model.

[0668] In embodiments of the present application, the AI ​​model can be used for data transmission or data reception. Data reception may include channel decoding and / or demodulation, and data transmission may include channel coding and / or modulation. For example, for downlink transmission, the input of the AI ​​model is the downlink received signal, and the output is the data detection result, such as the source bit stream; for uplink transmission, the input of the AI ​​model is the data to be transmitted, and the output is the modulated and coded transmit signal.

[0669] The training of AI models can depend on the MCS or transport block size of uplink and downlink data. In other words, under different MCS configurations or transport block size configurations, the terminal device may train different AI models.

[0670] In some embodiments, a terminal device can train a unified AI model for all MCS configurations or transport block size configurations supported by the protocol. This AI model can be highly generalizable and can be used for any MCS configuration or transport block size configuration indicated by the network device. However, this training method requires a large dataset, complicates model parameter adjustment, and can be difficult to guarantee transmission performance.

[0671] In other embodiments, the terminal device can use different AI models for training according to different MCS configurations or transmission block size configurations. Each AI model only targets a portion of the configuration. This not only reduces the complexity of training, but also improves the final data transmission performance. The AI ​​model obtained by this method has poor generalization. When the network device adopts a different MCS configuration or transmission block size configuration, the terminal device needs to switch the AI ​​model accordingly.

[0672] That is to say, different terminal devices can adopt different training methods to obtain AI models with different generalization capabilities. At this time, the terminal device can report the generalization capability to the network device, so that the network device can determine whether the adjustment of the MCS configuration or the transmission block size configuration will cause the AI ​​model of the terminal device to switch.

[0673] In some embodiments, the scope of the MCS configuration corresponding to the AI ​​model may include one or more of the following:

[0674] One AI model can be used for all MCSs; for example, all MCSs supported by the protocol use the same AI model on the terminal side. Each MCS corresponds to a modulation method and a bit rate.

[0675] Different AI models are used for different MCSs. For example, different AI models are used on the terminal device side for different MCSs supported in the protocol.

[0676] The MCS index range applicable to different AI models; for example, MCS indexes 0-12 correspond to one AI model, and 12-27 correspond to another AI model.

[0677] One AI model is used for all modulation schemes; for example, QPSK, 16QAM, and 64QAM use the same AI model.

[0678] Different AI models are used for different modulation methods; for example, QPSK, 16QAM, 64QAM, and 256QAM use different AI models.

[0679] A set of modulation methods applicable to different AI models; for example, one AI model supports QPSK and 16QAM data transmission and reception, while another AI model supports 64QAM and 256QAM data transmission and reception.

[0680] One AI model is used for all bitrate ranges; for example, all bitrates supported by the protocol use the same AI model.

[0681] The bitrate range applicable to different AI models; for example, when the bitrate is less than 1 / 3, one AI model is used, and when it is greater than 1 / 3, another AI model is used

[0682] Exemplarily, the terminal device can report UE capabilities to the network device. The UE capabilities (e.g., 2 bits) are used to indicate one or more of the following: per UE (i.e., all MCS configurations share one AI model), per Modulation Order (i.e., one AI model corresponds to one modulation mode, and different AI models use different modulation modes), and per MCS (i.e., one model corresponds to one MCS, and different MCSs use different AI models). Based on the information reported by the terminal device, the network device can determine whether the adjustment of the MCS configuration will cause the AI ​​model to be switched on the terminal device side.

[0683] For example, the UE capability may be indicated in one or more of the following ways:

[0684] 1-bit information, used to indicate whether different MCSs use the same AI model;

[0685] 1-bit information, used to indicate whether different modulation modes use the same AI model;

[0686] 1-bit information, used to indicate whether different bit rates use the same AI model;

[0687] N1 bit information is used to indicate the MCS range supported by different AI models;

[0688] N2 bits of information are used to indicate the modulation modes supported by different AI models;

[0689] N3 bits of information are used to indicate the bit rate ranges supported by different AI models.

[0690] In some embodiments, the range of transport block size configurations corresponding to the AI ​​model may include one or more of the following:

[0691] One AI model is used for all transport block sizes;

[0692] Different AI models are used for different transmission block sizes;

[0693] The applicable transmission block size range for different AI models.

[0694] For example, the terminal device can use 1-bit information to indicate whether different transmission block sizes use the same AI model for data transmission or reception. 1 means that the same AI model is used for different sizes, and 0 means that different AI models are used when the transmission block size is greater than a certain threshold value (i.e., the fourth threshold value) and less than a certain threshold value.

[0695] In some embodiments, the terminal device can report different values ​​of the MCS configuration or the transport block size configuration to the network device in the form of a Bitmap, and whether the AI ​​model used by the terminal device is the same. For example, the terminal device can report 3 bits of information, where the first bit indicates whether different modulation modes use the same AI model, the second bit indicates whether different code rate ranges use the same AI model, and the third bit indicates whether different transport block size ranges use the same AI model. Among them, 1 indicates the use of different AI models, and 0 indicates the use of the same AI model. Based on the information reported by the terminal device, the network device can determine whether the adjustment of the MCS configuration or the transport block size configuration will cause the terminal device side to switch the AI ​​model.

[0696] 2. The network device can receive the generalization capability of the AI ​​model reported by the terminal device, where the generalization capability is represented by the range of the MCS configuration or transmission block size configuration corresponding to the AI ​​model.

[0697] 3. Based on the generalization capability, the network device may indicate to the terminal device the MCS configuration or transport block size configuration associated with the first AI model.

[0698] The network device can determine the MCS configuration based on the range of the MCS configuration corresponding to the AI ​​model reported by the terminal device; and determine the data transmission block size configuration based on the range of the transmission block size configuration corresponding to the AI ​​model reported by the terminal device.

[0699] When the AI ​​model reported by the terminal device can be used for any MCS configuration or transport block size configuration, the network device can determine that the adjustment of the MCS or transport block size does not need to consider the switching of the AI ​​model.

[0700] When the AI ​​model reported by the terminal device can be used for different modulation modes, the network device can determine the adjustment of the modulation mode, which needs to take into account the switching of the AI ​​model.

[0701] When the AI ​​model reported by the terminal device can be used in all bit rate ranges, the network device can determine the adjustment of the transmission bit rate without considering the switching of the AI ​​model.

[0702] In the case where the AI ​​model reported by the terminal device is trained for different values ​​of the MCS configuration or the transport block size configuration, the network device needs to consider the AI ​​model switching on the terminal device side that may be caused by the adjustment of the MCS configuration or the transport block size configuration when indicating the MCS configuration or the transport block size configuration. For example, in the case where the AI ​​model is trained for different modulation modes, the network device should not dynamically adjust the modulation mode (but can still dynamically adjust the bit rate and MCS) to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0703] The parameter values ​​indicated by the MCS configuration or transport block size configuration cannot correspond to two different AI models for data detection at the same time, otherwise the terminal device cannot determine which AI model to use.

[0704] In some embodiments, when the MCS configuration or transport block size configuration is dynamically indicated through DCI signaling, the MCS configuration or transport block size configuration cannot cause dynamic AI model switching on the terminal device side, that is, the parameter value (such as MCS index) indicated by the MCS configuration or transport block size configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model for data detection. This method can also be used when the MCS configuration or transport block size configuration is indicated through MAC CE signaling.

[0705] In other embodiments, the network device may receive the AI ​​model switching capability (i.e., the second capability) reported by the terminal device, which is used to indicate whether the terminal device supports AI model switching based on DCI signaling, or supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. In the case where the terminal device cannot support AI model switching of a certain signaling, if the network device uses the signaling to perform MCS configuration or transport block size configuration, the parameter value indicated by the MCS configuration or transport block size configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model. In other words, the value of the MCS configuration or transport block size configuration cannot cause the AI ​​model switching of the terminal device.

[0706] 4. The terminal device can receive the MCS configuration or transmission block size configuration indicated by the network device, and use the first AI model corresponding to the MCS configuration or transmission block size configuration to transmit with the network device.

[0707] In some embodiments, when the MCS configuration or the transport block size configuration is indicated by DCI signaling, the terminal device does not expect the MCS indicated by the MCS configuration to correspond to a different AI model than the original MCS (i.e., the first value), or the terminal device does not expect the transport block size indicated by the transport block size configuration to correspond to a different AI model than the original transport block size (i.e., the first value). In other words, the MCS configuration or the transport block size configuration cannot cause dynamic AI model switching on the terminal device side. For example, when the terminal device reports different AI models for different modulation modes, if the network device indicates the MCS through DCI signaling, it can dynamically switch between MCSs under the same modulation mode (i.e., different code rates), but cannot dynamically switch between MCSs of different modulation modes. This method can also be used when the MCS configuration or the transport block size configuration is indicated by MAC CE signaling.

[0708] In some embodiments, at least for terminals that do not have multi-model capabilities, the parameter values ​​indicated by the MCS configuration or the transport block size configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot use two AI models for data detection at the same time. For example, when the terminal device reports different code rate ranges (for example, greater than 1 / 3 and less than 1 / 3) using different AI models, if the network device indicates the MCS of two codewords through DCI signaling, the MCS of the two codewords cannot indicate the code rate ranges corresponding to different AI models (for example, one code rate is greater than 1 / 3 and the other code rate is less than 1 / 3).

[0709] In some embodiments, the terminal device may report the AI ​​model switching capability (i.e., the second capability) to the network device, which may be used to indicate whether the terminal device supports AI model switching based on DCI signaling, or supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. The terminal device does not expect the AI ​​model switching caused by the MCS configuration or the transport block size configuration to exceed the AI ​​model switching capability reported by the terminal device. For example, if the terminal device does not support AI model switching based on DCI signaling, and the transport block size configuration is indicated by DCI signaling, the transport block size configuration cannot enable the terminal device to switch the AI ​​model. For example, switching can be performed within the range of transport block sizes corresponding to the same AI model.

[0710] Embodiment 5: Configuring the number of antenna ports or the number of transmission layers.

[0711] 1. The generalization capability of the AI ​​model reported by the terminal device. The generalization capability is represented by the range of the number of antenna ports or transmission layers configured for the AI ​​model.

[0712] In the embodiments of the present application, the AI ​​model is used for data transmission or data reception. For example, for downlink transmission, the input of the AI ​​model is the downlink received signal, and the output is the data detection result, such as the source bit stream; for uplink transmission, the input of the AI ​​model is the data to be transmitted, and the output is the transmit signal on each antenna port.

[0713] Since data transmission and reception are based on antenna ports and transmission layers, AI model training depends on the number of antenna ports and transmission layers. The configuration of the number of antenna ports and transmission layers will affect the input of AI model training data. In other words, under different antenna port configurations or different transmission layer configurations, the terminal device may train different AI models for transmission or reception.

[0714] In some embodiments, a terminal device can train a unified AI model for different antenna port configurations or transmission layer configurations. This AI model can be highly generalizable and can be used with different antenna port configurations or transmission layer configurations indicated by the network device. However, this training method requires a large data set, complicates model parameter adjustment, and can be difficult to guarantee corresponding transmission performance.

[0715] In other embodiments, the terminal device can use different AI models for training according to different antenna port configurations or transmission layer configurations. Each AI model only takes values ​​for a portion of them. This not only reduces the complexity of training, but also improves the final data transmission performance. The AI ​​model obtained by this method has poor generalization capabilities. When the network device uses a different antenna port configuration or transmission layer configuration, the terminal device needs to switch the corresponding AI model.

[0716] That is to say, different terminal devices can adopt different training methods to obtain AI models with different generalization capabilities. At this time, the terminal device can report this generalization capability to the network device, so that the network device can determine whether the adjustment of the number of antenna ports or the number of transmission layers will cause the AI ​​model of the terminal device to switch.

[0717] In some embodiments, the range of antenna port number configuration corresponding to the AI ​​model may include one or more of the following:

[0718] One AI model is used for different numbers of antenna ports;

[0719] Different AI models are used for different numbers of antenna ports;

[0720] The range of antenna port numbers applicable to different AI models.

[0721] For example, UE capabilities can be used to indicate whether different antenna port configurations use the same AI model; or, to indicate the range of antenna port numbers applicable to different AI models. For example, a bitmap is used to indicate the number of antenna ports applicable to the AI ​​model supported by each terminal device, where each bit corresponds to an antenna port number. For example, 11110000 indicates that the number of antenna ports applicable to the AI ​​model is 2, 4, 8, or 12.

[0722] In some embodiments, the range of the number of transmission layers configured for the AI ​​model may include one or more of the following:

[0723] One AI model is used for all transmission layers;

[0724] Different AI models are used for different numbers of transmission layers;

[0725] The range of transmission layers applicable to different AI models.

[0726] For example, UE capabilities can be used to indicate whether different transmission layers use the same AI model; or to indicate the range of transmission layers applicable to different AI models. For example, a bitmap is used to indicate the number of transmission layers applicable to the AI ​​model supported by each terminal device, where each bit corresponds to a transmission layer number. For example, 1100 indicates that the number of transmission layers applicable to the AI ​​model is 1 or 2.

[0727] In some embodiments, the terminal device can report to the network device in a Bitmap manner whether the AI ​​model used by the terminal device is the same for different antenna port number configurations or transmission layer number configurations. For example, the terminal device can report 2 bits of information, where the first bit indicates whether different antenna port number configurations use the same AI model, and the second bit indicates whether different transmission layer number configurations use the same AI model. Among them, 1 indicates using different AI models, and 0 indicates using the same AI model (or vice versa). Based on the above information reported by the terminal device, the network device can determine whether the adjustment of the antenna port number configuration or the transmission layer number configuration will cause the AI ​​model to be switched on the terminal device side.

[0728] 2. The network device receives the generalization capability of the AI ​​model reported by the terminal device. The generalization capability is represented by the range of the number of antenna ports configured or the range of the number of transmission layers configured corresponding to the AI ​​model.

[0729] 3. Based on the generalization capability, the network device may indicate to the terminal device the antenna port number configuration or the transmission layer number configuration related to the first AI model.

[0730] The network device can determine the antenna port number configuration based on the range of antenna port number configuration corresponding to the AI ​​model reported by the terminal device; determine the transmission layer number configuration based on the range of transmission layer number configuration corresponding to the AI ​​model reported by the terminal device.

[0731] When the AI ​​model reported by the terminal device can be used for any antenna port number configuration, the network device does not need to consider the switching of the AI ​​model when determining the antenna port number configuration.

[0732] When the AI ​​model reported by the terminal device can be used for any transmission layer configuration, the network device does not need to consider the switching of the AI ​​model when determining the transmission layer configuration.

[0733] If the AI ​​model reported by the terminal device is trained separately for different values ​​of the antenna port number configuration or the transmission layer number configuration, the network device needs to consider the possible AI model switching on the terminal device side caused by the antenna port number configuration or the transmission layer number configuration when indicating the antenna port number configuration or the transmission layer number configuration. For example, if the AI ​​model is trained separately for different antenna port number configurations, the network device should not dynamically adjust the antenna port number configuration to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device.

[0734] The number of transmission layers or antenna ports indicated by the network device cannot correspond to two different AI models for data detection at the same time; otherwise, the terminal device cannot determine which AI model to use.

[0735] In some embodiments, when the antenna port number configuration or the transmission layer number configuration is dynamically indicated through DCI signaling, the antenna port number configuration or the transmission layer number configuration cannot cause dynamic AI model switching on the terminal device side, that is, the parameter value (such as the number of antenna ports) indicated by the antenna port number configuration or the transmission layer number configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model. The above method can also be used when the antenna port number configuration or the transmission layer number configuration is indicated through MAC CE signaling.

[0736] In some embodiments, the network device may receive the AI ​​model switching capability (i.e., the second capability) reported by the terminal device, which is used to indicate whether the terminal device supports AI model switching based on DCI signaling, or supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. If the terminal device cannot support AI model switching based on a certain signaling, when the network device uses the signaling to configure the number of antenna ports or the number of transmission layers, the parameter value indicated by the antenna port configuration or the transmission layer configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model. In other words, the value of the antenna port configuration or the transmission layer configuration cannot cause the AI ​​model switching of the terminal device.

[0737] 4. The terminal device can receive the antenna port number configuration or the transmission layer number configuration indicated by the network device, and use the first AI model corresponding to the antenna port number configuration or the transmission layer number configuration to transmit with the network device.

[0738] In some embodiments, when the antenna port number configuration or the transmission layer number configuration is indicated by DCI signaling, the terminal device does not expect the parameter value indicated by the antenna port number configuration to correspond to a different AI model than the original parameter value (i.e., the first value), or the terminal device does not expect the parameter value indicated by the transmission layer configuration to correspond to a different AI model than the original parameter value (i.e., the first value). In other words, the antenna port number configuration or the transmission layer number configuration cannot cause dynamic AI model switching on the terminal device side. For example, when the terminal device reports different AI models for different antenna port number configurations, if the network device indicates the antenna port number configuration through DCI signaling, dynamic AI model switching cannot be performed between different antenna port number configurations. The above method can also be used when the antenna port number configuration or the transmission layer number configuration is indicated by MAC CE signaling.

[0739] In some embodiments, at least for terminal devices that do not have multi-model capabilities, the parameter values ​​indicated by the antenna port number configuration or the parameter values ​​indicated by the transmission layer number configuration cannot correspond to two different AI models at the same time, otherwise the terminal device cannot use two AI models for data detection at the same time. For example, in the case where different AI models are used for different antenna port number configurations, if multiple DCI signalings simultaneously schedule the terminal device for uplink or downlink transmission, the multiple DCI signalings cannot simultaneously indicate two different antenna port number configurations; in the case where different AI models are used for different transmission layer number configurations, if multiple DCI signalings simultaneously schedule the terminal device for uplink or downlink transmission, the multiple DCI signalings cannot simultaneously indicate two different transmission layer number configurations.

[0740] In some embodiments, the terminal device may report the AI ​​model switching capability (i.e., the second capability) to the network device, which may be used to indicate whether the terminal device supports AI model switching based on DCI signaling, or supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. The terminal device does not expect the AI ​​model switching caused by the adjustment of the number of transmission layer configuration or the number of antenna ports configuration to exceed the AI ​​model switching capability reported by the terminal device. For example, if the terminal device does not support AI model switching based on DCI signaling, and the number of transmission layer configuration is indicated by DCI signaling, the number of transmission layer configuration cannot enable the terminal device to perform AI model switching, for example, it can only switch within the range of transmission layer numbers corresponding to the same AI model.

[0741] In some embodiments, if different transmission layer configurations or antenna port configurations use different AI models, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, through DCI format 0_0 or 1_0 scheduling), and the fallback DCI does not include the transmission layer configuration or the antenna port configuration, then the terminal device does not use the AI ​​model for transmission with the network device, for example, the terminal device does not use the AI ​​model for sending or receiving data. This is because most terminal devices can support dynamic switching between sending and receiving methods based on AI models and sending and receiving methods based on non-AI models, but cannot support dynamic switching between different AI models (model deployment takes time).

[0742] In some embodiments, if different transmission layer configurations or antenna port configurations use different AI models, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, scheduled through DCI format 0_0 or 1_0), and the transmission layer configuration or antenna port configuration corresponding to the fallback DCI is different from the transmission layer configuration or antenna port configuration corresponding to the non-fallback DCI, then the terminal device does not use the AI ​​model for transmission with the network device. For example, if the transmission scheduled by the fallback DCI uses a default number of transmission layers (such as a transmission layer of 1), and the default number of transmission layers is different from the number of transmission layers previously indicated by the non-fallback DCI, then for the transmission scheduled by the fallback DCI, the terminal device does not use data sending or receiving based on the AI ​​model, but uses a sending and receiving method based on a non-AI model.

[0743] Example 6: Transmission bandwidth configuration / carrier frequency configuration / subcarrier spacing configuration

[0744] 1. The generalization capability of the AI ​​model used by the terminal device. The generalization capability is represented by the range of the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration corresponding to the AI ​​model.

[0745] In the embodiments of the present application, the AI ​​model is used to transmit or receive signals (such as reference signals) or data. For example, for downlink transmission, the input of the AI ​​model is the downlink received signal, and the output is the data detection result or channel measurement result. For uplink transmission, the input of the AI ​​model is the data or signal to be transmitted, and the output is the transmit signal after processing by the transmitter.

[0746] Since data / signal transmission and reception are based on a certain transmission bandwidth, carrier frequency, or subcarrier spacing, AI model training depends on the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration used for transmission. The transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration value affects the input of AI model training data. In other words, under different transmission bandwidth configurations, carrier frequency configurations, or subcarrier spacing configurations, terminal devices may train different AI models for transmission or reception.

[0747] In some embodiments, a terminal device can train a unified AI model for different transmission bandwidth configurations, different carrier frequency configurations, or different subcarrier spacing configurations. This AI model can have good generalization and can be used for different transmission bandwidth configurations, carrier frequency configurations, or subcarrier spacing configurations indicated by the network device. However, this training method requires a large data set, complex model parameter adjustment, and the corresponding transmission performance is difficult to guarantee.

[0748] In other embodiments, the terminal device can use different AI models for training according to different transmission bandwidth configurations, carrier frequency configurations or subcarrier spacing configurations, and each AI model is only for part of the configuration. This not only reduces the complexity of training, but also improves the final data or signal transmission performance. The AI ​​model obtained by this method has poor generalization. When the network device adopts different transmission bandwidth configurations, carrier frequency configurations or subcarrier spacing configurations, the terminal device needs to perform corresponding AI model switching. Different terminal devices can adopt different training methods to obtain AI models with different generalization capabilities. At this time, this generalization capability can be reported to the network device, so that the network device can determine whether the adjustment of the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration will cause the terminal device's AI model to switch.

[0749] In some embodiments, the range of the transmission bandwidth configuration corresponding to the AI ​​model may include one or more of the following:

[0750] One AI model is used for all transmission bandwidth;

[0751] Different AI models are used for different transmission bandwidths;

[0752] The transmission bandwidth range applicable to different models.

[0753] For example, UE capabilities can be used to indicate whether different transmission bandwidth configurations use the same AI model; or to indicate the transmission bandwidth range applicable to different AI models. For example, a bitmap is used to indicate the transmission bandwidth applicable to the AI ​​model supported by each terminal device, with each bit corresponding to a transmission bandwidth. For example, the transmission bandwidth can be 5MHz, 10MHz, 20MHz, 100MHz, and so on.

[0754] In some embodiments, the range of carrier frequency configuration corresponding to the AI ​​model may include one or more of the following:

[0755] One model is used for all carrier frequencies;

[0756] Different AI models are used for different carrier frequencies;

[0757] The carrier frequency range applicable to different models.

[0758] For example, UE capabilities can be used to indicate whether different carrier frequency configurations use the same AI model; or to indicate the carrier frequency ranges applicable to different AI models. For example, a bitmap is used to indicate the carrier frequency ranges applicable to the AI ​​model supported by each terminal device, with each bit corresponding to a carrier frequency range. For example, the carrier frequency range can be predefined FR1, FR2, and FR3, etc. Alternatively, the terminal device can indicate that FR1 uses one AI model and FR2 uses another AI model.

[0759] In some embodiments, the range of subcarrier spacing configuration corresponding to the AI ​​model may include one or more of the following:

[0760] One AI model is used for different subcarrier spacings;

[0761] Different AI models are used for different subcarrier spacings;

[0762] The subcarrier spacing range applicable to different AI models.

[0763] For example, UE capabilities can be used to indicate whether different subcarrier spacing configurations use the same AI model; or to indicate the subcarrier spacing applicable to different AI models, for example, using a bitmap to indicate the subcarrier spacing applicable to the AI ​​model supported by each terminal device, with each bit corresponding to a subcarrier spacing. Alternatively, the terminal device can indicate that one AI model is used for frequencies less than or equal to 60kHz, and another AI model is used for frequencies greater than 60kHz. For example, the subcarrier spacing can be predefined as 15kHz, 30kHz, 120kHz, etc.

[0764] In some embodiments, the terminal device can report to the network device in a Bitmap manner whether the AI ​​model used by the terminal device is the same for different transmission bandwidth configurations, carrier frequency configurations, or subcarrier spacing configurations. For example, the terminal device can report 3 bits of information, where the first bit indicates whether different transmission bandwidth configurations use the same AI model, the second bit indicates whether different carrier frequency configurations use the same AI model, and the third bit indicates whether different subcarrier spacing configurations use the same AI model. Among them, 1 indicates using different AI models, and 0 indicates using the same AI model (or vice versa). Based on the information reported by the terminal device, the network device can determine whether the adjustment of the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration will result in the switching of the AI ​​model on the terminal device side.

[0765] 2. The network device receives the generalization capability of the AI ​​model reported by the terminal device. The generalization capability can be represented by the range of the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration corresponding to the AI ​​model.

[0766] 3. Based on the generalization capability, the network device may indicate to the terminal device the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration related to the first AI model.

[0767] The network device can determine the transmission bandwidth configuration based on the range of the transmission bandwidth configuration corresponding to the AI ​​model reported by the terminal device; determine the carrier frequency configuration based on the range of the carrier frequency configuration corresponding to the AI ​​model reported by the terminal device; and determine the subcarrier spacing configuration based on the range of the subcarrier spacing configuration corresponding to the AI ​​model reported by the terminal device.

[0768] If the AI ​​model reported by the terminal device can be used for any transmission bandwidth configuration, the network device can determine that the adjustment of the transmission bandwidth configuration does not need to consider the switching of the AI ​​model.

[0769] If the AI ​​model reported by the terminal device can be used for any carrier frequency configuration, the network device does not need to consider the switching of the AI ​​model when determining the adjustment of the carrier frequency configuration.

[0770] If the AI ​​model used in the terminal device's reporting can be used for any subcarrier spacing configuration, the network device does not need to consider the switching of the AI ​​model when determining the adjustment of the subcarrier spacing configuration.

[0771] If the AI ​​model reported by the terminal device is trained for different values ​​of the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration, the network device needs to consider the AI ​​model switching on the terminal device side that may be caused by the corresponding configuration when indicating the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration. For example, if the AI ​​model is trained for FR1 and FR2 respectively, the network device should not dynamically switch between FR1 and FR2 to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0772] In some embodiments, when the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration is dynamically indicated through DCI signaling, the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration cannot cause dynamic AI model switching on the terminal device side, that is, the parameter value (such as subcarrier spacing) indicated by the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model. This method can also be used when the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration is indicated through MAC CE signaling.

[0773] In some embodiments, the network device can receive the AI ​​model switching capability (i.e., the second capability) reported by the terminal device, which can be used to indicate whether the terminal device supports AI model switching based on DCI signaling, or supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. If the terminal device cannot support AI model switching based on a certain signaling, when the network device uses the signaling to configure the transmission bandwidth, carrier frequency, or subcarrier spacing, the parameter value indicated by the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration needs to correspond to the same AI model as the original parameter value (i.e., the first value). In other words, the value of the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration cannot cause the AI ​​model switching of the terminal device.

[0774] 4. The terminal device receives the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration indicated by the network device, and uses the first AI model corresponding to the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration to transmit with the network device.

[0775] In some embodiments, in order to avoid dynamic AI model switching of the terminal device, when the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration is indicated by DCI signaling, the terminal device does not expect the parameter value indicated by the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration to correspond to a different AI model than the original parameter value (i.e., the first value). In other words, the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration cannot cause dynamic AI model switching on the terminal device side. For example, when the terminal reports different subcarrier spacing configurations using different AI models, if the network device indicates BWP switching through DCI signaling (different BWPs may use different carrier spacings), dynamic switching between different subcarrier spacings cannot be performed, that is, the indicated BWP and the original BWP need to use the same subcarrier spacing. This method can also be used when the transmission bandwidth configuration, carrier frequency configuration or subcarrier spacing configuration is indicated by MAC CE signaling.

[0776] In some embodiments, the terminal device may report the AI ​​model switching capability (i.e., the second capability) to the network device, which may be used to indicate whether the terminal device supports AI model switching based on DCI signaling, or supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. The terminal device does not expect the AI ​​model switching caused by the adjustment of the transmission bandwidth configuration, carrier frequency configuration, or subcarrier spacing configuration to exceed the AI ​​model switching capability reported by the terminal device. For example, if the terminal device does not support AI model switching based on MAC layer signaling and DCI signaling, and the transmission bandwidth configuration is indicated by DCI signaling or MAC layer signaling, the transmission bandwidth configuration cannot enable the terminal device to switch the AI ​​model. For example, it can only switch within the transmission bandwidth range corresponding to the same AI model.

[0777] In some embodiments, if different carrier frequency configurations or subcarrier spacing configurations use different AI models, the transmission between the terminal device and the network device is scheduled through fallback DCI (for example, scheduled through DCI format 0_0 or 1_0), and the carrier frequency configuration or subcarrier spacing configuration corresponding to the fallback DCI is different from the carrier frequency configuration or subcarrier spacing configuration corresponding to the non-fallback DCI, then the terminal device does not use the AI ​​model for transmission between the network device. For example, if the transmission scheduled by the fallback DCI uses a default carrier frequency (such as FR1) or subcarrier spacing (such as 30kHz), and the default carrier frequency or subcarrier spacing is different from the carrier frequency or subcarrier spacing previously indicated by the non-fallback DCI, then for the transmission scheduled by the fallback DCI, the terminal device does not use AI model-based sending or receiving, but uses a non-AI-based sending and receiving method. This is because most terminal devices can support dynamic switching between sending or receiving based on the AI ​​model and sending and receiving based on the non-AI model, but cannot support dynamic switching between different AI models (model deployment takes time).

[0778] Example 7: Cell ID configuration.

[0779] 1. The generalization capability of the AI ​​model used by the terminal device. The generalization capability is represented by the range of cell ID configurations corresponding to the AI ​​model.

[0780] In the embodiments of the present application, the AI ​​model is used to transmit or receive signals (such as reference signals) or data. For example, for downlink transmission, the input of the AI ​​model is the downlink received signal, and the output is the data detection result or channel measurement result. For uplink transmission, the input of the AI ​​model is the data or signal to be transmitted, and the output is the transmit signal after processing by the transmitter.

[0781] Since the terminal device needs to access the cell before sending and receiving data / signals, the signal transmission is all between the terminal device and the service cell. Therefore, the training of the AI ​​model can be based on the cell. When the service cell changes, for example, when the terminal device accesses a cell with a different cell ID, the AI ​​model used may also be different. In other words, the terminal device can train a unified AI model for different cells, or train a different model for each cell, which depends on the generalization capability of the AI ​​model on the terminal device side. Since different terminal devices can adopt different training methods, thereby obtaining AI models with different generalization capabilities, this generalization capability can be reported to the network device at this time, so that the network device can determine whether the cell switching will cause the AI ​​model of the terminal device to switch.

[0782] In some embodiments, the range of the cell ID configuration corresponding to the AI ​​model may include one or more of the following:

[0783] One AI model is used for all cell IDs;

[0784] Different AI models are used for different cell IDs;

[0785] The range of cell IDs applicable to different AI models.

[0786] For example, UE capabilities can use 1-bit information to indicate whether the same AI model is used for data reception when the terminal device accesses cells with different cell IDs. For another example, UE capabilities can indicate the range of cell IDs to which different AI models supported by the terminal device are applicable, that is, each AI model can be used in cells corresponding to a group of cell IDs.

[0787] 2. The network device receives the generalization capability of the AI ​​model reported by the terminal device, where the generalization capability can be represented by the range of the cell ID configuration corresponding to the AI ​​model.

[0788] 3. Based on the generalization capability, the network device can indicate the cell ID configuration related to the first AI model to the terminal device.

[0789] The network device can determine whether cell switching will cause AI model switching on the terminal device side based on the range of cell ID configuration corresponding to the AI ​​model reported by the terminal device.

[0790] If the AI ​​model reported by the terminal device can be used in any cell, the network device determines that the cell switching will not cause the AI ​​model switching on the terminal device side. Therefore, there is no need to reserve time for the terminal device to switch the AI ​​model when performing the handover.

[0791] If the terminal device reports using different AI models for different cells, the network device determines that the cell switching will cause the AI ​​model switching on the terminal device side. Therefore, time needs to be reserved for the terminal device to switch the AI ​​model when performing Handover.

[0792] If the terminal device reports the cell ID ranges applicable to different AI models, the network device can determine whether the cell switching will cause the AI ​​model switching on the terminal device side based on the cell ID of the switched cell and the cell ID of the original cell, so as to consider reserving time for the terminal device to switch the AI ​​model when performing handover.

[0793] 4. The terminal device receives the cell ID configuration indicated by the network device, and uses the first AI model corresponding to the cell ID configuration to transmit with the network device.

[0794] The cell ID configuration may be carried by the synchronization signal of the target cell.

[0795] In some embodiments, the terminal device can determine whether to perform a cell handover and whether the cell handover will cause an AI model switch based on the cell ID configuration indicated by the network device. If the cell handover may cause an AI model switch, the terminal device needs to reserve time for the AI ​​model switch; otherwise, the terminal device can still use the original AI model for data transmission or reception.

[0796] Embodiment 8: MIMO transmission solution configuration.

[0797] 1. The terminal device can report the generalization capability of the AI ​​model used, where the generalization capability is represented by the range of MIMO transmission scheme configuration corresponding to the AI ​​model.

[0798] In the embodiments of the present application, the AI ​​model is used for downlink CSI measurement or for data transmission or reception. For example, for downlink transmission, the AI ​​model input is the downlink received signal, and the output is the data detection result, such as the source bit stream; for uplink transmission, the AI ​​model input is the data to be transmitted, and the output is the transmit signal on each antenna port; for CSI measurement, the AI ​​model input is the downlink received signal, and the output is the CSI measurement result, such as RSRP or PMI bits.

[0799] Since the terminal device sends or receives data, or measures CSI, it is all based on a certain MIMO transmission scheme. Therefore, if the AI ​​model is used for these functions, it needs to be trained based on the assumption of a certain MIMO transmission scheme. The terminal can train a unified AI model for different MIMO transmission scheme configurations, or it can train a different AI model for each MIMO transmission scheme configuration, depending on the generalization capability of the AI ​​model on the terminal device side. In other words, different terminal devices can use different training methods to obtain AI models with different generalization capabilities. At this time, this generalization capability can be reported to the network device, so that the network device can determine whether the adjustment of the MIMO transmission scheme configuration will cause the AI ​​model of the terminal device to switch.

[0800] In some embodiments, the range of MIMO transmission scheme configuration corresponding to the AI ​​model may include one or more of the following:

[0801] One AI model is used for single-TRP transmission scheme and multi-TRP collaboration scheme;

[0802] Different AI models are used for single-TRP transmission solutions and multi-TRP collaboration solutions;

[0803] One AI model is used for both open-loop and closed-loop transmission solutions;

[0804] Different AI models are used for open-loop transmission solutions and closed-loop transmission solutions.

[0805] In an embodiment of the present application, the MIMO transmission scheme may include a single TRP transmission scheme, a multi-TRP collaborative scheme, an open-loop transmission scheme, a closed-loop transmission scheme, etc.

[0806] In an embodiment of the present application, a single TRP transmission scheme or a multi-TRP collaboration scheme can be determined by the number of TCI states configured by a network device for a certain signal or channel, or a special signaling can be used to indicate whether a multi-TRP collaboration scheme is adopted, such as RRC signaling. Among them, the multi-TRP collaboration scheme can include a scheme for multiple TRPs to perform NC-JT, a scheme for multiple TRPs to perform CJT, a scheme for multiple TRPs to perform SFN, a scheme for multiple TRPs to perform Repetition, etc.

[0807] In the embodiment of the present application, the open-loop transmission scheme refers to a transmission scheme that does not require the terminal device to feedback precoding information, and the closed-loop transmission scheme refers to a transmission scheme that requires the terminal device to feedback precoding information (such as PMI).

[0808] In some embodiments, the terminal device can report to the network device in a Bitmap manner whether the AI ​​model used by the terminal device is the same for different MIMO transmission scheme configurations. For example, the terminal device can report 2 bits of information, where the first bit indicates whether the single TRP transmission scheme and the multi-TRP collaboration scheme use the same AI model, and the second bit indicates whether the open-loop transmission scheme and the closed-loop transmission scheme use the same AI model. Among them, 1 indicates the use of different AI models, and 0 indicates the use of the same AI model (or vice versa). Based on the information reported by the terminal device, the network device can determine whether the adjustment of the MIMO transmission scheme configuration will result in the switching of the AI ​​model on the terminal device side.

[0809] 2. The network device receives the generalization capability of the AI ​​model reported by the terminal device, where the generalization capability is represented by the range of MIMO transmission scheme configuration corresponding to the AI ​​model.

[0810] 3. Based on the generalization capability, the network device can indicate the MIMO transmission scheme configuration related to the first AI model to the terminal device.

[0811] The network device can determine the MIMO transmission scheme configuration used by the terminal device for CSI measurement or data transmission or reception based on the range of the MIMO transmission scheme configuration corresponding to the AI ​​model reported by the terminal device.

[0812] If the AI ​​model reported by the terminal device can be used for any MIMO transmission scheme configuration, the network device does not need to consider the switching of the AI ​​model when indicating the MIMO transmission scheme configuration.

[0813] If the AI ​​model reported by the terminal device is trained for different MIMO transmission scheme configurations, the network device needs to consider the possible AI model switching on the terminal device side caused by the MIMO transmission scheme configuration when indicating the MIMO transmission scheme configuration. For example, if the AI ​​model is trained for a single TRP transmission scheme and a multi-TRP collaborative scheme, the network device should not dynamically switch back and forth between the two schemes to avoid dynamic AI model switching on the terminal device side, which exceeds the capabilities of the terminal device side.

[0814] In some embodiments, the network device cannot instruct two signals to be sent or received at the same time, and the MIMO transmission scheme configurations used by the two signals correspond to different AI models, otherwise the terminal device cannot use two AI models to send or receive these signals at the same time.

[0815] In some embodiments, when the MIMO transmission scheme configuration is dynamically indicated via DCI signaling, the MIMO transmission scheme configuration cannot cause dynamic AI model switching on the terminal device side. In other words, the parameter value indicated by the MIMO transmission scheme configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model. This method can also be used when the MIMO transmission scheme configuration is indicated via MAC CE signaling.

[0816] In other embodiments, the network device may receive the AI ​​model switching capability (i.e., the second capability) reported by the terminal device, which may be used to indicate whether the terminal device supports AI model switching based on DCI signaling, or supports AI model switching based on MAC signaling, or supports AI model switching based on RRC signaling. If the terminal device cannot support AI model switching based on a certain signaling, when the network device uses the signaling to configure the MIMO transmission scheme, the parameter value indicated by the MIMO transmission scheme configuration and the original parameter value (i.e., the first value) need to correspond to the same AI model. In other words, the value of the MIMO transmission scheme configuration cannot cause the AI ​​model switching of the terminal device.

[0817] 4. The terminal device receives the MIMO transmission scheme configuratio...

Claims

1. A communication method, applied to a terminal device, the method comprising: Sending a first capability to a network device, the first capability being used to characterize the range of parameter configurations corresponding to one or more AI models supported by the terminal device; Receiving the parameter configuration from the network device, and communicating with the network device based on the parameter configuration.

2. The method according to claim 1, wherein, The range of parameter configurations corresponding to the one or more AI models includes: for different values of the parameter configuration, whether the terminal device uses the same AI model.

3. The method according to claim 1 or 2, wherein The parameter configuration includes the configuration of one or more of the following parameters: Demodulation Reference Signal (DMRS), Channel State Information Reference Signal (CSI-RS), Channel State Information (CSI) reporting content, Modulation and Coding Scheme (MCS), Transport Block Size, Number of Transmission Layers, Transmission Bandwidth, Number of Antenna Ports, Carrier Frequency, Subcarrier Spacing, Transmission Configuration Indicator (TCI) state, Cell Identifier (ID), Multiple-Input Multiple-Output (MIMO) transmission scheme.

4. The method according to any one of claims 1 to 3, wherein When the one or more AI models are used for channel estimation or data reception, the first capability characterizes the range of DMRS configurations corresponding to the one or more AI models.

5. The method according to claim 4, wherein The range of DMRS configurations corresponding to the one or more AI models includes one or more of the following: One AI model corresponds to all DMRS configurations; Different AI models correspond to different DMRS base sequences; Different AI models correspond to different DMRS Code Division Multiplexing (CDM) groups; Different AI models correspond to different DMRS port multiplexing methods; Different AI models correspond to different numbers of DMRS symbols; Different AI models correspond to different DMRS port sets.

6. The method according to any one of claims 1 to 5, wherein, When the one or more AI models are used for CSI measurement, the first capability characterizes the range of CSI-RS configurations corresponding to the one or more AI models, and / or the range of CSI reporting content configurations corresponding to the one or more AI models.

7. The method according to claim 6, wherein, The range of CSI-RS configurations corresponding to the one or more AI models includes one or more of the following: One AI model corresponds to all CSI-RS configurations; Different AI models correspond to different CSI-RS base sequences; Different AI models correspond to different CSI-RS port number sets; Different AI models correspond to different CSI-RS port multiplexing methods; Different AI models correspond to different numbers of CSI-RS symbols; Different AI models correspond to different CSI-RS transmission bandwidths.

8. The method according to claim 6 or 7, wherein The range of CSI reporting content configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for Reference Signal Received Power (RSRP) reporting and CSI reporting; Different AI models are respectively used for RSRP reporting and CSI reporting; One AI model is used for CSI compression and CSI prediction; Different AI models are respectively used for CSI compression and CSI prediction; One AI model is used for CSI compression, CSI prediction, and CSI prediction and compression; Different AI models are respectively used for CSI compression, CSI prediction, and CSI prediction and compression.

9. The method according to any one of claims 1 to 8, wherein In the case where the one or more AI models are used for data transmission or data reception, the first capability characterizes one or more of the following: The range of MCS configurations corresponding to the one or more AI models; The range of transport block size configurations corresponding to the one or more AI models; The range of transmission layer number configurations corresponding to the one or more AI models; The range of transmission bandwidth configurations corresponding to the one or more AI models; The range of antenna port number configurations corresponding to the one or more AI models; The range of carrier frequency configurations corresponding to the one or more AI models; The range of subcarrier spacing configurations corresponding to the one or more AI models; The range of TCI state configurations corresponding to the one or more AI models; The range of cell ID configurations corresponding to the one or more AI models.

10. The method according to claim 9, wherein, The range of MCS configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all MCSs; Different AI models are used for different MCSs; The range of MCS index applicable to different AI models; One AI model is used for all modulation schemes; Different AI models are used for different modulation schemes; The set of modulation schemes applicable to different AI models; One AI model is used for all code rate ranges; The code rate ranges applicable to different AI models.

11. The method according to claim 9 or 10, wherein The range of transport block size configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all transport block sizes; Different AI models are used for different transport block sizes; The range of transport block sizes applicable to different AI models.

12. The method according to any one of claims 9 to 11, wherein, The range of transmission layer number configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all transmission layer numbers; Different AI models are used for different transmission layer numbers; The range of transmission layer numbers applicable to different AI models.

13. The method according to any one of claims 9 to 12, wherein, The range of transmission bandwidth configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all transmission bandwidths; Different AI models are used for different transmission bandwidths; The range of transmission bandwidths applicable to different AI models.

14. The method according to any one of claims 9 to 13, wherein The range of antenna port number configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all antenna port numbers; Different AI models are used for different antenna port numbers; The range of antenna port numbers applicable to different AI models.

15. The method according to any one of claims 9 to 14, wherein, The range of carrier frequency configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all carrier frequencies; Different AI models are used for different carrier frequencies; The range of carrier frequencies applicable to different AI models.

16. The method according to any one of claims 9 to 15, wherein The range of subcarrier spacing configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all subcarrier spacings; Different AI models are used for different subcarrier spacings; The range of subcarrier spacings applicable to different AI models.

17. The method according to any one of claims 9 to 16, wherein, The range of TCI state configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all TCI states; Different AI models are used for different TCI states; An AI model is used for single TCI state and multiple TCI states; Different AI models are respectively used for single TCI state and multiple TCI states.

18. The method according to any one of claims 9 to 17, wherein, The range of cell ID configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all cell IDs; Different AI models are used for different cell IDs; The range of cell IDs applicable to different AI models.

19. The method according to any one of claims 1 to 18, wherein, When the one or more AI models are used for CSI measurement, data transmission or data reception, the first capability characterizes the range of MIMO transmission scheme configurations corresponding to the one or more AI models.

20. The method according to claim 19, wherein The range of MIMO transmission scheme configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for single transmission and reception point (TRP) transmission scheme and multi-TRP cooperation scheme; Different AI models are respectively used for single TRP transmission scheme and multi-TRP cooperation scheme; One AI model is used for open-loop transmission scheme and closed-loop transmission scheme; Different AI models are respectively used for open-loop transmission scheme and closed-loop transmission scheme.

21. The method according to any one of claims 1 to 20, wherein Communicating with the network device based on the parameter configuration includes: Communicating with the network device based on the first AI model corresponding to the value of the parameter configuration; wherein, the first AI model belongs to the one or more AI models.

22. The method according to any one of claims 1 to 21, wherein, The method further includes: Sending a second capability to the network device, the second capability being used to indicate whether the terminal device supports one or more of the following: Switching between AI models based on downlink control information (DCI) signaling; Switching between AI models based on media access control (MAC) signaling; Switching between AI models based on radio resource control (RRC) signaling.

23. The method according to claim 22, wherein, The method further includes: When the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value are different AI models, it is not expected that the switching between the first AI model and the second AI model exceeds the second capability; wherein, the first value is the parameter value before receiving the parameter configuration.

24. The method according to any one of claims 21 to 23, wherein The method further includes: It is not expected that the first AI model corresponding to the value of the parameter configuration includes at least two different AI models.

25. The method according to any one of claims 21 to 24, wherein The method further includes: When the parameter configuration is indicated by DCI signaling and / or MAC signaling, it is not expected that the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value are different AI models; wherein, the first value is the parameter value before receiving the parameter configuration.

26. According to the method according to any one of claims 1 to 20, wherein, When the parameter configuration is indicated by DCI signaling and / or MAC signaling and the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value are different AI models, communication is not performed with the network device based on the first AI model; wherein, the first value is the parameter value before receiving the parameter configuration.

27. According to the method according to any one of claims 1 to 20, wherein, When the first AI model corresponding to the value of the parameter configuration includes at least two different AI models, communication with the network device is not based on the first AI model; wherein, the first value is the parameter value before receiving the parameter configuration.

28. A communication method, applied to a network device, the method includes: Receiving a first capability from a terminal device, the first capability being used to characterize the range of parameter configurations corresponding to one or more AI models supported by the terminal device; Based on the first capability, sending the parameter configuration to the terminal device; wherein, the parameter configuration is used for communication between the terminal device and the network device.

29. The method according to claim 28, wherein The range of parameter configurations corresponding to the one or more AI models includes: for different values of the parameter configuration, whether the terminal device uses the same AI model.

30. The method according to claim 28 or 29, wherein, The parameter configuration includes the configuration of one or more of the following parameters: Demodulation Reference Signal (DMRS), Channel State Information Reference Signal (CSI-RS), Channel State Information (CSI) reporting content, Modulation and Coding Scheme (MCS), Transport Block Size, Number of Transmission Layers, Transmission Bandwidth, Number of Antenna Ports, Carrier Frequency, Subcarrier Spacing, Transmission Configuration Indicator (TCI) state, Cell Identifier (ID), Multiple-Input Multiple-Output (MIMO) transmission scheme.

31. The method according to any one of claims 28 to 30, wherein When the one or more AI models are used for channel estimation or data reception, the first capability characterizes the range of DMRS configurations corresponding to the one or more AI models.

32. The method according to claim 31, wherein, The range of DMRS configurations corresponding to the one or more AI models includes one or more of the following: One AI model corresponds to all DMRS configurations; Different AI models correspond to different DMRS base sequences; Different AI models correspond to different DMRS Code Division Multiplexing (CDM) groups; Different AI models correspond to different DMRS port multiplexing methods; Different AI models correspond to different numbers of DMRS symbols; Different AI models correspond to different DMRS port sets.

33. The method according to any one of claims 28 to 32, wherein, When the one or more AI models are used for CSI measurement, the first capability characterizes the range of CSI-RS configurations corresponding to the one or more AI models, and / or, the range of CSI reporting content configurations corresponding to the one or more AI models.

34. The method according to claim 33, wherein, The range of CSI-RS configurations corresponding to the one or more AI models includes one or more of the following: One AI model corresponds to all CSI-RS configurations; Different AI models correspond to different CSI-RS base sequences; Different AI models correspond to different CSI-RS port number sets; Different AI models correspond to different CSI-RS port multiplexing methods; Different AI models correspond to different numbers of CSI-RS symbols; Different AI models correspond to different CSI-RS transmission bandwidths.

35. The method according to claim 33 or 34, wherein, The range of CSI reporting content configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for Reference Signal Received Power (RSRP) reporting and CSI reporting; Different AI models are respectively used for RSRP reporting and CSI reporting; One AI model is used for CSI compression and CSI prediction; Different AI models are respectively used for CSI compression and CSI prediction; One AI model is used for CSI compression, CSI prediction, and CSI prediction and compression; Different AI models are respectively used for CSI compression, CSI prediction, and CSI prediction and compression.

36. The method according to any one of claims 28 to 35, wherein When the one or more AI models are used for data transmission or data reception, the first capability characterizes one or more of the following: The range of MCS configurations corresponding to the one or more AI models; The range of transport block size configurations corresponding to the one or more AI models; The range of transmission layer number configurations corresponding to the one or more AI models; The range of transmission bandwidth configurations corresponding to the one or more AI models; The range of antenna port number configurations corresponding to the one or more AI models; The range of carrier frequency configurations corresponding to the one or more AI models; The range of subcarrier spacing configurations corresponding to the one or more AI models; The range of TCI state configurations corresponding to the one or more AI models; The range of cell ID configurations corresponding to the one or more AI models.

37. The method according to claim 36, wherein, The range of MCS configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all MCSs; Different AI models are used for different MCSs; The range of MCS indices applicable to different AI models; One AI model is used for all modulation schemes; Different AI models are used for different modulation schemes; The set of modulation schemes applicable to different AI models; One AI model is used for all code rate ranges; The code rate ranges applicable to different AI models.

38. The method according to claim 36 or 37, wherein, The range of transport block size configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all transport block sizes; Different AI models are used for different transport block sizes; The range of transport block sizes applicable to different AI models.

39. The method according to any one of claims 36 to 38, wherein The range of transmission layer number configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all transmission layer numbers; Different AI models are used for different transmission layer numbers; The range of transmission layer numbers applicable to different AI models.

40. The method according to any one of claims 36 to 39, wherein, The range of transmission bandwidth configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all transmission bandwidths; Different AI models are used for different transmission bandwidths; The range of transmission bandwidths applicable to different AI models.

41. The method according to any one of claims 36 to 40, wherein, The range of antenna port number configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all antenna port numbers; Different AI models are used for different antenna port numbers; The range of antenna port numbers applicable to different AI models.

42. The method according to any one of claims 36 to 41, wherein, The range of carrier frequency configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all carrier frequencies; Different AI models are used for different carrier frequencies; The range of carrier frequencies applicable to different AI models.

43. The method according to any one of claims 36 to 42, wherein The range of subcarrier spacing configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all subcarrier spacings; Different AI models are used for different subcarrier spacings; The subcarrier spacing ranges applicable to different AI models.

44. The method according to any one of claims 36 to 43, wherein, The range of TCI state configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all TCI states; Different AI models are used for different TCI states; One AI model is used for single TCI state and multiple TCI states; Different AI models are respectively used for single TCI state and multiple TCI states.

45. The method according to any one of claims 36 to 44, wherein, The range of cell ID configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for all cell IDs; Different AI models are used for different cell IDs; The cell ID ranges applicable to different AI models.

46. The method according to any one of claims 28 to 45, wherein, In the case where the one or more AI models are used for CSI measurement, data transmission, or data reception, the first capability characterizes the range of MIMO transmission scheme configurations corresponding to the one or more AI models.

47. The method according to claim 46, wherein, The range of MIMO transmission scheme configurations corresponding to the one or more AI models includes one or more of the following: One AI model is used for single transmission and reception point (TRP) transmission scheme and multi-TRP cooperation scheme; Different AI models are respectively used for single-TRP transmission scheme and multi-TRP cooperation scheme; One AI model is used for open-loop transmission scheme and closed-loop transmission scheme; Different AI models are respectively used for open-loop transmission scheme and closed-loop transmission scheme.

48. The method according to any one of claims 28 to 47, wherein The value of the parameter configuration corresponds to the first AI model for communication between the terminal device and the network device; wherein, the first AI model belongs to the one or more AI models.

49. The method according to claim 48, wherein The number of the first AI models corresponding to the value of the parameter configuration is one.

50. The method according to any one of claims 28 to 49, wherein The method further includes: Receiving a second capability from the terminal device, the second capability being used to indicate whether the terminal device supports one or more of the following: Switching between AI models based on downlink control information (DCI) signaling; Switching between AI models based on media access control (MAC) signaling; Switching between AI models based on radio resource control (RRC) signaling.

51. The method according to claim 50, wherein Sending the parameter configuration to the terminal device based on the first capability includes: Sending the parameter configuration to the terminal device based on the first capability and the second capability; wherein, In the case where the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value are different AI models, the switching between the first AI model and the second AI model does not exceed the second capability, and the first value is the parameter value before sending the parameter configuration.

52. The method according to any one of claims 48 to 50, wherein, In the case where the parameter configuration is indicated by DCI signaling and / or MAC signaling, the first AI model corresponding to the value of the parameter configuration and the second AI model corresponding to the first value are the same AI model; wherein, the first value is the parameter value before sending the parameter configuration.

53. A terminal device, the device includes: A first communication unit configured to send a first capability to a network device, the first capability being used to characterize the range of parameter configurations corresponding to one or more AI models supported by the terminal device; The first communication unit is further configured to receive the parameter configuration from the network device and communicate with the network device based on the parameter configuration.

54. A network device, the device comprising: A second communication unit, configured to receive a first capability from a terminal device, the first capability being used to characterize a range of parameter configurations corresponding to one or more AI models supported by the terminal device; The second communication unit is further configured to send the parameter configuration to the terminal device based on the first capability; wherein the parameter configuration is used for communication between the terminal device and the network device.

55. A terminal device, the terminal device comprising: A memory for storing a computer program; A processor, connected to the memory, for calling and running the computer program from the memory to implement the method according to any one of claims 1 to 27; A transceiver for receiving and sending information during the process of receiving and sending information between the terminal device and other external devices.

56. A network device, the network device comprising: A memory for storing a computer program; A processor, connected to the memory, for calling and running the computer program from the memory to implement the method according to any one of claims 28 to 52; A transceiver for receiving and sending information during the process of receiving and sending information between the network device and other external devices.

57. A chip, wherein, The chip comprises: A memory for storing a computer program; A processor, connected to the memory, for calling and running the computer program from the memory, so that the device installed with the chip executes the method according to any one of claims 1 to 27, or executes the method according to any one of claims 28 to 52; A transceiver for receiving and sending information during the process of receiving and sending information between the chip and other devices or chips.

58. A computer-readable storage medium, wherein, The computer-readable storage medium stores a computer program, which when executed by at least one processor, implements the method according to any one of claims 1 to 27, or implements the method according to any one of claims 28 to 52.

Images