Communication method and device and storage medium

Abstract

The embodiment of the invention provides a communication method and device and a storage medium, relates to the technical field of communication, and is used for improving the matching accuracy between an information processing unit on a terminal side and an information processing unit on a base station side. The method is applied to a first node, and comprises the following steps: sending a first signaling to a second node, the first signaling comprising an identifier of a first information processing unit; receiving a second signaling sent by a second node, wherein the second signaling is used for determining a target first information processing unit; a target first information processing unit is determined from the first information processing units based on the second signaling.

Description

Technical Field

[0001] The present disclosure relates to the field of communication technologies, and in particular, to a communication method, apparatus, and storage medium. Background Art

[0002] The application of technologies such as artificial intelligence and deep learning in the wireless air interface has received extensive attention and emphasis in the industry. However, limited by the generalization of the artificial intelligence information processing unit (such as a model) and limited computing resources, the information processing unit trained on the terminal side can usually only work in specific configurations, scenarios, sites, or environments, and is highly dependent on the software / hardware environment and mobility characteristics of the terminal side. As a result, in many cases, the information processing unit on the terminal side cannot be accurately matched with the information processing unit on the base station side, resulting in a decline in the performance of the information processing unit. Summary of the Invention

[0003] The present disclosure provides a communication method, apparatus, and storage medium for improving the matching accuracy between the information processing unit on the terminal side and the information processing unit on the base station side.

[0004] To achieve the above object, the present disclosure adopts the following technical solutions:

[0005] In a first aspect, a communication method is provided. The method is applied to a first node and includes:

[0006] Sending a first signaling to a second node, the first signaling including an identifier of a first information processing unit;

[0007] Receiving a second signaling sent by the second node, the second signaling being used to determine a target first information processing unit;

[0008] Based on the second signaling, determining a target first information processing unit from the first information processing units.

[0009] In a second aspect, a communication method is provided. The method is applied to a second node and includes:

[0010] Receiving a first signaling sent by the first node, the first signaling including an identifier of a first information processing unit;

[0011] Sending a second signaling to the first node, the second signaling being used to determine a target first information processing unit.

[0012] In a third aspect, a communication apparatus is provided. The apparatus is applied to a first node and includes:

[0013] A sending unit, configured to send a first signaling to a second node, the first signaling including an identifier of a first information processing unit;

[0014] A receiving unit, configured to receive a second signaling sent by the second node, where the second signaling is used to determine a target first information processing unit;

[0015] A processing unit, configured to determine the target first information processing unit from the first information processing units based on the second signaling.

[0016] In a fourth aspect, a communication device is provided. The device is applied to a second node and includes:

[0017] A receiving unit, configured to receive a first signaling sent by a first node, where the first signaling includes an identifier of a first information processing unit;

[0018] A sending unit, configured to send a second signaling to the first node, where the second signaling is used to determine a target first information processing unit.

[0019] In a fifth aspect, a communication device is provided, including: a processor and a memory; the memory stores instructions executable by the processor; when the processor is configured to execute the instructions, the communication device implements any of the methods provided in the first aspect or the second aspect above.

[0020] In a sixth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores computer instructions, and when the computer instructions run on a computer, the computer executes any of the methods provided in the first aspect or the second aspect.

[0021] In a seventh aspect, a computer program product including computer instructions is provided. When the computer instructions run on a computer, the computer executes any of the methods provided in the first aspect or the second aspect.

[0022] In the embodiments of the present disclosure, the target first information processing unit of the first node is determined based on the second signaling sent by the second node. In this way, the matching accuracy between the information processing unit of the first node and the information processing unit of the second node is improved, and the performance degradation of the information processing unit can be avoided. Description of the Drawings

[0023] The drawings are used to provide a further understanding of the technical solutions of the present invention, and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the technical solutions of the present invention, and do not constitute a limitation to the technical solutions of the present invention.

[0024] Figure 1 It is a schematic diagram of the architecture of a communication system provided by an embodiment of the present disclosure;

[0025] Figure 2 It is a schematic diagram of a bilateral model provided by an embodiment of the present disclosure;

[0026] Figure 3Schematic flowchart of a communication method provided by an embodiment of the present disclosure;

[0027] Figure 4 Schematic diagram of a sorting method for identifiers of a model provided by an embodiment of the present disclosure;

[0028] Figure 5 Schematic diagram of another sorting method for identifiers of a model provided by an embodiment of the present disclosure;

[0029] Figure 6 Schematic diagram of another sorting method for identifiers of a model provided by an embodiment of the present disclosure;

[0030] Figure 7 Schematic diagram of another sorting method for identifiers of a model provided by an embodiment of the present disclosure;

[0031] Figure 8 Schematic diagram of another sorting method for identifiers of a model provided by an embodiment of the present disclosure;

[0032] Figure 9 Schematic diagram of another sorting method for identifiers of a model provided by an embodiment of the present disclosure;

[0033] Figure 10 Schematic diagram of a model training process provided by an embodiment of the present disclosure;

[0034] Figure 11 Schematic flowchart of another communication method provided by an embodiment of the present disclosure;

[0035] Figure 12 Schematic diagram of the composition of a communication device provided by an embodiment of the present disclosure;

[0036] Figure 13 Schematic diagram of the composition of another communication device provided by an embodiment of the present disclosure;

[0037] Figure 14 Schematic diagram of the structure of a communication device provided by an embodiment of the present disclosure. Detailed implementation manners

[0038] Next, the technical solutions in the embodiments of the present disclosure will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present disclosure without creative efforts shall fall within the protection scope of the present disclosure.

[0039] Unless the context requires otherwise, throughout the specification and claims, the term "comprise" and its other forms, such as the third-person singular form "comprises" and the present participle form "comprising", are interpreted in an open, inclusive sense, i.e., "including, but not limited to". In the description of the specification, the terms "one embodiment", "some embodiments", "exemplary embodiments", "example", "specific example", or "some examples", etc., are intended to indicate that the specific features, structures, materials, or characteristics related to the embodiment or example are included in at least one embodiment or example of the present disclosure. The schematic representations of the above terms do not necessarily refer to the same embodiment or example. In addition, the specific features, structures, materials, or characteristics may be included in any one or more embodiments or examples in any appropriate manner.

[0040] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present disclosure, unless otherwise specified, the meaning of "a plurality" is two or more.

[0041] In the embodiments of the present disclosure, words such as "exemplary" or "for example" are used to represent examples, illustrations, or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of the present disclosure should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0042] In addition, the use of "based on" means open and inclusive, because a process, step, calculation, or other action "based on" one or more of the stated conditions or values may, in practice, be based on additional conditions or values beyond those stated.

[0043] The applications of technologies such as artificial intelligence and deep learning in the wireless air interface have received extensive attention and emphasis in the industry. The industrial community is conducting feasibility studies and standard support for artificial intelligence technologies in the wireless air interface, including channel state information feedback, beam management, positioning, etc. based on artificial intelligence / deep learning. Take artificial intelligence-based beam management as an example. In the traditional beam scanning process, the base station configures multiple reference signal resources for beam measurement for the terminal, and these reference signal resources are respectively carried on different downlink transmission beams. The terminal measures these reference signals and reports the beam measurement results to the base station. Since the beams are usually selected from a pre-determined analog beam codebook, exhaustive scanning of all beams in the codebook is an optimal beam training scheme. However, this may lead to excessive training overhead, measurement power consumption, and processing delay. In the artificial intelligence-based beam management method, the base station only needs to transmit reference signal resources in part of the beam space or at part of the time, and uses artificial intelligence algorithms to predict the full beam space information and the optimal beam at all times, thereby effectively reducing the beam training overhead and the terminal measurement power consumption.

[0044] However, limited by the generalization of the artificial intelligence information processing unit (such as the model) and limited computing resources, taking the information processing unit as the model as an example, the model trained on the terminal side usually can only work under specific configurations, scenarios, sites, or environments, and is highly dependent on the software / hardware environment and mobility characteristics on the terminal side. This results in many cases where the second node cannot accurately instruct and operate the model on the terminal side, and further leads to the inability to accurately match the model on the terminal side with the model on the second node, resulting in a decline in the performance of the model. How to improve the matching accuracy between the information processing unit on the terminal side and the information processing unit on the base station side is an urgent problem to be solved.

[0045] Based on this, the embodiments of the present disclosure provide a communication method, device, and storage medium. The target first information processing unit of the first node is determined based on the second signaling sent by the second node. In this way, the matching accuracy between the information processing unit of the first node and the information processing unit of the second node is improved, and the performance degradation of the information processing unit can be avoided.

[0046] The technical solutions provided by the embodiments of the present disclosure can be applied to various mobile communication networks. For example, the new radio (NR) mobile communication network adopting the fifth generation mobile networks (5G), future mobile communication networks, or various communication convergence systems, etc. The embodiments of the present disclosure do not limit this.

[0047] In the embodiments of the present disclosure, the network architecture of a mobile communication network (including but not limited to the third-generation 3G, fourth-generation 4G, fifth-generation 5G, and future mobile communication networks, such as the sixth-generation 6G) may include network-side devices (e.g., including but not limited to base stations) and receiving-side devices (e.g., including but not limited to terminals). It should be understood that, in this example, in the downlink, the first communication node (which may also be referred to as the first communication node device, the first node) may be the second node device, and the second communication node (which may also be referred to as the second communication node device, the second node) may be the terminal-side device. Of course, in the uplink, the first communication node may also be the terminal-side device, and the second communication node may also be the second node device. In device-to-device communication between two communication nodes, both the first communication node and the second communication node may be base stations or terminals. The first communication node and the second communication node may be abbreviated as the first node and the second node, respectively.

[0048] In a wireless communication scenario, the first communication node communicates with the second communication node via a wireless channel. For example, the first communication node is a terminal, the second communication node is a base station, and communication occurs between the base station and the terminal via a wireless channel. Another example is that the first communication node is a terminal, the second communication node is a wireless router, and communication occurs between the wireless router and the terminal via a wireless channel. Another example is that the first communication node is a first base station, the second communication node is a second base station, and communication occurs between the first base station and the second base station via a wireless channel. Another example is that the first communication node is a first terminal, the second communication node is a second terminal, and communication occurs between the first terminal and the second terminal via a wireless channel. Another example is that the first communication node is a repeater, the second communication node is a base station, and communication occurs between the base station and the repeater via a wireless channel. Another example is that the first communication node is a terminal, the second communication node is a repeater, and communication occurs between the repeater and the terminal via a wireless channel. Another example is that the first communication node is a first repeater, the second communication node is a second repeater, and communication occurs between the first repeater and the second repeater via a wireless channel. Another example is that the first communication node is a base station, the second communication node is a satellite, and communication occurs between the satellite and the base station via a wireless channel. Another example is that the first communication node is a satellite, the second communication node is a base station, and communication occurs between the base station and the satellite via a wireless channel. Another example is that the first communication node is a terminal, the second communication node is a satellite, and communication occurs between the satellite and the terminal via a wireless channel. Another example is that the first communication node is a satellite, the second communication node is a terminal, and communication occurs between the terminal and the satellite via a wireless channel. Another example is that the first communication node is a ground device, the second communication node is an aircraft, and communication occurs between the aircraft and the ground device via a wireless channel. Another example is that the first communication node is a first aircraft, the second communication node is a second aircraft, and communication occurs between the first aircraft and the second aircraft via a wireless channel.

[0049] The "first" communication node, "second" communication node, "first" method, "second" method, "first" matrix,

[0050] "second" matrix, "first" part, "second" part are only used for descriptive distinction and do not represent front-back or sequence, nor do they represent superiority or inferiority. That is to say, the "first" and "second" in the embodiments of the present disclosure are only used for descriptive distinction and do not represent front-back or sequence, nor superiority or inferiority.

[0051] Exemplarily, taking the network-side device as a base station and the receiving-side device as a terminal as an example, as Figure 1 shown, it is a schematic diagram of the architecture of a communication system provided by an embodiment of the present disclosure. As Figure 1 shown, the communication system 10 includes multiple base stations (such as base station 21 and base station 22) and multiple terminals (such as terminal 31, terminal 32, terminal 33, and terminal 34). Among them, the multiple base stations and the multiple terminals can be communicatively connected.

[0052] In the present disclosure, the base station can be a base station in Long Term Evolution (LTE), Long Term Evolution Advanced (LTEA), or an evolved Node B (eNB or eNodeB), a base station device in a 5G network, or a base station in a future communication system (such as 6G), etc. The base station can include various macro base stations, micro base stations, home base stations, remote radio heads, Reconfigurable Intelligent Surfaces (RISs), routers, Wireless Fidelity (WIFI) devices, or various network-side devices such as a primary cell and a secondary cell.

[0053] In the present disclosure, a terminal is a device with wireless transceiver capabilities, which can be deployed on land, including indoors or outdoors, handheld, wearable or vehicle-mounted; it can also be deployed on water (such as a ship, etc.); it can also be deployed in the air (such as an airplane, a balloon, a satellite, etc.). The terminal can be a mobile phone, a tablet computer (Pad), a computer with wireless transceiver capabilities, a virtual reality (VR) terminal, an augmented reality (AR) terminal, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, a wireless terminal in smart home, and so on. The embodiments of the present disclosure do not limit the application scenarios. Sometimes, a terminal can also be referred to as a user, a user equipment (UE), an access terminal, a UE unit, a UE station, a mobile station, a mobile unit, a remote station, a remote terminal, a mobile device, a UE terminal, a wireless communication device, a UE agent, or a UE device, etc. The embodiments of the present disclosure do not limit this.

[0054] In some embodiments, the high-layer signaling includes but is not limited to radio resource control (RRC), and media access control-control element (MAC CE), or other high-layer signaling above the physical layer. The physical layer signaling includes but is not limited to: downlink control information and uplink control information. As an example, between a base station and a terminal, physical layer signaling can be transmitted on a physical downlink control channel (PDCCH), physical layer signaling can be transmitted on a physical uplink control channel (PUCCH), data can be transmitted on a physical downlink shared channel (PDSCH), and data can be transmitted on a physical uplink shared channel (PUSCH).

[0055] In some embodiments, an indicator of a parameter, which may also be referred to as an index or an identifier (ID), is an equivalent concept among an indicator, an identifier, and an index. For example, a resource identifier of a wireless system may also be referred to as a resource indicator or a resource index. Among them, the resource identifier of the wireless system includes, but is not limited to, one of the following: reference signal resources, reference signal resource groups, reference signal resource configurations, channel state information (CSI) reports, CSI report sets, terminals, base stations, panels, neural networks, sub-neural networks, identifiers corresponding to neural network layers, etc. The base station may indicate the identifier of one or a group of resources to the terminal through various high-layer signaling or physical layer signaling. The terminal may feedback the identifier of one or a group of resources to the base station through various high-layer signaling and / or physical layer signaling.

[0056] In some embodiments, a time slot may be a slot or a mini slot. A time slot or a mini slot includes at least one symbol. A symbol refers to a time unit in a subframe, a frame, or a time slot. For example, it may be an orthogonal frequency division multiplexing (OFDM) symbol, a single-carrier frequency division multiple access (SC-FDMA) symbol, an orthogonal frequency division multiple access (OFDMA) symbol, etc.

[0057] In some embodiments, in order to calculate channel state information or perform channel estimation, mobility management, positioning, etc., it is necessary for the base station or user to send reference signals (RS). The reference signals include, but are not limited to, channel-state information reference signals (CSI-RS), which include zero power CSI-RS (ZP CSI-RS) and non-zero power CSI-RS (NZP CSI-RS), channel-state information-interference measurement (CSI-IM), sounding reference signal (SRS), synchronization signals block (SSB), physical broadcast channel (PBCH), synchronization signals block / physical broadcast channel (SSB / PBCH). NZP CSI-RS can be used to measure the channel or interference, and CSI-RS can also be used for tracking, called CSI-RS for Tracking (TRS), while CSI-IM is generally used to measure interference and SRS is used to measure the uplink channel. In addition, the time-frequency resources used to transmit reference signals include a set of resource elements (RE) called reference signal resources, such as CSI-RS resource, SRS resource, CSI-IM resource, SSB resource; the time-frequency resources also include resource element groups (REG), physical resource blocks (PRB), resource block groups (RBG), wideband, and subbands. In this article, SSB includes synchronization signals block and / or physical broadcast channel.

[0058] In some embodiments, in order to save signaling overhead, etc., multiple reference signal resources may be divided into multiple sets (such as CSI-RS resource set, CSI-IM resource set, SRS resource set). A reference signal resource set includes at least one reference signal resource, and multiple reference signal resource sets may all come from the same reference signal resource setting (such as CSI-RS resource setting, SRS resource setting, where CSI-RS resource setting may be combined with CSI-IM resource setting and both are referred to as CSI-RS resource setting) to configure parameter information.

[0059] In some embodiments, a beam includes a transmit beam, a receive beam, a receive beam and a transmit beam pair, a transmit beam and a receive beam pair. The beam index may be replaced by a resource index (such as a reference signal resource index) because a beam may be transmission-bound with some time-frequency code resources. A beam may also be a transmission (transmit / receive) mode; the transmission mode may include spatial division multiplexing, frequency domain / time domain diversity, beamforming, etc. In some embodiments, a beam pair includes a combination of a transmit beam and a receive beam.

[0060] In some embodiments, a beam is equivalent to a beam state, a quasi-co-location (QCL) state, a transmission configuration indicator (TCI) state, a spatial relation, spatial relation information, a reference signal (RS), a reference signal resource, a spatial filter, and precoding. In some embodiments, a transmit beam is equivalent to a QCL state, a TCI state, a spatial relation, a spatial relation state, an uplink / downlink reference signal (such as CSI-RS, SSB, DMRS, SRS, PRACH), a transmit spatial filter, and transmit precoding. In some embodiments, a receive beam is equivalent to a QCL state, a TCI state, a spatial relation, a spatial relation state, a spatial reception parameter, a spatial filter, a receive spatial filter, and receive precoding. Among them, the spatial filter, also known as the spatial domain filter, can be either on the second node or on the UE side. In some embodiments, the base station can perform a Quasi co-location (QCL) configuration for two reference signals and inform the user equipment to describe the channel characteristic assumptions. The parameters involved in the said quasi co-location at least include: Doppler spread, Doppler shift, delay spread, average delay, average gain, and spatial parameters (Spatial Rx parameter, or Spatial parameter); among them, the spatial parameters can include spatial reception parameters, angle information, spatial correlation of the receive beam, average delay, and correlation of the time-frequency channel response (including phase information). The angle information can include at least one of the following: angle of arrival (AOA), angle of departure (AOD), zenith angle of departure (ZOD), and zenith angle of arrival (ZOA). The spatial domain filtering can be at least one of the following: a DFT vector, a precoding vector, a DFT matrix, a precoding matrix, or a vector composed of a linear combination of multiple DFTs, a vector composed of a linear combination of multiple precoding vectors. In some embodiments, the concepts of vector and vector can be interchanged.

[0061] In some examples, in order to better transmit data or signals, a base station or a terminal needs to obtain measurement parameters. The measurement parameters may include channel state information or other parameters for characterizing a channel. Among them, the channel state information may include at least one of the following: Channel State Information-Reference Signal Resource Indicator (CSI-RS resource indicator, CRI), Synchronization Signals Block Resource Indicator (SSBRI), Layer 1 Reference Signal Received Power (L1 reference signal received power, L1-RSRP or RSRP), Differential RSRP (Differential RSRP); Layer 1 Signal to Interference Noise Ratio (L1 signal to interference noise ratio, L1-SINR or SINR), Differential L1-SINR (Differential L1-SINR); Reference Signal Received Quality (RSRQ), Channel Quality Indicator (CQI), Precoding Matrix Indicator (PMI), Layer Indicator (LI), Rank Indicator (RI), and related precoding information. The precoding information includes a first type of precoding information, such as precoding information based on a codebook. Here, the precoding matrix indicator is one of the precoding information based on a codebook. The precoding information also includes a non-codebook implementation method. For example, a second type of precoding information, such as precoding information obtained based on advanced technologies such as artificial intelligence.

[0062] In some embodiments, the beam parameter information may also be referred to as beam quality information, or channel measurement result, or beam measurement result, or measurement result, or measurement parameter. In some embodiments, the beam parameter information is a subset of the channel state information, that is, the beam parameter information is the channel state information, and the channel state information belongs to the measurement parameters. In some embodiments, the measurement parameters, channel state information, and beam parameter information all belong to the measurement result, or processing result, or generation result.

[0063] In some embodiments, to transmit channel state information at the physical layer, the terminal and the base station define a CSI report (CSI report or CSI report congfig), where the CSI report defines at least one of the following parameters: time-frequency resources for feedbacking CSI, the report quality (report Quantity) included in the CSI, the time-domain category (reportConfigType) of CSI feedback, channel measurement resources, interference measurement resources, information such as the bandwidth size of the measurement, etc. The CSI report can be transmitted on the uplink transmission resources, where the uplink transmission resources include PUSCH and PUCCH, and the CSI report also includes time-domain characteristics, including periodic CSI report (periodic CSI report, P-CSI), aperiodic CSI report (aperiodic CSI report, AP-CSI), and semi-persistent CSI report (semi-persistent CSI report, SP-CSI).

[0064] In some embodiments, the base station configures NC CSI reports (CSI report) that need to be feedbacked to the base station to the terminal through higher-layer signaling and / or physical-layer signaling. Each CSI report has an identity (ID), called CSIreportID. The terminal can select MC CSI reports from the NC CSI reports according to its own computing power or processing power, as well as the requirements of the base station. And according to the uplink feedback resources, feedback at least one CSI report among the MC CSI reports, where NC and MC are positive integers, and MC <= NC. In one example, MC CSI reports need to be feedbacked, but the feedback resources of at least two of the MC reports conflict. The conflict of the feedback resources of the two reports means that at least one symbol and / or at least one subcarrier in the transmission resources (such as PUCCH or PUSCH) corresponding to the feedback of the two reports are the same. In some embodiments, feedbacking CSI can also be referred to as transmitting CSI or sending CSI. For example, the channel state information is carried on the uplink transmission resources for feedback or transmission. The uplink transmission resources and the corresponding CSI are both indicated by a channel state information report. In some embodiments, feedbacking or transmitting a CSI report means feedbacking the channel state information configured by the CSI report. In some embodiments, feedbacking or transmitting a CSI report means transmitting the content that needs to be transmitted configured by the CSI report through the transmission resources.

[0065] In some embodiments, artificial intelligence (AI) includes machine learning (ML), deep learning, reinforcement learning, transfer learning, deep reinforcement learning, meta-learning, etc., which are devices, components, software, and modules with self-learning capabilities. In some embodiments, artificial intelligence is implemented through an artificial intelligence network (or neural network). The neural network includes multiple layers, and each layer includes at least one node. In one example, the neural network includes an input layer, an output layer, and at least one hidden layer. Each layer of the neural network includes, but is not limited to, at least one of a fully connected layer, a dense layer, a convolutional layer, a transposed convolutional layer, a direct connection layer, an activation function, a normalization layer, a pooling layer, etc. In some embodiments, each layer of the neural network may include a sub-neural network, such as a Residual Network block (or Resnet block), a Densenet Block, a Recurrent Neural Network (RNN), a Transformer, etc. The artificial intelligence network can be implemented through a model, where the model may include a neural network model. The neural network model includes a neural network model structure and / or neural network model parameters. Here, the neural network model structure can be abbreviated as the model structure, and the neural network model parameters can be abbreviated as network parameters or model parameters. A model structure defines the architecture of the neural network, including the number of layers, the size of each layer, the activation function, the connection situation, the convolution kernel and its size, the convolution step, the type of convolution (such as 1D convolution, 2D convolution, 3D convolution, dilated convolution, transposed convolution, separable convolution, grouped convolution, depthwise convolution, etc.). The network parameters are the weights and / or biases of each layer in the neural network model and their values. A model structure can correspond to multiple sets of different neural network model parameter values to adapt to different scenarios. The neural network model parameters are obtained through online training or offline training. For example, by inputting at least one sample and label, the neural network model is trained to obtain the neural network model parameters.

[0066] In some embodiments, a model refers to a general term used to describe a processing method, function, feature, or group of features that a terminal can execute. In some embodiments, a model is equivalent to a function / functionality, a functional module, a functional entity, a processing method, an information processing method, an implementation, a feature, a feature group, an information processing unit, an information processing manner, etc. In some embodiments, each model corresponds to a model indicator (Model indicator, Model ID) or a functionality indicator or a model identity (Model identity, Model ID) or a functionality identity. In some embodiments, the model identity may also have one of the following other equivalent names or concepts: model index, first identity, functionality identity, model indicator, etc.

[0067] In some embodiments, a model refers to a data stream from the original input of a sample to the output target passing through multiple linear or non-linear components. The said model includes a neural network model, a non-artificial intelligence module for processing information or its corresponding model, and a functional component or function that maps input information to output information (where the mapping includes linear mapping and non-linear mapping).

[0068] In some examples, a model includes a model structure and model parameters. For example, if the model is a neural network model, the neural network model includes a neural network model structure and neural network model parameters, which are respectively used to describe the structure of the neural network and the parameter values of the neural network. A neural network model structure can correspond to multiple neural network model parameters, that is, the neural network model structures can be the same, but the corresponding neural network model parameter values can be different.

[0069] In some examples, a model includes a one-sided model and a two-sided model. For example, a one-sided model is a model deployed / inferred and calculated on the terminal side or the second node, and a two-sided model is a model deployed on the terminal side and the second node respectively. The models on both sides need to cooperate to perform inference calculation (such as the structure of an autoencoder (AE)). Exemplarily, the two-sided model can be as Figure 2 shown.

[0070] In some embodiments, the slash symbol ' / ' represents 'and'. In some embodiments, the slash symbol ' / ' represents 'or'. In some embodiments, the slash symbol ' / ' represents 'and / or'.

[0071] The communication method provided by the embodiments of the present disclosure is applicable to a communication system including at least one first node and at least one second node. Hereinafter, taking the first node as a terminal and the second node as a base station, the communication method provided by the embodiments of the present disclosure will be introduced.

[0072] Next, as Figure 3 shown, the embodiments of the present disclosure provide a communication method, which is applied to the first node. The first node may be the terminal 31 shown above Figure 1 shown, and the method may include the following steps:

[0073] S101. Send a first signaling to the second node.

[0074] In a frequency division duplexing (FDD) communication system, since the uplink channel and the downlink channel are at different frequency points, the reciprocity of the two channels is weak. The second node (i.e., the base station) cannot obtain the downlink channel condition based on the uplink channel condition obtained by measuring the uplink reference signal. Therefore, the first node (i.e., the terminal) needs to measure the downlink reference signal sent by the second node to estimate the downlink channel and feedback the channel state information (CSI) to the second node for the second node to obtain the downlink channel state information. Usually, considering the limited uplink transmission resources, the first node quantizes and compresses the complete CSI through a codebook or other means, and feeds back the CSI to the second node with less feedback overhead. However, in this case, some information of the CSI will be lost, resulting in a decrease in the accuracy of the feedback CSI. Therefore, how to feedback the CSI with higher accuracy without significantly increasing the uplink feedback overhead is an urgent problem to be solved. Based on this, in the CSI compression feedback method based on artificial intelligence, the first node quantizes and compresses the channel measurement result into AI-based CSI through artificial intelligence / machine learning / deep learning models or other means and feeds it back to the second node. The second node further restores the original channel measurement result as much as possible through artificial intelligence / machine learning / deep learning models or other means for the second node to use. The information processing unit (such as a model) of the first node needs to match the information processing unit of the second node to achieve high-precision CSI restoration. At the same time, this method can also further reduce the CSI feedback overhead of the first node.

[0075] Taking the information processing unit as an example of the model, since the model inputs, model outputs, required reference signal resource configurations, CSI reporting configurations, model applicable scenarios, etc. corresponding to different models may all be different. Therefore, before activating the model of the first node, the first node needs to indicate the model-related information to the second node, so as to achieve a consistent understanding of the model between the second node and the first node and lay the foundation for subsequent model matching. This model-related information may include some static information (required reference signal resource configuration, CSI reporting configuration, etc.) or dynamic information (scenarios, sites, dataset information, etc. corresponding to the model during training), as well as the pairing indication information of the model (for example: model ID / pairing ID / dataset ID / training period / stage ID, etc. information). Among them, the static information and the pairing indication information of the model can be reported to the second node through the capability report of the first node, but are not limited to this. It should be noted that the information processing unit in the embodiments of the present disclosure may also have one of the following other equivalent names or concepts, for example, information processing method, information processing approach, model, etc.

[0076] In some embodiments, in order to improve the matching accuracy between the information processing unit on the terminal side and the information processing unit on the base station side, that is, to improve the matching accuracy between the information processing unit of the first node and the information processing unit of the second node, when the first node determines the information processing unit to be used by the first node, the first node sends a first signaling to the second node, and the first signaling includes the identifier of the first information processing unit. Correspondingly, the second node receives the first signaling sent by the first node.

[0077] In some embodiments, the first information processing unit can be understood as the information processing unit supported by the first node. It should be understood that the first signaling sent by the first node to the second node includes the identifier of the first information processing unit supported by the first node, so that the second node can determine the first information processing unit supported by the first node based on the identifier of the first information processing unit included in the first signaling, and then indicate the information processing unit that the first node should use based on the first information processing unit supported by the first node, thereby improving the matching accuracy between the information processing unit of the first node and the information processing unit of the second node.

[0078] In some embodiments, the identifier of the first information processing unit is used to uniquely indicate a first information processing unit. For example, it can be the name of the first information processing unit. The identifier of the first information processing unit included in the first signaling can also be replaced by the indication information of the first information processing unit.

[0079] In some embodiments, the identifier of the first information processing unit is a single-layer structure, and the identifier of the first information processing unit is indicated, allocated or configured by the second node. For example, taking the first information processing unit as UE_Model#1 as an example, assuming that UE_Model#1 is trained based on the data set Dataset#1, the identifier of the first information processing unit can be represented by UE_ID#1, indicating that UE_Model#1 is trained under Dataset#1. For another example, taking the first node supporting multiple first information processing units UE_Model#1 and UE_Model#2 under a certain scenario / configuration / function / feature / feature group / implementation method as an example, wherein the first information processing units UE_Model#1 and UE_Model#2 are both trained in the same training period / stage Session#1, the identifier of UE_Model#1 can be represented as UE_ID#1, indicating that it is UE_Model#1 trained under Session#1.

[0080] In some embodiments, the sorting method of the identifiers of the first information processing unit is related to the sorting method of the identifiers of the data sets corresponding to the first information processing unit. Figure 4 FIG. 1 is a schematic diagram of a method for sorting the identifiers of a model provided in an embodiment of the present disclosure. It should be noted that the method for sorting the identifiers of the first information processing unit is not limited to Figure 4 In some embodiments, the sorting method of the first information processing unit identifiers is related to the generation order of the first information processing unit. Figure 5 As shown in FIG. 1 , it is a schematic diagram of another method for sorting the identifiers of a model provided in an embodiment of the present disclosure. Figure 5 As shown in Figure 1, when a new model is generated (developed), the model's identity document (ID) / instruction information is added to the last number of the existing ID by 1. Figure 5 , ID#1 to ID#6 have been sorted according to the model development order. When a new model #2 is developed based on data set #3, the identifier of model #2 is set to ID#7 according to the development order.

[0081] In some embodiments, the sorting method of the identifiers of the first information processing unit is related to the sorting method of the identifiers of the training phase corresponding to the first information processing unit. Taking the information processing unit as an example, for example, Figure 6 FIG. 1 is a schematic diagram of another method for sorting the identifiers of a model provided in an embodiment of the present disclosure. It should be noted that the method for sorting the identifiers of the first information processing unit is not limited to Figure 6For the sorting method shown, the sorting method of the identifiers of the first information processing units can also be sorted according to the generation order of the first information processing units. When a new first information processing unit is generated, the identifier of the new first information processing unit can be incremented by 1 based on the identifier of the last existing first information processing unit. Exemplarily, as Figure 7 shown, it is a schematic diagram of the sorting method of the identifiers of another model provided by an embodiment of the present disclosure. According to the generation order of the models, ID#1 to ID#6 have been sorted. When a new model#2 is newly generated based on dataset#3, the identifier of the new model is designated as ID#7.

[0082] In some embodiments, when an already developed information processing unit (i.e., a model) is updated (for example: the parameters of the information processing unit or the structure of the information processing unit are updated, etc.), the identifier / indication information ID of the information processing unit can include but is not limited to a new identifier / indication information, that is, a new identifier / indication information is assigned according to the generation order of the information processing units, or the current identifier / indication information can remain unchanged, etc.

[0083] In some embodiments, when multiple first information processing units are trained and developed based on a subset of a dataset, the multiple first information processing units can belong to the same Dataset_ID or different Dataset_IDs. Further, when the multiple first information processing units belong to different Dataset_IDs, the first node needs to inform the second node which subset of the dataset each first information processing unit is trained on to keep the understanding of the conditions for developing the information processing units consistent on both sides (i.e., the first node side and the second node side), and it can also help the second node select a suitable second node side information processing unit.

[0084] In some embodiments, the identifier of the first information processing unit has a multi-layer structure. Taking the first information processing unit as a model as an example, as Figure 8 shown, it is a schematic diagram of the sorting method of the identifiers of another model provided by an embodiment of the present disclosure. Refer to Figure 8 , the identifier of model#1 can be expressed as ID#1-1, indicating that model#1 is trained and completed by dataset#1. Or, as Figure 9 shown, it is a schematic diagram of the sorting method of the identifiers of another model provided by an embodiment of the present disclosure. Refer to Figure 9 , the identifier of model#1 can be expressed as ID#1-1, indicating that model#1 is trained and completed in the Session#1 stage.

[0085] In some embodiments, the identifier of the first information processing unit may be determined based on the identifier set information corresponding to the first information processing unit and the identifier information of the first information processing unit. The identifier set information includes at least one of the following: the identifier information of the data set corresponding to the first information processing unit, the identifier information of the training phase corresponding to the first information processing unit, and the description information of the first information processing unit.

[0086] As a possible example, take the case where the identifier of the first information processing unit is determined based on the identifier information of the data set corresponding to the first information processing unit and the identifier information of the first information processing unit. For example, the identifier of the first information processing unit is composed of the identifier information of the data set corresponding to the first information processing unit + the identifier information of the first information processing unit. The identifier of UE_Model#1 can be UE_ID#1-1, indicating UE_Model#1 trained under Dataset#1 by the first information processing unit. It should be noted that in order for the second node and the first node to have a common understanding of the identifier of the information processing unit and to better manage the first information processing unit on the first node side, the assignment of the identifier of the first information processing unit may include, but is not limited to, being assigned by the second node.

[0087] In some embodiments, when there is only one first information processing unit under a certain data set, the first node may only feedback the identifier Dataset_ID of the data set and does not need to feedback the identifier information of the first information processing unit; the second node may also only indicate Dataset_ID when indicating the first information processing unit. For example, when there is only 1 first information processing unit under Dataset#3, the first node or the second node may only need to feedback Dataset_ID#3. When the first node supports all the first information processing units under a certain data set, it may also only feedback Dataset_ID and does not need to feedback the identifier information of the first information processing unit. For example, when the first node supports all 3 first information processing units under Dataset#1, the first node only needs to feedback Dataset_ID#1. However, since there are 3 first information processing units under Dataset#1, the second node still needs to indicate the specific first information processing unit when indicating which first information processing unit for the first node to use. Combining the above Figure 8 , assuming that the second node indicates to use the second model under data set #1 for the first node, the second node may indicate ID#1-2 to indicate that the first node uses the second model in Dataset#1, that is, Model#2.

[0088] As another possible example, take the case where the identifier of the first information processing unit can be determined based on the identifier information of the training phase corresponding to the first information processing unit and the identifier information of the first information processing unit. For example, the identifier of the first information processing unit is composed of the identifier information of the training phase + the identifier information of the first information processing unit. That is, the identifier of UE_Model#1 can be UE_ID#1-1, indicating UE_Model#1 trained under Session#1.

[0089] In some embodiments, when there is only one first information processing unit in a certain training phase, the first node can only feedback the Session_ID and does not need to feedback the identifier information of the first information processing unit; when the second node indicates this first information processing unit, it can also only indicate the Dataset_ID. For example, when there is only 1 first information processing unit under Session#3, the first node or the second node can only feedback Session_ID#3. When the first node supports all the first information processing units in a certain training phase, the first node can also only feedback the Session_ID and does not need to feedback the identifier information of the first information processing unit. For example, when the first node supports all 3 first information processing units under Session#1, the first node only needs to feedback Session_ID#1. However, since there are 3 first information processing units under Session#1, when the second node indicates which first information processing unit for the first node to use, it still needs to indicate to the specific first information processing unit. Combining the above Figure 9 , assuming that the second node indicates to use the second model under Session#1 for the first node, the second node can indicate ID#1-2 to indicate that the first node uses the second model in Session#1, that is, model#2.

[0090] It should be noted that: The second node can indicate the first information processing unit that the first node should use according to some relevant information (the relevant information can be but is not limited to information such as scenario / configuration / implementation method / auxiliary information provided by the first node (such as: UE speed, movement trajectory, etc.)), and can include but is not limited to the first information processing unit with the best performance. Taking the information processing unit as a model as an example, for example: Model A is a model trained based on an indoor scene dataset, and Model B is a model trained based on an indoor and outdoor mixed scene dataset. Of course, Model A has better performance in an indoor environment than Model B, but Model B can have a certain generalization ability in an outdoor scene and can work in an outdoor scene, while Model A may not be able to work in an outdoor environment. Therefore, when the movement trajectory of the first node is about to go out of the indoor to the outdoor, the second node can select Model B as the model that the first node should use according to some trajectory information of the first node, so that the performance will not suddenly drop.

[0091] In some embodiments, the first signaling may include, but is not limited to, the capability report of the first node.

[0092] In some embodiments, the first signaling may include at least one of the following: UCI signaling, MAC CE signaling, and high-layer signaling.

[0093] In some embodiments, when the first node sends the first signaling to the second node, it may be that the first node dynamically sends the first signaling to the second node.

[0094] S102. Receive the second signaling sent by the second node.

[0095] Wherein, the second signaling may include at least one of the following: RRC signaling, MAC CE signaling, and DCI signaling.

[0096] In some embodiments, after the second node receives the first signaling sent by the first node, in response to the first signaling, it sends the second signaling to the first node. Correspondingly, the first node receives the second signaling sent by the second node. Wherein, the second signaling is used to determine the target first information processing unit, and the target first information processing unit can be understood as the information processing unit that the first node should use.

[0097] In some embodiments, the content included in the second signaling is related to the content included in the first signaling. The specific content included in the first signaling and the second signaling can refer to the relevant description in step S103 below, and will not be elaborated here.

[0098] S103. Based on the second signaling, determine the target first information processing unit from the first information processing units.

[0099] In some embodiments, there is one or more first information processing units, that is, the number of information processing units supported by the first node is one or more. When the number of information processing units supported by the first node is one, that is, when the first signaling includes the identifier of one first information processing unit, after the second node receives the first signaling, based on the identifier of the first information processing unit included in the first signaling, it determines the second information processing unit that matches the first information processing unit, and then sends the second signaling to the first node. The second signaling includes the identifier of the target first information processing unit, and the identifier of the target first information processing unit is also the identifier of the first information processing unit supported by the first node, to instruct the first node to use one first information processing unit supported by the first node to perform subsequent inference calculation work.

[0100] In some embodiments, the second information processing unit matches the first information processing unit and may be obtained by jointly training the first information processing unit and the second information processing unit. That the first information processing unit and the second information processing unit are obtained by joint training may be that after the second node jointly trains the first information processing unit and the second information processing unit on the second node side, the first information processing unit is transmitted to the first node; or that in the same training stage, the first node trains the first information processing unit on the first node side and the second node trains the second information processing unit on the second node side; or that after the second node trains the second information processing unit on the second node side, the data set generated during the training of the second information processing unit is sent to the first node, and the first node trains the first information processing unit based on the data set generated during the training of the second information processing unit sent by the second node to obtain the trained first information processing unit, or that the first node sends the data set generated during the training of the first information processing unit on the first node side to the second node, and the second node trains the second information processing unit based on the data set generated during the training of the first information processing unit sent by the first node. It should be noted that the training period may be the offline stage or the online stage, and the training order may be determined through offline negotiation or configured by the second node, etc. The embodiments of the present disclosure do not limit this.

[0101] Taking the information processing unit as the model, the identifier of the information processing unit as the indication information of the model, and the first signaling as the capability report of the first node as an example, as a possible example, for a certain scenario / configuration / function / feature / feature group / implementation manner, the first node supports the model UE_Model#1, and reports the indication information UE_ID#1 of the model to the second node through the capability report. After receiving the indication information UE_ID#1 of the model, the second node selects / confirms the matching second node side model NW_Model#1. The reason why the two match may be that UE_Model#1 and NW_Model#1 are paired models jointly trained offline. Therefore, after the second node confirms the match, it instructs the first node to use the model UE_Model#1 for subsequent inference calculation work.

[0102] As another possible example, for a certain scenario / configuration / function / feature / feature group / implementation, the first node supports the model UE_Model#1. This model UE_Model#1 is trained based on the dataset Dataset#1 sent by the second node. Therefore, the first node reports the indication information Dataset_ID#1 of this model to the second node through the capability report. After receiving the model indication information Dataset_ID#1, the second node selects / confirms the matching second-node model NW_Model#1. The reason they match could be that the dataset Dataset#1 is generated during the training process of the second-node model NW_Model#1. Therefore, after the second node confirms the match, it instructs the first node to use the model UE_Model#1 for subsequent inference calculation work.

[0103] As another possible example, for a certain scenario / configuration / function / feature / feature group / implementation, the first node supports the model UE_Model#1. This model UE_Model#1 is trained based on the training session / stage Session#1. Therefore, the first node reports the indication information Session_ID#1 of this model to the second node through the capability report. After receiving the model indication information Session_ID#1, the second node selects / confirms the matching second-node model NW_Model#1. The reason they match could be that UE_Model#1 and NW_Model#1 are paired models jointly trained during the offline training session / stage Session#1. Therefore, after the second node confirms the match, it instructs the first node to use the model UE_Model#1 for subsequent inference calculation work.

[0104] In some embodiments, the second node jointly trains the first information processing unit and the second information processing unit on the second-node side. It can be that the second node trains the first information processing unit corresponding to each of the multiple first nodes for the multiple first nodes, and trains the second information processing unit corresponding to the second node, and then sends the first information processing unit corresponding to each first node to each first node, that is, the second node trains the information processing units one-to-many; it can also be that the second node trains the first information processing unit corresponding to one first node, and trains the second information processing unit corresponding to the second node, and then sends the trained first information processing unit to the first node, that is, the second node trains the information processing units one-to-one.

[0105] In some embodiments, when the number of information processing units supported by the first node includes multiple, that is, when the first signaling includes the identifiers of multiple first information processing units, several examples are combined below to illustrate the determination of the target first information processing unit from the first information processing units based on the second signaling.

[0106] Example 1: The first signaling further includes first information and second information, and the second signaling includes the identifier of the target first information processing unit.

[0107] Among them, the second information is obtained by the first node based on the processing of the first information by the first information processing unit, and the target first information processing unit is determined by the second node from the first information processing units based on the performance evaluation index between the third information and the first information. The third information is obtained by the second node based on the processing of the second information by the second information processing unit.

[0108] In some embodiments, the first information can be understood as the original channel state information. The first information can also have other names, such as initial channel state information, target channel state information, note channel state information, etc. The first information can include the information processed by possible information preprocessing methods. The second information can be understood as the channel state information compressed by the first information processing unit, and the third information can be understood as the channel state information decompressed (restored) based on the second information processing unit.

[0109] As a possible example, when the first signaling further includes first information and second information, after receiving the first signaling, the second node can determine the multiple first information processing units supported by the second node based on the identifiers of the first information processing units included in the first signaling, and then determine the second information processing units matching each first information processing unit based on the multiple first information processing units. Then the second node inputs the second information into the multiple second information processing units to perform decompression processing on the second information, obtaining multiple third information. Then the second node determines the target first information processing unit from the first information processing units based on the performance evaluation index between each third information and the first information corresponding to each third information. After the second node determines the target first information processing unit from the first information processing units, the second node can send the second signaling to the first node, and the second signaling includes the identifier of the target first information processing unit.

[0110] Based on this, the first node determines the target first information processing unit from the first information processing units based on the second signaling. Specifically, the first node can determine the target first information processing unit based on the identifier of the target first information processing unit included in the second signaling, that is, the first information processing unit corresponding to the identifier of the target first information processing unit in the first information processing units is used as the target first information processing unit.

[0111] It should be noted that a third piece of information is obtained after the second node processes the second information based on a second information processing unit, and a performance evaluation metric is obtained based on a third piece of information and a first piece of information. That is, a third piece of information corresponds to a second information processing unit and a performance evaluation metric, and a second information processing unit matches a first information processing unit. Based on this, the target first information processing unit can be the first information processing unit that matches the second information processing unit corresponding to the optimal performance evaluation metric among the performance evaluation metrics between each third piece of information and the first piece of information corresponding to each third piece of information.

[0112] In some embodiments, the above-mentioned performance evaluation metrics include the first type of evaluation metric, the second type of evaluation metric, and the third type of evaluation metric; among them, the first type of evaluation metric includes the evaluation metric based on distance error and the evaluation metric based on similarity; the evaluation metric based on distance error includes at least one of the following: Euclidean distance (ED), mean squared error (MSE), and normalized mean squared error (NMSE); the evaluation metric based on similarity includes at least one of the following: cosine similarity (CS), generalization cosine similarity (GCS), and cross entropy (CE);

[0113] The second type of evaluation metrics includes throughput-based evaluation metrics and channel quality-based evaluation metrics; the throughput-based evaluation metrics include at least one of the following: system throughput (Throughput), user throughput (userthroughput, UPT), average user throughput, 5% user throughput, spectral efficiency (SE); the channel quality-based evaluation metrics include at least one of the following: bit error rate (BER), block error rate (BLER), hypothetical bit error rate, hypothetical block error rate, signal-to-noise ratio (SNR), signal-to-interference-plus-noise ratio (SINR), modulation and coding scheme (MCS) transmission level;

[0114] The third type of evaluation metrics includes data distribution error-based evaluation metrics, and the data distribution error-based evaluation metrics include at least one of the following: power spectrum, energy spectrum, amplitude spectrum, phase spectrum.

[0115] It should be noted that the performance evaluation metrics may also include, but are not limited to, the test performance after the information processing unit is trained, that is, the information processing unit can be used only after passing the detection of the information processing unit. Specifically, the performance evaluation metrics may be average performance metrics, pass rate, etc.

[0116] Taking the first information processing unit as a model, the identifier of the first information processing unit as the indication information of the model, and the first signaling as the capability report of the first node as an example, as a possible example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple model UE_Model#1 and UE_Model#2. The first node reports the indication information UE_ID#1, UE_ID#2 of these models, as well as the first information and the second information to the second node through the capability report. After receiving the capability report of the first node, the second node selects / confirms multiple second-node-side models NW_Model#1, NW_Model#2 that match (for example: NW_Model#1 is jointly trained with UE_Model#1, and NW_Model#2 is jointly trained with UE_Model#2, so one second-node-side model can match one first-node-side model), and inputs the second information into the corresponding second-node-side model to recover the third information. The second node calculates the performance evaluation index of the third information and the first information fed back by the first node to compare which model's indication information has better model performance indicators, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node instructs the first node to use the model with the best performance indicators to perform subsequent inference calculation work, that is, instructs the first information processing unit corresponding to the optimal performance evaluation index as the target first information processing unit. Or, after receiving the capability report of the first node, the second node selects / confirms one second-node-side model NW_Model#1 that matches (for example: NW_Model#1 is jointly trained with UE_Model#1 and UE_Model#2, so one second-node-side model can match multiple first-node-side models), and inputs the second information into the corresponding second-node-side model to recover the third information. The second node calculates the performance evaluation index of the third information and the first information to compare which model's indication information has better model performance indicators, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node instructs the first node to use the model with the best performance indicators to perform subsequent inference calculation work, that is, sends the second signaling to the first node, and the second signaling includes the identifier of the target first information processing unit.

[0117] As another possible example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple models UE_Model#1 and UE_Model#2, where the model UE_Model#1 is trained based on the dataset Dataset#1, and the model UE_Model#2 is trained based on the dataset Dataset#2. The training process can be as Figure 10 shownFigure 10 The dataset #1 in it can correspond to Dataset #1, and the dataset #2 can correspond to Dataset #2. It should be noted that the dataset can include but is not limited to the dataset sent by the second node to the first node / the dataset indicated by the second node for the first node to collect (for example: a certain spatial area / time period / scenario / configuration / function / feature / feature group / implementation method, etc.). Moreover, different datasets can be completely different / partially overlapping / in a subset relationship. The specific dataset sending method can be that the second node broadcasts to users in the entire cell / a certain specific area or can be transmitted separately to one or more first nodes. Therefore, after the first node reports its own capability report (including Dataset_ID#1, Dataset_ID#2, the first information, and the second information) to the second node, the second node can select / confirm the corresponding second node side models NW_Model#1 and NW_Model#2. Then, the second node can input the second information into NW_Model#1 and NW_Model#2 to recover two pieces of third information. The second node calculates the performance evaluation indicators of the two pieces of third information and the first information to compare which model has better model performance indicators under the indication information, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node instructs the first node to use the model with the best performance indicators for subsequent inference calculation work, that is, sends a second signaling to the first node. The second signaling includes the identifier of the target first information processing unit.

[0118] It should be noted that the second node can perform performance comparison based on the information partially reported by the first node. For example, the second node can judge the difference size and change range between the current channel information and the channel information at the previous moment by the third information recovered by processing the second information according to different second information processing units, so as to indirectly judge the performance difference and lay a foundation for the indication and selection of the first information processing unit. This method is applicable to all embodiments of the present disclosure and will not be elaborated below.

[0119] As another possible example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple models UE_Model#1 and UE_Model#2, where the models UE_Model#1 and UE_Model#2 are both trained based on the same dataset Dataset#1. As an example, when the identifier of the first information processing unit is a single-layer structure, that is, when the identifier / indication information of the model is a single-layer structure, the identifier information of UE_Model#1 can be UE_ID#1, indicating that it is UE_Model#1 trained under Dataset#1. After the first node reports the capabilities of the first node (including UE_ID#1, UE_ID#2, the first information, and the second information) to the second node, the second node determines and selects a matching second node-side model NW_Model#1 (for example: both UE_Model#1 and UE_Model#2 are trained based on the dataset Dataset#1 sent by the second node, where Dataset#1 can be generated after training NW_Model#1, so one second node-side model can match multiple first node-side models), and inputs the second information into the corresponding second node-side model to recover the third information (for example: the third channel state information). The second node calculates the performance evaluation metrics of the third information and the first information to compare which model has better performance metrics under the indication information of the model, that is, the model pairing matching degree is the highest or most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node instructs the first node to use the model with the best performance metrics for subsequent inference calculation work, that is, sends a second signaling to the first node, and the second signaling includes the identifier of the target first information processing unit.

[0120] As another example, when the identifier of the first information processing unit is a multi-layer structure, that is, when the identifier / indication information of the model is a multi-layer structure. For example, taking the identifier of the first information processing unit being determined based on the identifier information of the dataset corresponding to the first information processing unit and the identifier information of the first information processing unit as an example, assume that the models supported by the first node are UE_Model#1 and UE_Model#2. The identifier information of UE_Model#1 can be UE_ID#1-1, indicating UE_Model#1 trained under Dataset#1, and the identifier information of UE_Model#2 can be UE_ID#1-2, indicating UE_Model#2 trained under Dataset#1. After the first node reports the capability report of the first node (including UE_ID#1-1, UE_ID#1-2, the first information, and the second information) to the second node, the second node selects / confirms a matching second node-side model NW_Model#1 (for example: both UE_Model#1 and UE_Model#2 are trained based on the dataset Dataset#1 sent by the base station, where Dataset#1 can be generated after training NW_Model#1, so one second node-side model can match multiple first node-side models), and inputs the two second information corresponding to the two first node-side models into the corresponding second node-side model to recover two third information (for example: the third channel state information). The second node calculates the performance evaluation metrics of the two third information and the first information to compare which model has better model performance metrics under the indication information, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node instructs the first node to use the model with the best performance metrics for subsequent inference calculation work, that is, sends a second signaling to the first node, and the second signaling includes the identifier of the target first information processing unit.

[0121] Taking the first information processing unit as an example, the indication information of the first information processing unit as the model, and the first signaling as the capability report of the first node, as a possible example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple model UEs_Model#1 and UEs_Model#2, where the model UEs_Model#1 is trained based on the training period / stage Session#1, and the model UEs_Model#2 is trained based on the training period / stage Session#2. After the first node reports the capability report of the first node (including UEs_Model#1, UEs_Model#2, the first information, and the second information) to the second node, the second node selects / confirms the corresponding second node-side models NW_Model#1 and NW_Model#2, and then inputs the second information into NW_Model#1 and NW_Model#2 to recover two pieces of third information. Furthermore, the second node calculates the performance evaluation index between the two pieces of third information and the first information to compare which model has better model performance indicators under the indication information of the model, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node instructs the first node to use the model with the best performance indicators for subsequent inference calculation work, that is, sends the second signaling to the first node, and the second signaling includes the identifier of the target first information processing unit.

[0122] As another possible example, for a certain scenario / configuration / function / feature / feature group / implementation, the first node supports multiple models UE_Model#1 and UE_Model#2, where both models UE_Model#1 and UE_Model#2 are trained during the same training period / stage Session#1. As an example, when the identifier of the first information processing unit is a single-layer structure, that is, when the identifier / indicator information of the model is a single-layer structure, the identifier information of UE_Model#1 can be UE_ID#1, indicating UE_Model#1 trained under Session#1. After the first node reports the capabilities of the first node (including UE_ID#1, UE_ID#2, the first information, and the second information) to the second node, the second node selects / confirms the matching second-node-side model NW_Model#1 (for example, UE_Model#1, UE_Model#2, and NW_Model#1 are all trained under Session#1, so one second-node-side model can match multiple first-node-side models). Then, the second node inputs the two second information corresponding to the two first-node-side models into the corresponding second-node-side model to recover two third information (for example: the third channel state information). The second node calculates the performance evaluation metrics of the two third information and the first information to compare which model has better model performance metrics under the indicator information of the model, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation. Furthermore, the second node instructs the first node to use the model with the best performance metrics for subsequent inference calculation work, that is, sends a second signaling to the first node, and the second signaling includes the identifier of the target first information processing unit.

[0123] As another example, when the identifier of the first information processing unit is a multi-layer structure, that is, when the identifier / indication information of the model is a multi-layer structure. For example, taking the identifier of the first information processing unit being determined based on the identifier information of the training phase corresponding to the first information processing unit and the identifier information of the first information processing unit as an example, the identifier information of UE_Model#1 can be UE_ID#1-1, indicating that UE_Model#1 is trained under Session#1. After the first node reports the capability report of the first node (including UE_ID#1-1, UE_ID#1-2, the first information, and the second information) to the second node, the second node selects / confirms the matching second node side model NW_Model#1 (for example: UE_Model#1, UE_Model#2, and NW_Model#1 are all trained based on Session#1, so one second node side model can match multiple first node side models). Then the second node inputs the two second information corresponding to the two first node side models into the second node side model to recover two third information (for example: the third channel state information). The second node calculates the performance evaluation index of the two third information and the first information to compare which model has better model performance indicators under the indication information, that is, the model pairing matching degree is the highest or most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node instructs the first node to use the model with the best performance indicator for subsequent inference calculation work, that is, sends a second signaling to the first node, and the second signaling includes the identifier of the target first information processing unit.

[0124] In some embodiments, after the second node determines the target first information processing unit, the second node can select a target second information processing unit that matches the target first information processing unit for subsequent inference calculation work.

[0125] Example 2: The first signaling further includes the second information, and the second signaling includes the third information.

[0126] Among them, the third information is obtained by the second node processing the second information based on the second information processing unit.

[0127] In the example shown in the above Example 1, the first node reports both the first information and the second information to the second node, and the second node determines the target first information processing unit based on the performance evaluation index between the third information and the first information. In some embodiments, the first node may also only report the identifier of the first information processing unit supported by itself and the second information to the second node. After receiving the first signaling, the second node determines the first information processing unit supported by the first node based on the identifier of the first information processing unit included in the first signaling, and then determines the second information processing unit that matches the first information processing unit. Then, the second information is processed based on the second information processing unit to obtain the third information, and then the second signaling including the third information is sent to the first node, so that the first node can determine the target first information processing unit based on the performance evaluation index between the third information in the second signaling and the first information.

[0128] In some embodiments, the information or signaling involved in the embodiments of the present disclosure, such as the first information, the second information, the first signaling, and the second signaling, may all be information or signaling after being quantized by a quantization method, and the quantization method may be a scalar quantization, a vector quantization, a codebook quantization, or the like.

[0129] Based on this, in the case where the first signaling further includes the second information and the second signaling includes the third information, based on the second signaling, determining the target first information processing unit from the first information processing units may include the following steps:

[0130] A1. Determine the performance evaluation index between the first information and the third information.

[0131] For the relevant description of the performance evaluation index, reference may be made to the corresponding description in the above Example 1, which will not be elaborated here.

[0132] A2. Based on the performance evaluation index between the first information and the third information, determine the target first information processing unit from the first information processing units.

[0133] In some embodiments, the first node may determine the first information processing unit that matches the second information processing unit corresponding to the optimal performance evaluation index as the target first information processing unit.

[0134] In some embodiments, in the case where the first signaling further includes the second information and the second signaling includes the third information, after the first node determines the target first information processing unit from the first information processing units based on the second signaling, the first node may also send the identifier of the target first information processing unit to the second node, so that the second node can use the target second information processing unit that matches the target first information processing unit for subsequent inference calculation work based on the identifier of the target first information processing unit.

[0135] In some embodiments, the transmission of the third information may, but is not limited to, be carried out by means of scalar quantization, vector quantization, or codebook quantization (e.g., Type I codebook, Type II codebook, various enhanced Type I / II codebooks, etc.). The data / information / parameters / models / signaling, etc. in the transmission process involved in the embodiments of the present disclosure are all applicable to the above quantization methods. For example, the transmission of the first signaling and the second signaling are both applicable to the above quantization methods, and details thereof will not be elaborated herein.

[0136] The following uses several specific examples to illustrate the determination of the target first information processing unit from the first information processing unit based on the second signaling in the case where the first signaling further includes the second information and the second signaling includes the third information.

[0137] Taking the first information processing unit as a model, the identifier of the first information processing unit as the indication information of the model, and the first signaling as the capability report of the first node as an example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple models UE_Model#1 and UE_Model#2, and reports the indication information UE_ID#1, UE_ID#2 of these models and the second information to the second node through the capability report / dynamic report (e.g., CSI report). After receiving the capability report of the first node, the second node selects one or more second node-side models (i.e., second information processing units) that are adapted based on UE_ID#1 and UE_ID#2, and then inputs the second information into one or more second node-side models for processing to obtain one or more third information, and then sends the second signaling to the first node. The second signaling includes one or more third information. After receiving the second signaling, the first node calculates the performance evaluation index of each third information and the first information to compare which model has better model performance indicators under the indication information of the model, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Then, the first node uses the model UE_Model#1 with the best model performance indicator (i.e., the optimal performance evaluation index) to perform subsequent inference calculation work. At the same time, the first node needs to report UE_ID#1 to the second node, and the second node selects NW_Model#1 corresponding to UE_ID#1 to perform subsequent inference calculation work.

[0138] It should be noted that when the second node transmits multiple adapted third information, it may include but is not limited to binding the relevant information indicated by the identifier of the model to the third information and indicating it to the first node, in order to help the first node understand the adapted second information for performance evaluation (for example: the third information output by NW_Model#1 is the third information output by the second node side model that matches UE_ID#1). Moreover, the number of the third information is the same as the number of models supported by the first node. When the second node only transmits one adapted third information, it may not indicate the relevant information indicated by the identifier of the model to the first node. For example, the first node only reports the identifier / indication information of one model.

[0139] As a possible example, after the second node receives the capability report sent by the first node, the second node determines the corresponding second node side models NW_Model#1 and NW_Model#2 based on UE_ID#1 and UE_ID#2 included in the capability report (for example: NW_Model#1 is jointly trained with UE_Model#1, and NW_Model#2 is jointly trained with UE_Model#2, so one second node side model can match one first node side model). The second node sends the two second information included in the capability report into the corresponding NW_Model#1 and NW_Model#2 to obtain two corresponding third information, and transmits the two third information to the first node, that is, sends a second signaling to the first node, and the second signaling includes two third information. After the first node receives the second signaling, the first node calculates the performance evaluation index of each third information and the first information to compare which model's indication information has better model performance index, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Then, the first node uses the model UE_Model#1 with the best performance index to perform subsequent inference calculation work. At the same time, the first node needs to report UE_ID#1 to the second node, and the second node selects NW_Model#1 corresponding to UE_ID#1 to perform subsequent inference calculation work.

[0140] As another possible example, after the second node receives the capability report sent by the first node, the second node determines the matching second-node-side model NW_Model#1 based on UE_ID#1 and UE_ID#2 included in the capability report (for example: NW_Model#1 is jointly trained with UE_Model#1 and UE_Model#2, so one second-node-side model can match multiple first-node-side models). After that, the second node inputs the second information corresponding to UE_ID#1 and the second information corresponding to UE_ID#2 included in the capability report into NW_Model#1 respectively, obtains two pieces of third information, and transmits the two pieces of third information to the first node, that is, sends a second signaling to the first node, and the second signaling includes the two pieces of third information. After the first node receives the second signaling, the first node calculates the performance evaluation metrics of each third information and the first information to compare which model has better model performance metrics under the indication information of which model, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the first node uses the model UE_Model#1 with the best performance metrics to perform subsequent inference calculation work. At the same time, the first node needs to report UE_ID#1 to the second node, and the second node selects NW_Model#1 corresponding to UE_ID#1 to perform subsequent inference calculation work.

[0141] As another possible example, the multiple models supported by the first node, UE_Model#1 and UE_Model#2, can but are not limited to being trained based on the same dataset Dataset#1. After the second node receives the capability report sent by the first node, the second node determines the corresponding second-node side model NW_Model#1 based on the UE_ID#1 and UE_ID#2 included in the capability report (for example: NW_Model#1 is trained based on the same dataset Dataset#1 as UE_Model#1 and UE_Model#2, so one second-node side model can match multiple first-node side models). After that, the second node inputs the second information corresponding to UE_ID#1 and the second information corresponding to UE_ID#2 included in the capability report into NW_Model#1 respectively, obtains two pieces of third information, and transmits the two pieces of third information to the first node, that is, sends a second signaling to the first node, and the second signaling includes the two pieces of third information. After the first node receives the second signaling, the first node calculates the performance evaluation index of each third information and the first information to compare which model has better model performance indicators under the indication information of the model, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the first node uses the model UE_Model#1 with the best performance indicators to perform subsequent inference calculation work. At the same time, the first node needs to report UE_ID#1 to the second node, and the second node selects NW_Model#1 corresponding to UE_ID#1 to perform subsequent inference calculation work.

[0142] As another possible example, the multiple models supported by the first node, UE_Model#1 and UE_Model#2, can be but are not limited to being trained based on the same time period / stage Session#1. After the second node receives the capability report sent by the first node, the second node determines the corresponding second-node side model NW_Model#1 based on the UE_ID#1 and UE_ID#2 included in the capability report (for example: NW_Model#1 is trained based on the same time period / stage Session#1 as UE_Model#1 and UE_Model#2, so one second-node side model can match multiple first-node side models). After that, the second node inputs the second information corresponding to UE_ID#1 and the second information corresponding to UE_ID#2 included in the capability report into NW_Model#1 respectively, obtains two pieces of third information, and transmits the two pieces of third information to the first node, that is, sends a second signaling to the first node, and the second signaling includes the two pieces of third information. After the first node receives the second signaling, the first node calculates the performance evaluation index of each third information and the first information to compare which model has better model performance under the indication information of the model, that is, the model pairing matching degree is the highest or most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the first node uses the model UE_Model#1 with the best performance index to perform subsequent inference calculation work. At the same time, the first node needs to report UE_ID#1 to the second node, and the second node selects NW_Model#1 corresponding to UE_ID#1 to perform subsequent inference calculation work.

[0143] Example 3: The second signaling includes a third information processing unit.

[0144] Among them, the third information processing unit is an information processing unit determined by the second node based on the identifier of the first information processing unit and matching the first information processing unit.

[0145] In the above Example 2, the second signaling sent by the second node includes the third information, so that the first node can determine the target first information processing unit based on the performance evaluation index between the third information and the first information. In some embodiments, the first signaling may also only include the identifier of the first information processing unit. After receiving the first signaling, the second node may determine the third information processing unit that matches the first information processing unit based on the identifier of the first information processing unit, and then send the second signaling to the first node. The second signaling includes the third information processing unit, so that the first node can process the second information based on the third information processing unit to obtain the fourth information, and then determine the target first information processing unit based on the performance evaluation index between the first information and the fourth information. Among them, the third information processing unit may be the same as the above-mentioned second information processing unit or different from the above-mentioned second information processing unit. The embodiments of the present disclosure do not limit this. In the case where the third information processing unit is the same as the above-mentioned second information processing unit, the fourth information may be the same as the above-mentioned third information. It should be understood that the computing power resources of the first node are limited. In the case where the computing power resources required by the second information processing unit are relatively high, the third information processing unit included in the second signaling may be a simplified second information processing unit, and in the case where the computing power resources required by the second information processing unit are relatively low, the third information processing unit included in the second signaling may be the second information processing unit.

[0146] Based on this, in the case where the first signaling further includes the second information and the second signaling includes the third information, determining the target first information processing unit from the first information processing units based on the second signaling may include the following steps:

[0147] B1. Process the second information based on the third information processing unit to obtain the fourth information.

[0148] In some embodiments, in the case where the second signaling includes the third information processing unit, after receiving the second signaling, the first node may input the second information corresponding to each first information processing unit among the multiple first information processing units into the third information processing unit to obtain multiple fourth information.

[0149] B2. Determine the performance evaluation index between the first information and the fourth information.

[0150] B3. Based on the performance evaluation index between the first information and the fourth information, determine the target first information processing unit from the first information processing units.

[0151] For the description of B2 - B3, reference may be made to the above description of A1 - A2, which will not be elaborated here.

[0152] In some embodiments, when the second signaling includes a third information processing unit, after the first node determines the target first information processing unit from the first information processing units based on the second signaling, the first node may further send the identifier of the target first information processing unit to the second node, so that the second node can use the target second information processing unit that matches the target first information processing unit to perform subsequent inference calculation work based on the identifier of the target first information processing unit.

[0153] As a possible example, after the first node determines the performance evaluation index between the first information and the fourth information, the first node may also send the performance evaluation index between the first information and the fourth information to the second node. The second node determines the target first information processing unit from the first information processing units based on the performance evaluation index between the first information and the fourth information, and then sends the identifier of the target first information processing unit to the first node. The first node determines the target first information processing unit from the first information processing units based on the identifier of the target first information processing unit. That is to say, when the second signaling includes a third information processing unit, after step S102, the first node may send the third signaling to the second node. The third signaling includes the performance evaluation index between the first information and the fourth information. After that, the first node receives the fourth signaling sent by the second node. The fourth signaling includes the identifier of the target first information processing unit. Then, the first node determines the target first information processing unit from the first information processing units based on the identifier of the target first information processing unit.

[0154] As another possible example, after the first node determines the performance evaluation index between the first information and the fourth information, the first node may also determine the candidate information processing units from the first information processing units based on the performance evaluation index between the first information and the fourth information, and then send the information processing unit recommendation information to the second node. The information processing unit recommendation information includes the identifier of the candidate first information processing unit. The second node determines the target first information processing unit from the candidate first information processing units based on the identifier of the candidate first information processing unit included in the information processing unit recommendation information, and then the second node sends the identifier of the target first information processing unit to the first node. The first node determines the target first information processing unit from the first information processing units based on the identifier of the target first information processing unit. That is to say, when the second signaling includes a third information processing unit, after step S102, the first node may send the fifth signaling to the second node. The fifth signaling includes the information processing unit recommendation information. After that, the first node receives the sixth signaling sent by the second node. The sixth signaling includes the identifier of the target first information processing unit. Then, the first node determines the target first information processing unit from the first information processing units based on the identifier of the target first information processing unit.

[0155] It should be noted that the example in which the first node determines candidate first information processing units based on the performance evaluation metrics and then recommends information to the second node information processing unit is also applicable to other embodiments in the present disclosure where the first node determines the performance evaluation metrics, and will not be elaborated below.

[0156] The following uses several specific examples to illustrate the determination of the target first information processing unit from the first information processing units based on the second signaling in the case where the second signaling includes the third information processing unit.

[0157] Taking the first information processing unit as a model, the identifier of the first information processing unit as the indication information of the model, and the first signaling as the capability report of the first node as an example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple models UE_Model#1 and UE_Model#2, and reports the indication information UE_ID#1, UE_ID#2 of these models to the second node through the capability report / dynamic report (for example, CSI report). After receiving the capability report of the first node, the second node selects one or more second node-side models (i.e., the third information processing unit) that are adapted based on UE_ID#1 UE_ID#2, and then sends the second signaling to the first node. The second signaling includes one or more second node-side models that are adapted.

[0158] As a possible example, for a certain scenario / configuration / function / feature / feature group / implementation, the first node supports multiple models UE_Model#1 and UE_Model#2. The models UE_Model#1 and UE_Model#2 may include, but are not limited to, those trained based on the same dataset Dataset#1. The first node reports UE_ID#1 and UE_ID#2 to the second node through a capability report. After receiving the capability report sent by the first node, the second node determines the corresponding models NW_Model#1 and NW_Model#2 on the second node side based on UE_ID#1 and UE_ID#2 included in the capability report. Then, the second node sends a second signaling to the first node, and the second signaling includes NW_Model#1 and NW_Model#2. After receiving the second signaling, the first node inputs multiple second information into NW_Model#1 and NW_Model#2 respectively, obtains the fourth information corresponding to NW_Model#1 and the fourth information corresponding to NW_Model#2, and further compares which model has better model performance indicators under the indication information of which model based on the performance evaluation indicators between the two fourth information and the first information, that is, the model pairing has the highest matching degree or is most suitable for the current scenario / configuration / function / feature / feature group / implementation. Furthermore, the first node uses the model UE_Model#1 with the best performance indicator to perform subsequent inference calculation work. At the same time, the first node needs to report UE_ID#1 to the second node, and the second node selects NW_Model#1 corresponding to UE_ID#1 to perform subsequent inference calculation work.

[0159] As another possible example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple models UE_Model#1 and UE_Model#2. The models UE_Model#1 and UE_Model#2 can include, but are not limited to, those trained based on the same training period / stage Session#1. The first node reports UE_ID#1 and UE_ID#2 to the second node through a capability report. After receiving the capability report sent by the first node, the second node determines the corresponding models NW_Model#1 and NW_Model#2 on the second node side based on UE_ID#1 and UE_ID#2 included in the capability report. Then, the second node sends a second signaling to the first node, and the second signaling includes NW_Model#1 and NW_Model#2. After receiving the second signaling, the first node inputs multiple second information into NW_Model#1 and NW_Model#2 respectively, obtains the fourth information corresponding to NW_Model#1 and the fourth information corresponding to NW_Model#2, and then compares which model has better model performance indicators under the indication information of which model based on the performance evaluation index between the two fourth information and the first information, that is, the model pairing matching degree is the highest or most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the first node uses the model UE_Model#1 with the best performance indicator to perform subsequent inference calculation work. At the same time, the first node needs to report UE_ID#1 to the second node, and the second node selects NW_Model#1 corresponding to UE_ID#1 to perform subsequent inference calculation work.

[0160] Example 4: The first signaling further includes performance index information, and the second signaling includes the identifier of the target first information processing unit.

[0161] Among them, the performance index information includes the performance evaluation index between the first information and the fifth information. The fifth information is obtained by the first node based on the fourth information processing unit processing the second information. The identifier of the target first information processing unit is determined by the second node from the first information processing units based on the performance index information.

[0162] In the above Example 3, the second node sends the third information processing unit that matches the first information processing unit to the first node. The first node processes the second information based on the third information processing unit to obtain the fourth information, and then determines the target first information processing unit from the first information processing units based on the performance evaluation index between the fourth information and the first information. In some embodiments, the first node is pre-configured or pre-stores the fourth information processing unit that matches the first information processing unit. After the first node processes the first information based on multiple first information processing units to obtain multiple second information, it can process the multiple second information based on multiple fourth information processing units to obtain multiple fifth information, and then determine the performance evaluation index between the multiple fifth information and the first information, and then send a first signaling to the second node. The first signaling includes the identifier of the first information processing unit and the performance index information, so that the second node can determine the target first information processing unit from the first information processing units based on the performance index information. After receiving the first signaling, the second node determines the target first information processing unit from the first information processing units based on the performance index information, and then sends a second signaling to the first node. The second signaling includes the identifier of the target first information processing unit.

[0163] It should be noted that in the case where the second node has the same understanding as the first node regarding the order of the first information processing units of the first node, the first node only needs to feedback the performance evaluation index according to the order of the first information processing units, without having to feedback the identifier of the first information processing unit corresponding to each performance evaluation index. In this way, the feedback overhead of the first node can be reduced.

[0164] Based on this, in the case where the first signaling further includes the performance index information and the second signaling includes the identifier of the target first information processing unit, determining the target first information processing unit from the first information processing units based on the second signaling may be that the first node determines the target first information processing unit from the first information processing units based on the identifier of the target first information processing unit included in the second signaling.

[0165] In some embodiments, in the case where the first signaling further includes the performance index information and the second signaling includes the identifier of the target first information processing unit, after the first node determines the target first information processing unit from the first information processing units based on the second signaling, the first node may further send the identifier of the target first information processing unit to the second node, so that the second node can use the target second information processing unit that matches the target first information processing unit to perform subsequent inference calculation work based on the identifier of the target first information processing unit.

[0166] The following uses several specific examples to illustrate the case where the first signaling further includes performance metric information and the second signaling includes the identifier of the target first information processing unit. Based on the second signaling, the target first information processing unit is determined from the first information processing units.

[0167] Taking the first information processing unit as a model, the indication information of the identifier of the first information processing unit as a model, and the first signaling as the capability report of the first node as an example, for a certain scenario / configuration / function / feature / feature group / implementation manner, the first node supports multiple model UE_Model#1 and UE_Model#2 (where UE_Model#1 and UE_Model#2 may include but are not limited to being trained based on the same data set, in the same training phase, or based on the same second node-side model). The first node calculates the corresponding performance information #1 and performance information #2 through the fourth information processing unit (performance information #1 is better than performance information #2), and reports the performance metric information (including performance information #1 and performance information #2) and / or the first information and the indication information of the model UE_ID#1, UE_ID#2 to the second node through the capability report / dynamic report (such as: CSI report, etc.). After receiving the capability report / dynamic report, the second node determines the better-performing UE_Model#1 based on the performance metric information, and instructs the better-performing UE_Model#1 to the first node for its subsequent inference calculation, that is, sends the second signaling to the first node. The second signaling includes the identifier of the target first information processing unit.

[0168] Example 5: The first signaling further includes performance metric information, and the second signaling includes the identifier of the candidate first information processing unit.

[0169] Among them, the candidate first information processing unit is determined by the second node from the first information processing units based on the performance metric information.

[0170] In the above Example 4, the second node directly determines the target first information processing unit from the first information processing units based on the performance metric information. In some embodiments, the second node may also determine the candidate first information processing unit from the first information processing units based on the performance metric information, and then send the second signaling to the first node. The second signaling includes the identifier of the candidate first information processing unit, and the first node determines the target first information processing unit from the candidate first information processing units.

[0171] Exemplarily, assume that the number of first information processing units supported by the first node is M. The second node can determine N candidate first information processing units from the M first information processing units based on the performance metric information, and then send a second signaling to the first node. The second signaling includes the identifiers of the N candidate first information processing units, and the first node determines the target first information processing unit from the N candidate first information processing units. Here, 1 < N < M.

[0172] Based on this, in the case where the first signaling further includes performance metric information and the second signaling includes the identifiers of the candidate first information processing units, based on the second signaling, determining the target first information processing unit from the first information processing units may be that the first node determines the target first information processing unit from the first information processing units based on the attribute information of the first node and the identifiers of the candidate first information processing units.

[0173] Among them, the attribute information of the first node includes at least one of the following: the capability information of the first node, the resource occupancy of the first node. The capability information of the first node includes at least one of the following: software capability, hardware capability, and the resource occupancy of the first node includes at least one of the following: CSI processing unit occupancy (which can be understood as the occupancy of the central processing unit (CPU)), power consumption.

[0174] In some embodiments, in the case where the first signaling further includes performance metric information and the second signaling includes the identifiers of the candidate first information processing units, after the first node determines the target first information processing unit from the first information processing units based on the second signaling, the first node may further send the identifier of the target first information processing unit to the second node, so that the second node can use the target second information processing unit that matches the target first information processing unit to perform subsequent inference calculation work based on the identifier of the target first information processing unit.

[0175] The following uses a specific example to illustrate the determination of the target first information processing unit from the first information processing units based on the second signaling in the case where the first signaling further includes performance metric information and the second signaling includes the identifiers of the candidate first information processing units.

[0176] Taking the first information processing unit as a model, the identifier of the first information processing unit as the indication information of the model, and the first signaling as the capability report of the first node as an example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple models UE_Model#1, UE_Model#2, and UE_Model#3. Then, the first node reports the indication information UE_ID#1, UE_ID#2, UE_ID#3 of the multiple models, as well as the performance metric information to the second node through the capability report. The second node determines which model has better model performance metrics based on the relevant performance metrics in the performance metric information, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node determines that the performances of UE_ID#2 and UE_ID#3 can both meet certain performance evaluation metrics, that is, the models corresponding to UE_ID#2 and UE_ID#3 are candidate models. Then, the second node indicates to the first node that the performances of UE_ID#2 and UE_ID#3 can both meet certain evaluation metrics. The first node further selects the model UE_Model#2 corresponding to UE_ID#2 as the target model according to its own attribute information, and reports the indication information UE_ID#2 of the model corresponding to UE_Model#2 to the second node. The second node further selects NW_Model#2 that matches UE_ID#2 for subsequent inference calculation work.

[0177] Example 6: The first signaling further includes first information and second information, and the second signaling includes the identifier of the candidate first information processing unit.

[0178] Among them, the candidate first information processing unit is determined by the second node from the first information processing units based on the performance evaluation metrics between the third information and the first information, and the third information is obtained by the second node after processing the second information based on the second information processing unit.

[0179] In the above Example 1, the second node directly determines the target first information processing unit from the first information processing units based on the performance evaluation metrics between the third information and the first information, and then indicates the target first information processing unit to the first node. In some embodiments, the second node may also determine the candidate first information processing unit from the first information processing units based on the performance evaluation metrics between the third information and the first information, and then indicate the candidate first information processing unit to the first node. The first node determines the target first information processing unit from the first information processing units based on the candidate first information processing unit.

[0180] Based on this, when the first signaling further includes first information and second information, and the second signaling includes the identifier of the candidate first information processing unit, the target first information processing unit may be determined from the first information processing units based on the second signaling. It may be that the first node determines the target first information processing unit from the first information processing units based on the attribute information of the first node and the identifier of the candidate first information processing unit.

[0181] For the description of the attribute information of the first node, reference may be made to the relevant description in Example 5 above. For the description of how the second node determines the candidate first information processing unit from the first information processing units based on the performance evaluation index between the third information and the first information, reference may be made to the description in Example 1 above on how the second node determines the target first information processing unit from the first information processing units based on the performance evaluation index between the third information and the first information, which will not be elaborated here.

[0182] The following uses a specific example to illustrate the determination of the target first information processing unit from the first information processing units based on the second signaling when the first signaling further includes performance index information and the second signaling includes the identifier of the candidate first information processing unit.

[0183] Taking the first information processing unit as a model, the identifier of the first information processing unit as the indication information of the model, and the first signaling as the capability report of the first node as an example, as a possible example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple model UEs, namely UE_Model#1, UE_Model#2, and UE_Model#3. Then, the first node reports the indication information of the multiple models, UE_ID#1, UE_ID#2, UE_ID#3, the first information, and the second information to the second node through the capability report. After receiving the capability report, the second node selects / confirms the corresponding multiple second node side models, NW_Model#1, NW_Model#2, NW_Model#3 (for example: NW_Model#1 is jointly trained with UE_Model#1, NW_Model#2 is jointly trained with UE_Model#2, and NW_Model#3 is jointly trained with UE_Model#3. Therefore, one second node side model can match one first node side model), and inputs each second information into the corresponding second node side model to recover multiple third information. The base station calculates the performance evaluation index between each third information and the first information to compare which model has better model performance index under the indication information of the model, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node determines that the performances of UE_ID#2 and UE_ID#3 can both meet certain performance evaluation indexes, that is, the models corresponding to UE_ID#2 and UE_ID#3 are candidate models. Then, the second node indicates to the first node that the performances of UE_ID#2 and UE_ID#3 can both meet certain evaluation indexes. The first node further selects the model UE_Model#2 corresponding to UE_ID#2 as the target model according to its own attribute information, and reports the indication information UE_ID#2 of the model corresponding to UE_Model#2 to the second node. The second node further selects NW_Model#2 that matches UE_ID#2 for subsequent inference calculation work.

[0184] As another possible example, for a certain scenario / configuration / function / feature / feature group / implementation method, the first node supports multiple models UE_Model#1, UE_Model#2, and UE_Model#3. Then, the first node reports the indication information UE_ID#1, UE_ID#2, UE_ID#3, the first information, and the second information of the multiple models to the second node through a capability report. After receiving the capability report, the second node selects / confirms the corresponding model NW_Model#1 on the second node side (for example: NW_Model#1 is jointly trained with UE_Model#1, UE_Model#2, and UE_Model#3, so one model on the second node side can match multiple models on the first node side), and inputs each second information into the model on the second node side to recover multiple third information. The base station calculates the performance evaluation index between each third information and the first information to compare which model has better model performance under the indication information of which model, that is, the highest model pairing matching degree or the most suitable for the current scenario / configuration / function / feature / feature group / implementation method. Furthermore, the second node determines that the performances of UE_ID#2 and UE_ID#3 can both meet certain performance evaluation indexes, that is, the models corresponding to UE_ID#2 and UE_ID#3 are candidate models. Then, the second node indicates to the first node that the performances of UE_ID#2 and UE_ID#3 can both meet certain evaluation indexes. The first node further selects the model UE_Model#2 corresponding to UE_ID#2 as the target model according to its own attribute information, and reports the indication information UE_ID#2 of the model corresponding to UE_Model#2 to the second node. The second node further selects NW_Model#2 that matches UE_ID#2 for subsequent inference calculation work.

[0185] Based on Figure 3 In the embodiment shown, the target first information processing unit of the first node is determined based on the second signaling sent by the second node. In this way, the matching accuracy between the information processing unit of the first node and the information processing unit of the second node is improved, and the performance degradation of the information processing unit can be avoided.

[0186] In some embodiments, the method may further include the following steps:

[0187] C1. Receive the first indication information sent by the second node.

[0188] In some embodiments, the second node sends first indication information to the first node periodically / aperiodically / semi-persistently or based on an event trigger. Accordingly, the first node receives the first indication information sent by the second node. Among them, the event trigger includes that the second node performs an information processing unit switch, and the first indication information is used to indicate a fifth information processing unit. The fifth information processing unit can be understood as the information processing unit that the second node is about to switch to or is currently using. The fifth information processing unit may be the same as the above-mentioned second information processing unit or different from the above-mentioned second information processing unit. The embodiments of the present disclosure do not limit this.

[0189] In some embodiments, the first indication information includes at least one of the following: identification information of the fifth information processing unit, identification information of the data set corresponding to the fifth information processing unit, identification information of the training phase corresponding to the fifth information processing unit, description information of the fifth information processing unit, etc.

[0190] C2. Based on the first indication information, determine a sixth information processing unit that matches the fifth information processing unit.

[0191] After receiving the first indication information, the first node can, based on the first indication information, determine a sixth information processing unit that matches the fifth information processing unit, and then use the sixth information processing unit to perform subsequent inference calculation work. It should be understood that the sixth information processing unit is determined by the first indication information sent by the second node. Therefore, the first node uses the sixth information processing unit to perform subsequent inference calculation work, which improves the matching accuracy between the information processing unit of the first node and the information processing unit of the second node and can avoid the performance degradation of the information processing unit.

[0192] In some embodiments, after determining the sixth information processing unit, the first node can send the identification of the sixth information processing unit to the second node, so that the second node can know that the first node will use the sixth information processing unit to perform subsequent inference calculation work. In this way, a consistent understanding of the sixth information processing unit can be achieved between the first node and the second node.

[0193] In some embodiments, when the first node sends the identification of the sixth information processing unit or the identification of the target first information processing unit to the second node, it may be that the first node sends information processing unit identification information to the second node, and the information processing unit identification information includes the identification of the sixth information processing unit or the identification of the target first information processing unit. The information processing unit identification information may include, but is not limited to, determining the information processing unit identification information through the information processing unit identification / authentication process, and the information processing unit identification information may be a global / general identification information or a local / small-range managed identification information.

[0194] It should be noted that when the number of the sixth information processing units determined by the first node is 1, the first node may report the identifier of the sixth information processing unit to the second node, may also not report the identifier of the sixth information processing unit to the second node, or may report the indication information for indicating whether pairing is supported to the second node. For example, send 1-bit indication information to the second node. When the value of the indication information is the first value, it indicates that the first node supports pairing with the fifth information processing unit, that is, it can determine the sixth information processing unit that matches the fifth information processing unit. When the value of the indication information is the second value, it indicates that the first node does not support pairing with the fifth information processing unit. Herein, the first value is different from the second value. For example, the first value is a non-zero integer (such as 1), and the second value is 0, or the first value is 0, and the second value is a non-zero integer; or, the first value is true (TRUE), the second value is false (FALSE), or the first value is false (FALSE), the second value is true (TRUE). The embodiments of the present disclosure do not limit this.

[0195] When the first node determines that there is no sixth information processing unit paired with the fifth information processing unit, the first node may send the indication information for indicating non-support for pairing to the second node. For example, send 1-bit indication information to the second node. When the value of the indication information is the second value, it indicates that the first node does not support pairing with the fifth information processing unit. After receiving the indication information, the second node will not activate the fifth information processing unit or will not respond.

[0196] In some embodiments, the above indication information for indicating whether pairing is supported and the above indication information for indicating non-support for pairing may both be reported to the second node in the first signaling, that is, the first signaling may include the above indication information for indicating whether pairing is supported and the above indication information for indicating non-support for pairing.

[0197] As a possible example, the first indication information may be the identifier of the fifth information processing unit. When the second node sends the first indication information, it may be that the second node broadcasts / indicates the identifier of the fifth information processing unit to all / some of the first nodes within the cell / within a certain area. After receiving the identifier of the fifth information processing unit, the first node selects / confirms the matching sixth information processing unit, and further reports the identifier of the sixth information processing unit to the second node to inform the second node that the first node will use the sixth information processing unit for subsequent inference calculation work.

[0198] As another possible example, the first indication information is the identification information of the data set corresponding to the fifth information processing unit. When the second node sends the first indication information, it may broadcast / indicate the identification information of the data set corresponding to the fifth information processing unit to all / part of the first nodes within the cell / within a certain area. After receiving the identification information of the data set corresponding to the fifth information processing unit, the first node selects / confirms the sixth information processing unit that matches the fifth information processing unit, and further reports the identification information of the sixth information processing unit to the second node to inform the second node that the sixth information processing unit will be used for subsequent inference calculation work.

[0199] In some embodiments, when the first information processing unit of the first node is trained by the second node and needs to be transmitted to the first node for use, the second node transmits and configures the first information processing unit to the first node and assigns the identification information of the corresponding information processing unit, such as the ID information of the information processing unit, etc.; after receiving the identification information, the first node on the first node side selects / confirms the matching first information processing unit, which can naturally match the second information processing unit on the second node side. Then, the first node reports the identification information of the adapted first information processing unit to the second node to inform the second node that subsequent inference calculation work will be based on the first information processing unit.

[0200] In some embodiments, when the second node only transmits one first information processing unit to the first node for the first node to use, the first node may not need to report the identification information of the corresponding information processing unit.

[0201] In some embodiments, when the second node trains multiple first information processing units and transmits multiple first information processing units to the first node for the first node to use, the first node needs to report the identification information of one or more information processing units that meet the capabilities / conditions of the first node. For the subsequent specific model pairing and indication process, refer to the above Figure 3The illustrated embodiments. For example, the second node transmits two models, UE_Model#1 and UE_Model#2, to the first node. The corresponding identification information of these two models is UE_ID#1 and UE_ID#2 respectively. The first node selects / confirms UE_Model#1 that meets its current capabilities / conditions, and the first node reports the identification information UE_ID#1 corresponding to UE_Model#1 to the second node to complete the pairing and identification process between the two parties. Another example is that the second node transmits three models, UE_Model#1, UE_Model#2, and UE_Model#3, to the first node. The corresponding identification information of these three models is UE_ID#1, UE_ID#2, and UE_ID#3 respectively. The first node selects / confirms that the models that meet its current capabilities / conditions are UE_Model#1 and UE_Model#2. The first node reports the identification information UE_ID#1 and UE_ID#2 corresponding to UE_Model#1 and UE_Model#2 to the second node, and the second node decides the model that the first node should use.

[0202] In some embodiments, in order for the information processing unit of the first node to better match the information processing unit of the second node, the second node may send auxiliary information to the first node to help the first node select a more suitable information processing unit. Based on this, the method may further include the following steps:

[0203] D1. Receive the second indication information sent by the second node.

[0204] Wherein, the second indication information is used for the first node to select the information processing unit.

[0205] In some embodiments, the second indication information includes at least one of the following: the structure information of the information processing unit that the first node should use; the structure information of the information processing unit used by the second node.

[0206] In this way, the first can determine an information processing unit that can better match the second information processing unit of the second node based on the second indication information.

[0207] For example, if the information processing unit used by the second node is a Transformer network structure, then the second node instructing the first node to use an information processing unit of the Transformer structure can enable the information processing unit used by the first node to better adapt to the information processing unit used by the second node. In some embodiments, when the second indication information includes the structure information of the information processing unit that the first node should use and / or the structure information of the information processing unit used by the second node, the structure information can be indicated by a bitmap of K bits (for example: there are four information processing unit structures, where 0001 represents the Transformer structure, 0010 represents the CNN structure, 0100 represents the RNN structure, and 1000 represents the DNN structure); further, in order to reduce the bit overhead of the second indication information, M bits can also be used to indicate (for example: there are four information processing unit structures, where 00 represents the Transformer structure, 01 represents the CNN structure, 10 represents the RNN structure, and 11 represents the DNN structure).

[0208] In some embodiments, the second indication information further includes at least one of the following: the information processing unit complexity of the information processing unit; the computing complexity of the information processing unit; the inference time range of the information processing unit; the storage overhead range of the information processing unit.

[0209] For example, the second node instructs the first node to perform inference using an information processing method with less than 20M floating-point operations (FLOPs); or, for another example, the second node instructs the first node to use an information processing unit with an inference latency less than T time slots / subframes / symbols for inference.

[0210] As an example, the second indication information sent by the second node to the first node only assists the first node in selecting the information processing unit, and does not force the first node to select the information processing unit based on the second indication information. As another example, the second indication information sent by the second node to the first node forcibly instructs the first node to select the information processing unit based on the second indication information. For example, the second node forcibly instructs the first node to use an information processing unit of the Transformer structure.

[0211] It should be understood that since the information processing unit does not have good generalization and cannot be applied to all scenarios / configurations, this will lead to a decrease in performance. Based on this, the method may further include the following steps:

[0212] E1. Receive the third indication information sent by the second node.

[0213] In some embodiments, when the second node detects a decline in the performance of the information processing unit over a period of time, or detects that the performance of the information processing unit is less than or equal to a preset performance threshold over a period of time, the second node sends third indication information to the first node. Accordingly, the first node receives the third indication information sent by the second node, and the third indication information is used for at least one of the following: activation of the information processing unit group, deactivation of the information processing unit group, switching of the information processing unit group, update of the information processing unit group, activation of the information processing unit, deactivation of the information processing unit, switching of the information processing unit, update of the information processing unit, fallback to the default information processing mode.

[0214] Taking the information processing unit as an example of the model, as a possible example, when the second node detects a decline in the model performance over a period of time, the second node performs at least one of the following operations on the entire model group / function group / configuration group of the first node (where there can be one or more models in the model group / function group / configuration group, and the number in different model groups / function groups / configuration groups can be the same or different, and one model can belong to one or more different model groups / function groups / configuration groups): group activation / group deactivation / group switching, etc. (that is, the models in the entire group are all operated. For example: group deactivation means that each model in the group is deactivated). For example, the first node has two supported related function groups, UE_Functionality#1 and UE_Functionality#2, and the model currently used by the first node belongs to UE_Functionality#1. When it is detected that the performance of the current model has deteriorated, the second node can indicate to the first node to deactivate UE_Functionality#1 / switch to UE_Functionality#2 through high-layer or physical-layer signaling (for example: RRC / MAC CE / DCI) (that is, the third indication information), and select a model under UE_Functionality#2 for subsequent inference. It should be noted that the management decision of the second node is at the granularity of the model group / function group / configuration group, and which specific model the first node uses is transparent to the second node.

[0215] As another possible example, when the second node detects a decline in model performance over a period of time, the second node performs at least one of the following operations on the models within the first node's model group / function group / configuration group (where there can be one or more models in the model group / function group / configuration group, and the number of models in different model groups / function groups / configuration groups can be the same or different, and a model can belong to one or more different model groups / function groups / configuration groups): model activation / model deactivation / model switching, etc. For example, the first node has 1 supported related function group UE_Functionality#1, and there are two models UE_Model#1 and UE_Model#2 within the function group. The first node is currently using UE_Model#1. When the second node detects that the current model performance has declined to a poor level, the second node instructs the first node to deactivate UE_Model#1 / switch to UE_Model#2 for subsequent inference through high-layer or physical-layer signaling (e.g., RRC / MAC CE / DCI). It should be noted that the management decision of the second node is at the model granularity, and the specific model used by the first node is not transparent to the second node (i.e., the second node can make management decisions on the model).

[0216] As another possible example, when the second node detects a decline in model performance over a period of time, the second node performs at least one of the following operations on the models among the first node's model group / function group / configuration group (where there can be one or more models in the model group / function group / configuration group, and the number of models in different model groups / function groups / configuration groups can be the same or different, and a model can belong to one or more different model groups / function groups / configuration groups): model activation / model deactivation / model switching, etc. For example, the first node has 2 supported related function groups UE_Functionality#1 and UE_Functionality#2. There are two models UE_Model#1 and UE_Model#2 within the function group UE_Functionality#1, and there are two models UE_Model#3 and UE_Model#4 within the function group UE_Functionality#2. The first node is currently using UE_Model#1 within UE_Functionality#1. When the second node detects that the current model performance has declined to a poor level, the second node instructs the first node to deactivate UE_Model#1 / switch to UE_Model#3 within UE_Functionality#2 for subsequent inference through high-layer or physical-layer signaling (e.g., RRC / MAC CE / DCI). It should be noted that the management decision of the second node is at the model granularity, and the specific model used by the first node is not transparent to the second node (i.e., the second node can make management decisions on the model).

[0217] As another possible example, when a model has good generalization ability and can support multiple scenarios / configurations / functions, the model may belong to multiple different model groups / function groups / configuration groups. If the performance of the model is not good under the current performance function / configuration, when the second node performs operations such as activation / switching on the model group / function group / configuration group of the first node, the first node may not make a model switching selection. For example, the first node has two supported related function groups, UE_Functionality#1 and UE_Functionality#2, and the model UE_Model#1 belongs to both the function group UE_Functionality#1 and UE_Functionality#2. The first node model is currently working under the function / configuration UE_Functionality#1. When the second node detects that the current model performance has dropped to a poor level, the second node instructs the first node to deactivate / switch to UE_Functionality#2 for the model group / function group / configuration group through high-layer or physical-layer signaling (e.g., RRC / MAC CE / DCI). At this time, the terminal does not need to switch the model for subsequent inference work, which can further reduce the latency of the management process (e.g., activation / deactivation / switching operation latency).

[0218] In some embodiments, the method may further include the following steps:

[0219] F1. Receive the fourth indication information sent by the second node.

[0220] In some embodiments, when the second node detects that the performance of the information processing unit has decreased over a period of time, or detects that the performance of the information processing unit is less than or equal to a preset performance threshold over a period of time, the second node sends the fourth indication information to the first node. Correspondingly, the first node receives the fourth indication information sent by the second node, and the fourth indication information is used for at least one of the following: activation of some information processing units, deactivation of some information processing units, and switching of some information processing units.

[0221] F2. In response to the fourth indication information, send decision information to the second node.

[0222] Wherein, the decision information is used to characterize the decision made by the first node within a preset duration.

[0223] Taking the information processing unit as an example of the model, as a possible example, when the second node detects a decline in model performance over a period of time, the second node may instruct the authorized first node to perform at least one of the following operations on a part of the model: partial model activation / deactivation / switching, etc. Then, the first node will report the decision information within a preset duration to the second node, so that the second node can have a certain understanding of the decision of the first node and make subsequent decisions. It should be noted that the authorized partial model may be opaque to the second node (i.e., the second node can make management decisions on the model). For example, the first node has 4 related models, UE_Model#1 to UE_Model#4 (where the models may belong to the same or different model groups / function groups / configuration groups). The second node instructs the authorized first node to perform model activation / deactivation / switching only on UE_Model#1 and UE_Model#3 through high-layer or physical-layer signaling (e.g., RRC / MAC CE / DCI). Since the first node makes management operation decisions, this will greatly reduce the latency and signaling overhead of the model management process. When the second node / first node detects a decline in model performance, the first node deactivates the current model accordingly and activates / switches to the authorized partial model for the next inference process. The first node needs to report decision information (e.g., switch 4 times within 10ms, identification information of the switched model, etc.) to the second node periodically or aperiodically within a preset duration. The second node makes the next management decision based on the reported decision information, such as deactivating the model / model group, switching to another model / model group, updating the current model / model group, or reverting to the default method, etc.

[0224] In some embodiments, the method may further include the following steps:

[0225] G1. Receive the fifth indication information sent by the second node.

[0226] Wherein, the fifth indication information is used to indicate the switching duration threshold when the first node performs the information processing unit switching.

[0227] In this way, when the first node performs the information processing unit switching, it can perform the information processing unit switching within the switching duration threshold indicated by the second node based on the fifth indication information, so that the information processing unit on the first node side can match the information processing unit on the second node side, improving the matching accuracy between the information processing unit on the first node side and the information processing unit on the second node side, that is, improving the matching accuracy between the information processing unit on the terminal side and the information processing unit on the base station side, and preventing the performance decline of the information processing unit.

[0228] As a possible example, the fifth indication information can be used to indicate multiple candidate handover duration thresholds. After receiving the fifth indication information, the first node can determine a target handover duration threshold from the candidate handover duration thresholds based on the fifth indication information, and then send the target handover duration threshold to the second node. In this way, when the first node performs an information processing unit handover, it can perform the handover of the information processing unit within the target handover duration threshold.

[0229] As another possible example, the fifth indication information can be used to indicate the maximum handover duration threshold, so that when the first node performs an information processing unit handover, it can perform the handover of the information processing unit within the maximum handover duration threshold.

[0230] It should be noted that when the information processing unit is updated, a new identification information of the information processing unit can be assigned or the previous identification information can be used.

[0231] In some embodiments, the first node can also receive the authorization time sent by the second node, and the authorization time is the time length allowed for the first node to perform autonomous model operations. Within the configured time length, the first node is authorized to perform information processing unit operations without sending a request to the second node and obtaining the permission of the second node. The unit of the authorization time can be frame / sub-frame / slot / sub-slot / symbol.

[0232] In some embodiments, within the configured time length, after the first node performs autonomous information processing unit operations, it does not need to send any feedback information to the second node.

[0233] In some embodiments, within the configured time length, after the first node performs autonomous information processing unit operations, it needs to send feedback information to the second node, and this feedback information can include one of the following: the number of information processing unit operations (such as the number of information processing unit handovers, the number of information processing unit activations / deactivations, the number of information processing unit updates), the information processing unit identification (such as the model identification before / after the information processing unit handover, the identification of the activated information processing unit, the identification of the deactivated information processing unit).

[0234] In some embodiments, the method may further include the following steps:

[0235] H1. Receive the sixth indication information sent by the second node.

[0236] In some embodiments, the second node can send the sixth indication information to the first node, and correspondingly, the first node receives the sixth indication information sent by the second node. Among them, the sixth indication information includes a threshold value or a threshold condition, and the sixth indication information is used to indicate the determination of the target first information processing unit based on the threshold value or the threshold condition.

[0237] That is to say, the second node can recommend / indicate a threshold value / threshold condition to the first node, so that the first node can select the corresponding target first information processing unit according to whether the performance evaluation index meets the threshold value. For example, when the second node recommends / indicates that the first node selects the target first information processing unit, it is required that the inference performance index SGCS of the information processing unit > 0.9 can be used. It should be noted that this threshold value / threshold condition also applies to the embodiments in which the first node calculates other performance evaluation indexes, which will not be elaborated here.

[0238] In some embodiments, the method may further include the following steps:

[0239] I1. Send information processing unit status information to the second node.

[0240] Among them, the information processing unit status information includes at least one of the following:

[0241] The identifier of the seventh information processing unit;

[0242] The indication information used to indicate whether the transmission of the eighth information processing unit is supported;

[0243] The identifier of the eighth information processing unit;

[0244] The time information, which is used to characterize at least one of the following: the time required to switch to the eighth information processing unit, the time required to activate the eighth information processing unit, and the time required to deactivate the seventh information processing unit.

[0245] In some embodiments, the seventh information processing unit can be understood as the old information processing unit, and the eighth information processing unit can be understood as the new information processing unit. The identifier of the seventh information processing unit can be understood as the identifier of the old information processing unit that needs to be deactivated or switched, the indication information used to indicate whether the transmission of the eighth information processing unit is supported can be understood as the indication information used to indicate the existence of the eighth information processing unit, and the identifier of the eighth information processing unit can be understood as the identifier of the new information processing unit to be switched or activated or the identifier of the new information processing unit after the update of the old information processing unit. The time information can be understood as the time indicating how long it takes to switch to the new information processing unit or activate the new information processing unit or deactivate the old information processing unit. The time information can be a frame / sub - frame / time slot / sub - time slot / symbol. For example, the first node recommends in the report that the second node switches to the new information processing unit after a specific time slot.

[0246] Taking the information processing unit model as an example, it should be understood that the reference signal resources, CSI reporting, etc. configurations corresponding to different first node-side models may be different, and how to configure resources / reporting is completely controlled by the second node. Therefore, under normal circumstances, the second node should decide on model operations such as activation / deactivation / switching / selection of the first node-side model. However, in some cases, the first node may have a better prediction or understanding of the future working state of the model. For example, the model performance has a great relationship with the running load, computing resources on the first node side, as well as the running trajectory, moving speed, and surrounding environment perceived by the first node. If the first node can predict the future working conditions of the model based on the perception of the moving trajectory, surrounding environment, or its own software / hardware environment, it can actively report this information (i.e., the above-mentioned information processing unit status information) to the second node, thereby assisting the second node in performing pre-switching or pre-deactivation of the model, etc.

[0247] In some embodiments, after step I1, the method may further include the following steps:

[0248] I2. When the response information of the second node is monitored within a specific time window, perform the switching or deactivation operation of the information processing unit after a preset duration.

[0249] For example, after the first node sends the information processing unit status information to the second node, the first node needs to monitor a specific CORESET or a specific search space carrying the response information of the second node within a specific time window. When the response information of the second node is detected, the first node can actively perform information processing unit operations such as switching or deactivating the information processing unit after a preset duration (specific time). Otherwise, the first node should not actively perform information processing unit operations such as switching or deactivating the information processing unit.

[0250] In some embodiments, as Figure 11 shown, the embodiments of the present disclosure further provide a communication method, which is applied to the second node. The second node may be the base station 21 shown above Figure 2 shown, and the method may include the following steps:

[0251] S201. Receive the first signaling sent by the first node.

[0252] Wherein, the first signaling includes the identifier of the first information processing unit.

[0253] S202. Send the second signaling to the first node.

[0254] Wherein, the second signaling is used to determine the target first information processing unit.

[0255] For the descriptions of the first signaling and the second signaling, reference may be made to the aboveFigure 3 The corresponding description in the embodiments shown is not elaborated herein.

[0256] In some embodiments, the method further includes: receiving an identifier of a target first information processing unit sent by a first node, or a performance evaluation index between first information and fourth information. For the description of why the first node sends the identifier of the target first information processing unit or the performance evaluation index between the first information and the fourth information, reference may be made to the corresponding description in the above Figure 3 The corresponding description in the embodiments shown is not elaborated herein.

[0257] In some embodiments, the method further includes: sending first indication information to the first node, where the first indication information is used to indicate a fifth information processing unit. For the description of the first indication information, reference may be made to the corresponding description in the above embodiments and will not be elaborated herein.

[0258] In some embodiments, the method further includes: sending second indication information to the first node, where the second indication information is used for the first node to perform information processing unit selection. For the description of the second indication information, reference may be made to the corresponding description in the above embodiments and will not be elaborated herein.

[0259] In some embodiments, the method further includes: sending third indication information to the first node. For the description of why the second node sends the third indication information to the first node, reference may be made to the corresponding description in the above embodiments and will not be elaborated herein.

[0260] As a possible example, sending the third indication information to the first node may be that when the second node detects that the performance of the information processing unit has decreased within a period of time, or when it detects that the performance of the information processing unit is less than or equal to a preset performance threshold within a period of time, the second node sends the third indication information to the first node.

[0261] In some embodiments, the method further includes: sending fourth indication information to the first node. For the description of why the second node sends the fourth indication information to the first node, reference may be made to the corresponding description in the above embodiments and will not be elaborated herein.

[0262] As a possible example, sending the fourth indication information to the first node may be that when the second node detects that the performance of the information processing unit has decreased within a period of time, or when it detects that the performance of the information processing unit is less than or equal to a preset performance threshold within a period of time, the second node sends the fourth indication information to the first node.

[0263] In some embodiments, after sending the fourth indication information to the first node, the method may further include the following steps: receiving decision information sent by the first node. For the description of the decision information, reference may be made to the corresponding description in the above embodiments and will not be elaborated herein.

[0264] As a possible example, after receiving the decision information, the second node may make a next management decision based on the decision information, such as: deactivating the information processing unit / information processing unit group, switching to another information processing unit / information processing unit group, updating the current information processing unit / information processing unit group, or falling back to the default method, etc.

[0265] In some embodiments, the method further includes: sending fifth indication information to the first node. For the description of the fifth indication information, reference may be made to the corresponding description in the foregoing embodiments, which will not be elaborated herein.

[0266] In some embodiments, the method further includes: sending sixth indication information to the first node. For the description of the sixth indication information, reference may be made to the corresponding description in the foregoing embodiments, which will not be elaborated herein.

[0267] In some embodiments, the second node may further send an authorization time to the first node. For the description of the authorization time, reference may be made to the corresponding description in the foregoing embodiments, which will not be elaborated herein.

[0268] In some embodiments, the second node may further receive feedback information sent by the first node. The feedback information may include one of the following: the number of operations of the information processing unit (such as the number of times of information processing unit switching, the number of times of information processing unit activation / deactivation, the number of times of information processing unit update), the information processing unit identifier (such as the identifier before / after information processing unit switching, the identifier of the activated information processing unit, the identifier of the deactivated information processing unit).

[0269] In some embodiments, the method further includes: receiving the information processing unit status information sent by the first node. For the description of the information processing unit status information, reference may be made to the relevant description in the foregoing embodiments, which will not be elaborated herein.

[0270] In the embodiments of the present disclosure, a pre-operation mechanism for multiple information processing units is proposed, including activation, deactivation, selection, update, switching, etc. of the information processing units. In this way, on the one hand, before performing an operation on the information processing unit, the second node can indicate multiple candidate information processing units in advance and initiate a feasibility test of the information processing unit, and determine the final operation of the information processing unit according to the content reported by the first node (i.e., the first signaling). On the other hand, the first node can report the recommended operation of the information processing unit and the corresponding time information to the second node based on its own environmental perception to assist the second node in making a decision on the final operation of the information processing unit. In addition, the embodiments of the present disclosure propose a pre-authorization mechanism for the operation of the information processing unit, that is, the second node provides information such as the identifier of the authorized information processing unit, the authorization time, and the authorization content to the first node, allowing the first node to perform autonomous operations and information feedback on the information processing unit. Through the communication method provided by the embodiments of the present disclosure, accurate indication of the operation of the information processing unit on the first node side by the second node can be achieved, and signaling overhead and processing delay can be reduced, thereby improving the matching accuracy between the information processing unit on the first node side and the information processing unit on the second node side, and preventing the performance degradation of the information processing unit.

[0271] The above mainly introduces the solution provided by the present disclosure from the perspective of the interaction between each node. It can be understood that each node, such as the first node and the second node, includes corresponding hardware structures and / or software modules for implementing the above functions. Those skilled in the art should easily realize that, in combination with the algorithm steps of each example described in the embodiments disclosed herein, the present invention can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0272] Figure 12 The following shows a schematic diagram of the composition of a communication device provided by an embodiment of the present disclosure. As Figure 12 shown, the communication device 30 includes a sending unit 301, a receiving unit 302, and a processing unit 303.

[0273] The communication device 30 can be the above-mentioned first node or a chip in the first node. When the communication device 30 is used to implement the functions of the first node in the above embodiments, each unit is specifically used to implement the following functions.

[0274] The sending unit 301 is used to send a first signaling to the second node, and the first signaling includes the identifier of the first information processing unit;

[0275] A receiving unit 302, configured to receive a second signaling sent by a second node, where the second signaling is used to determine a target first information processing unit;

[0276] A processing unit 303, configured to determine a target first information processing unit from first information processing units based on the second signaling.

[0277] In some embodiments, the first signaling further includes first information and second information, where the second information is obtained by the first node after processing the first information based on a first information processing unit; the second signaling includes an identifier of the target first information processing unit, and the target first information processing unit is determined by the second node from first information processing units based on a performance evaluation metric between third information and the first information, and the third information is obtained by the second node after processing the second information based on a second information processing unit.

[0278] In some embodiments, the performance evaluation metric includes a first type of evaluation metric, a second type of evaluation metric, and a third type of evaluation metric; the first type of evaluation metric includes an evaluation metric based on distance error and an evaluation metric based on similarity; the evaluation metric based on distance error includes at least one of the following: Euclidean distance, mean square error, normalized mean square error; the evaluation metric based on similarity includes at least one of the following: cosine similarity, generalized cosine similarity, cross entropy; the second type of evaluation metric includes an evaluation metric based on throughput and an evaluation metric based on channel quality; the evaluation metric based on throughput includes at least one of the following: system throughput, user throughput, average user throughput, 5% user throughput, spectral efficiency; the evaluation metric based on channel quality includes at least one of the following: bit error rate, block error rate, assumed bit error rate, assumed block error rate, signal-to-noise ratio, signal-to-interference-plus-noise ratio, modulation and coding scheme (MCS) transmission level; the third type of evaluation metric includes an evaluation metric based on data distribution error, and the evaluation metric based on data distribution error includes at least one of the following: power spectrum, energy spectrum, amplitude spectrum, phase spectrum.

[0279] In some embodiments, the second information processing unit is an information processing unit matching the first information processing unit.

[0280] In some embodiments, the processing unit 303 is specifically configured to determine the target first information processing unit from first information processing units based on the identifier of the target first information processing unit included in the second signaling.

[0281] In some embodiments, the first signaling further includes second information, where the second information is obtained by the first node after processing the first information based on a first information processing unit; the second signaling includes third information, and the third information is obtained by the second node after processing the second information based on a second information processing unit.

[0282] In some embodiments, the processing unit 303 is specifically configured to determine a performance evaluation metric between the first information and the third information; and determine a target first information processing unit from the first information processing units based on the performance evaluation metric between the first information and the third information.

[0283] In some embodiments, the processing unit 303 is specifically configured to determine, as the target first information processing unit, the first information processing unit that matches the second information processing unit corresponding to the optimal performance evaluation metric.

[0284] In some embodiments, the sending unit 301 is further configured to send an identifier of the target first information processing unit to the second node.

[0285] In some embodiments, the second signaling includes a third information processing unit, and the third information processing unit is an information processing unit determined by the second node based on the identifier of the first information processing unit and that matches the first information processing unit.

[0286] In some embodiments, the processing unit 303 is specifically configured to process the first information based on the third information processing unit to obtain fourth information; determine a performance evaluation metric between the first information and the fourth information; and determine a target first information processing unit from the first information processing units based on the performance evaluation metric between the first information and the fourth information.

[0287] In some embodiments, the sending unit 301 is further configured to send an identifier of the target first information processing unit or the performance evaluation metric between the first information and the fourth information to the second node.

[0288] In some embodiments, the first signaling further includes performance metric information, and the performance metric information includes a performance evaluation metric between the first information and a fifth information, where the fifth information is obtained by the first node after processing the second information based on a fourth information processing unit, and the second information is obtained by the first node after processing the first information based on a first information processing unit; and the second signaling includes an identifier of the target first information processing unit, and the target first information processing unit is determined by the second node from the first information processing units based on the performance metric information.

[0289] In some embodiments, the processing unit 303 is specifically configured to determine the target first information processing unit from the first information processing units based on the identifier of the target first information processing unit included in the second signaling.

[0290] In some embodiments, the first signaling further includes performance metric information, which is a performance evaluation metric between the first information and the fifth information. The fifth information is obtained by the first node after processing the second information based on the fourth information processing unit, and the second information is obtained by the first node after processing the first information based on the first information processing unit; the second signaling includes an identifier of a candidate first information processing unit, and the candidate first information processing unit is determined by the second node from the first information processing units based on the performance metric information.

[0291] In some embodiments, the processing unit 303 is specifically configured to determine a target first information processing unit from the first information processing units based on the attribute information of the first node and the identifier of the candidate first information processing unit.

[0292] In some embodiments, the sending unit 301 is further configured to send the identifier of the target first information processing unit to the second node.

[0293] In some embodiments, the first signaling further includes the first information and the second information, where the second information is obtained by the first node after processing the first information based on the first information processing unit; the second signaling includes an identifier of a candidate first information processing unit, and the candidate first information processing unit is determined by the second node from the first information processing units based on the performance evaluation metric between the third information and the first information. The third information is obtained by the second node after processing the second information based on the second information processing unit.

[0294] In some embodiments, the processing unit 303 is specifically configured to determine a target first information processing unit from the first information processing units based on the attribute information of the first node and the identifier of the candidate first information processing unit.

[0295] In some embodiments, the sending unit 301 is further configured to send the identifier of the target first information processing unit to the second node.

[0296] In some embodiments, the identifier of the first information processing unit is a single-layer structure, and the identifier of the first information processing unit is indicated, assigned, or configured by the second node.

[0297] In some embodiments, the sorting method of the identifier of the first information processing unit is related to the generation order of the first information processing unit.

[0298] In some embodiments, the identifier of the first information processing unit is a multi-layer structure, and the identifier of the first information processing unit is determined based on the identifier information of the data set corresponding to the first information processing unit and the identifier information of the first information processing unit.

[0299] In some embodiments, the receiving unit 302 is further configured to receive first indication information sent by the second node, where the first indication information is used to indicate a fifth information processing unit;

[0300] The processing unit 303 is further configured to determine a sixth information processing unit that matches the fifth information processing unit based on the first indication information.

[0301] In some embodiments, the sending unit 301 is further configured to send an identifier of the sixth information processing unit to the second node.

[0302] In some embodiments, the receiving unit 302 is further configured to receive second indication information sent by the second node, where the second indication information is used for the first node to perform information processing unit selection.

[0303] In some embodiments, the second indication information includes at least one of the following: structure information of the information processing unit that the first node should use; structure information of the information processing unit used by the second node.

[0304] In some embodiments, the second indication information further includes at least one of the following: information processing unit complexity of the information processing unit; computing complexity of the information processing unit; inference time range of the information processing unit; storage overhead range of the information processing unit.

[0305] In some embodiments, the receiving unit 302 is further configured to receive third indication information sent by the second node, where the third indication information is used for at least one of the following: information processing unit group activation, information processing unit group deactivation, information processing unit group switching, information processing unit activation, information processing unit deactivation, information processing unit switching, information processing unit update, fallback to the default information processing method, information processing unit group update.

[0306] In some embodiments, the receiving unit 302 is further configured to receive fourth indication information sent by the second node, where the fourth indication information is used for at least one of the following: partial information processing unit activation, partial information processing unit deactivation, partial information processing unit switching;

[0307] The sending unit 301 is further configured to send decision information to the second node in response to the fourth indication information, where the decision information is used to represent the decision made by the first node within a preset time period.

[0308] In some embodiments, the receiving unit 302 is further configured to receive fifth indication information sent by the second node, where the fifth indication information is used to indicate a switching duration threshold when the first node performs information processing unit switching.

[0309] In some embodiments, the receiving unit 302 is further configured to receive sixth indication information sent by the second node, where the sixth indication information includes a threshold value or a threshold condition, and the sixth indication information is used to indicate a target first information processing unit determined based on the threshold value or the threshold condition.

[0310] In some embodiments, the sending unit 301 is further configured to send information processing unit status information to the second node, where the information processing unit status information includes at least one of the following: an identifier of a seventh information processing unit; indication information used to indicate whether transmission of an eighth information processing unit is supported; an identifier of the eighth information processing unit; time information, where the time information is used to characterize at least one of the following: the time required to switch to the eighth information processing unit, the time required to activate the eighth information processing unit, and the time required to deactivate the seventh information processing unit.

[0311] In some embodiments, the processing unit 303 is further configured to, when response information of the second node is monitored within a specific time window, perform a switching or deactivation operation of the information processing unit after a preset duration.

[0312] Figure 13 The following shows a schematic diagram of the composition of another communication device provided by an embodiment of the present disclosure. As Figure 13 shown, the communication device 40 includes a receiving unit 401 and a sending unit 402.

[0313] The communication device 40 may be the above-mentioned second node or a chip in the second node. When the communication device 40 is used to implement the functions of the second node in the above embodiments, each unit is specifically configured to implement the following functions.

[0314] The receiving unit 401 is configured to receive a first signaling sent by the first node, where the first signaling includes an identifier of a first information processing unit;

[0315] The sending unit 402 is configured to send a second signaling to the first node, where the second signaling is used to determine a target first information processing unit.

[0316] In some embodiments, the first signaling further includes first information and second information, where the second information is obtained by the first node after processing the first information based on the first information processing unit; the second signaling includes an identifier of the target first information processing unit, and the target first information processing unit is determined by the second node from the first information processing units based on a performance evaluation index between third information and the first information, and the third information is obtained by the second node after processing the second information based on the second information processing unit.

[0317] In some embodiments, the performance evaluation metrics include first - type evaluation metrics, second - type evaluation metrics, and third - type evaluation metrics; the first - type evaluation metrics include evaluation metrics based on distance error and evaluation metrics based on similarity; the evaluation metrics based on distance error include at least one of the following: Euclidean distance, mean square error, normalized mean square error; the evaluation metrics based on similarity include at least one of the following: cosine similarity, generalized cosine similarity, cross - entropy; the second - type evaluation metrics include evaluation metrics based on throughput and evaluation metrics based on channel quality; the evaluation metrics based on throughput include at least one of the following: system throughput, user throughput, average user throughput, 4% user throughput, spectral efficiency; the evaluation metrics based on channel quality include at least one of the following: bit error rate, block error rate, assumed bit error rate, assumed block error rate, signal - to - noise ratio, signal - to - interference - plus - noise ratio, modulation and coding scheme (MCS) transmission level; the third - type evaluation metrics include evaluation metrics based on data distribution error, and the evaluation metrics based on data distribution error include at least one of the following: power spectrum, energy spectrum, amplitude spectrum, phase spectrum.

[0318] In some embodiments, the second information processing unit is an information processing unit that matches the first information processing unit.

[0319] In some embodiments, the target first information processing unit is determined from the first information processing units based on the identifier of the target first information processing unit included in the second signaling.

[0320] In some embodiments, the first signaling further includes second information, where the second information is obtained by the first node after processing the first information based on the first information processing unit; the second signaling includes third information, and the third information is obtained by the second node after processing the second information based on the second information processing unit.

[0321] In some embodiments, the target first information processing unit is determined from the first information processing units based on the performance evaluation metrics between the first information and the third information.

[0322] In some embodiments, the target first information processing unit is the first information processing unit that matches the second information processing unit corresponding to the optimal performance evaluation metric.

[0323] In some embodiments, the receiving unit 401 is further configured to receive the identifier of the target first information processing unit sent by the first node.

[0324] In some embodiments, the second signaling includes a third information processing unit, and the third information processing unit is an information processing unit that the second node determines to match the first information processing unit based on the identifier of the first information processing unit.

[0325] In some embodiments, the target first information processing unit is determined from the first information processing units based on a performance evaluation metric between the first information and the fourth information, where the fourth information is obtained by processing the first information based on a third information processing unit.

[0326] In some embodiments, the receiving unit 401 is further configured to receive an identifier of the target first information processing unit sent by the first node or a performance evaluation metric between the first information and the fourth information.

[0327] In some embodiments, the first signaling further includes performance metric information, where the performance metric information includes a performance evaluation metric between the first information and the fifth information, and the fifth information is obtained by the first node processing the second information based on a fourth information processing unit, and the second information is obtained by the first node processing the first information based on a first information processing unit; the second signaling includes an identifier of the target first information processing unit, and the target first information processing unit is determined by the second node from the first information processing units based on the performance metric information.

[0328] In some embodiments, the target first information processing unit is determined from the first information processing units based on the identifier of the target first information processing unit included in the second signaling.

[0329] In some embodiments, the first signaling further includes performance metric information, where the performance metric information includes a performance evaluation metric between the first information and the fifth information, and the fifth information is obtained by the first node processing the second information based on a fourth information processing unit, and the second information is obtained by the first node processing the first information based on a first information processing unit; the second signaling includes an identifier of a candidate first information processing unit, and the candidate first information processing unit is determined by the second node from the first information processing units based on the performance metric information.

[0330] In some embodiments, the target first information processing unit is determined from the first information processing units based on the attribute information of the first node and the identifier of the candidate first information processing unit.

[0331] In some embodiments, the receiving unit 401 is further configured to receive an identifier of the target first information processing unit sent by the first node.

[0332] In some embodiments, the first signaling further includes the first information and the second information, where the second information is obtained by the first node processing the first information based on a first information processing unit; the second signaling includes an identifier of a candidate first information processing unit, and the candidate first information processing unit is determined by the second node from the first information processing units based on a performance evaluation metric between the third information and the first information, and the third information is obtained by the second node processing the second information based on a second information processing unit.

[0333] In some embodiments, the target first information processing unit is determined from the first information processing units based on the attribute information of the first node and the identifier of the candidate first information processing unit.

[0334] In some embodiments, the receiving unit 401 is further configured to receive the identifier of the target first information processing unit sent by the first node.

[0335] In some embodiments, the identifier of the first information processing unit is a single-layer structure, and the identifier of the first information processing unit is indicated, allocated, or configured by the second node.

[0336] In some embodiments, the sorting method of the identifier of the first information processing unit is related to the generation order of the first information processing unit.

[0337] In some embodiments, the identifier of the first information processing unit is a multi-layer structure, and the identifier of the first information processing unit is determined based on the identifier information of the data set corresponding to the first information processing unit and the identifier information of the first information processing unit.

[0338] In some embodiments, the sending unit 402 is further configured to send first indication information to the first node, where the first indication information is used to indicate the fifth information processing unit.

[0339] In some embodiments, the sending unit 402 is further configured to send second indication information to the first node, where the second indication information is used for the first node to select an information processing unit.

[0340] In some embodiments, the second indication information includes at least one of the following: the structure information of the information processing unit that the first node should use; the structure information of the information processing unit used by the second node.

[0341] In some embodiments, the second indication information further includes at least one of the following: the information processing unit complexity of the information processing unit; the computing complexity of the information processing unit; the inference time range of the information processing unit; the storage overhead range of the information processing unit.

[0342] In some embodiments, the sending unit 402 is further configured to send third indication information to the first node, where the third indication information is used for at least one of the following: information processing unit group activation, information processing unit group deactivation, information processing unit group switching, information processing unit activation, information processing unit deactivation, information processing unit switching, information processing unit update, fallback to the default information processing method, information processing unit group update.

[0343] In some embodiments, the sending unit 402 is further configured to send fourth indication information to the first node, where the fourth indication information is used for at least one of the following: activating a partial information processing unit, deactivating a partial information processing unit, and switching a partial information processing unit;

[0344] The receiving unit 401 is further configured to receive decision information sent by the first node, where the decision information is used to represent the decisions made by the first node within a preset time period.

[0345] In some embodiments, the sending unit 402 is further configured to send fifth indication information to the first node, where the fifth indication information is used to indicate a switching duration threshold when the first node performs an information processing unit switch.

[0346] In some embodiments, the receiving unit 401 is further configured to receive information processing unit status information sent by the first node, where the information processing unit status information includes at least one of the following: an identifier of a seventh information processing unit; indication information used to indicate whether transmission of an eighth information processing unit is supported; an identifier of the eighth information processing unit; time information, where the time information is used to represent at least one of the following: the time required to switch to the eighth information processing unit, the time required to activate the eighth information processing unit, and the time required to deactivate the seventh information processing unit.

[0347] Figure 12 and Figure 13 When each unit in is implemented in the form of a software functional module and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present disclosure, in essence, or the part that contributes to the prior art, or all or part of this technical solution, may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods of the various embodiments of the present disclosure. The storage media storing the computer software product include: USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, and other media that can store program codes.

[0348] In the case of implementing the functions of the above integrated module in the form of hardware, the embodiments of the present disclosure provide a structural schematic diagram of a communication device, and the communication device may be the above communication device 30 or communication device 40. As Figure 14 shown, the communication device 50 includes: a processor 502, a communication interface 503, and a bus 504. Optionally, the communication device 50 may further include a memory 501.

[0349] The processor 502 can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the present disclosure. The processor 502 can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the present disclosure. The processor 502 can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0350] The communication interface 503 is used to connect to other devices through a communication network. The communication network can be an Ethernet, a wireless access network, a wireless local area network (WLAN), etc.

[0351] The memory 501 can be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM), or other type of dynamic storage device that can store information and instructions. It can also be an electrically erasable programmable read-only memory (EEPROM), a magnetic disk storage medium, or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0352] As a possible implementation, the memory 501 can exist independently of the processor 502. The memory 501 can be connected to the processor 502 through the bus 504 for storing instructions or program code. When the processor 502 calls and executes the instructions or program code stored in the memory 501, the communication method provided by the embodiments of the present disclosure can be implemented.

[0353] In another possible implementation, the memory 501 can also be integrated with the processor 502.

[0354] The bus 504 can be an extended industry standard architecture (EISA) bus, etc. The bus 504 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 14It is represented only by a thick line, but it does not mean that there is only one bus or one type of bus.

[0355] Through the description of the above embodiments, those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above models is used as an example.

[0356] In practical applications, the above functions can be allocated to different models according to needs, that is, the internal structure of the base station or terminal is divided into different models to complete all or part of the functions described above.

[0357] The embodiments of the present disclosure also provide a computer-readable storage medium. All or part of the processes in the above method embodiments can be completed by computer instructions instructing relevant hardware. The program can be stored in the above computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. The computer-readable storage medium can be the memory of any of the foregoing embodiments. The above computer-readable storage medium can also be an external storage device of the first node or the second node, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the first node or the second node. Further, the above computer-readable storage medium can also include both the internal storage unit of the first node or the second node and the external storage device. The above computer-readable storage medium is used to store the above computer program and other programs and data required by the first node or the second node. The above computer-readable storage medium can also be used to temporarily store the data that has been output or will be output.

[0358] The embodiments of the present disclosure also provide a computer program product. The computer product includes a computer program. When the computer program product runs on a computer, the computer is caused to execute any one of the communication methods provided in the above embodiments.

[0359] Although the present disclosure has been described in conjunction with various embodiments herein, however, in the process of implementing the claimed present disclosure, those skilled in the art can understand and realize other variations of the disclosed embodiments by viewing the drawings, the disclosure content, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps.

[0360] "A" or "an" does not exclude a plurality of cases. A single processor or other unit can implement several functions recited in the claims. Certain measures are recited in mutually different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0361] Although the present disclosure has been described in connection with specific features and their embodiments, it will be apparent that various modifications and combinations can be made thereto without departing from the spirit and scope of the present disclosure. Accordingly, the specification and drawings are merely illustrative descriptions of the present disclosure defined by the appended claims and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present disclosure. Obviously, those skilled in the art can make various changes and modifications to the present disclosure without departing from the spirit and scope of the present disclosure. Thus, if these modifications and variations of the present disclosure fall within the scope of the claims of the present disclosure and their equivalent technologies, the present disclosure is also intended to include these changes and modifications.

[0362] The above is only the specific implementation manner of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions within the technical scope disclosed by the present disclosure should be covered by the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the protection scope of the claims.

Claims

1. A communication method, characterized in that, Applied to a first node, the method includes: Sending a first signaling to a second node, the first signaling including an identifier of a first information processing unit; Receiving a second signaling sent by the second node, the second signaling being used to determine a target first information processing unit; Determining the target first information processing unit from the first information processing units based on the second signaling.

2. The method according to claim 1, characterized in that, The first signaling further includes first information and second information, wherein the second information is obtained by the first node after processing the first information based on the first information processing unit; The second signaling includes an identifier of the target first information processing unit, and the target first information processing unit is determined by the second node from the first information processing units based on a performance evaluation metric between third information and the first information, and the third information is obtained by the second node after processing the second information based on a second information processing unit; 3. The method according to claim 2, wherein The performance evaluation metric includes a first type of evaluation metric, a second type of evaluation metric, and a third type of evaluation metric; The first type of evaluation metric includes an evaluation metric based on distance error and an evaluation metric based on similarity; the evaluation metric based on distance error includes at least one of the following: Euclidean distance, mean square error, normalized mean square error; the evaluation metric based on similarity includes at least one of the following: cosine similarity, generalized cosine similarity, cross entropy; The second type of evaluation metric includes an evaluation metric based on throughput and an evaluation metric based on channel quality; the evaluation metric based on throughput includes at least one of the following: system throughput, user throughput, average user throughput, 5% user throughput, spectral efficiency; the evaluation metric based on channel quality includes at least one of the following: bit error rate, block error rate, assumed bit error rate, assumed block error rate, signal-to-noise ratio, signal-to-interference-plus-noise ratio, modulation and coding scheme MCS transmission level; The third type of evaluation metric includes an evaluation metric based on data distribution error, and the evaluation metric based on data distribution error includes at least one of the following: power spectrum, energy spectrum, amplitude spectrum, phase spectrum.

4. The method according to claim 2, wherein The second information processing unit is an information processing unit matching the first information processing unit.

5. The method according to claim 2, wherein The determining the target first information processing unit from the first information processing units based on the second signaling includes: Determining the target first information processing unit from the first information processing units based on the identifier of the target first information processing unit included in the second signaling.

6. The method according to claim 1, characterized in that, The first signaling further includes second information, wherein the second information is obtained by the first node after processing first information based on the first information processing unit; The second signaling includes third information, and the third information is obtained by the second node after processing the second information based on the second information processing unit.

7. The method according to claim 6, characterized in that, The determining the target first information processing unit from the first information processing units based on the second signaling includes: Determining a performance evaluation metric between the first information and the third information; Determine the target first information processing unit from the first information processing units based on the performance evaluation index between the first information and the third information.

8. The method according to claim 7, characterized in that, The determining of the target first information processing unit from the first information processing units based on the performance evaluation index between the first information and the third information includes: Determine the first information processing unit that matches the second information processing unit corresponding to the optimal performance evaluation index as the target first information processing unit.

9. The method according to claim 7, wherein The method further includes: Send the identifier of the target first information processing unit to the second node.

10. The method according to claim 1, wherein The second signaling includes a third information processing unit, and the third information processing unit is an information processing unit that the second node determines to match the first information processing unit based on the identifier of the first information processing unit.

11. The method according to claim 10, characterized in that, The determining of the target first information processing unit from the first information processing units based on the second signaling includes: Process the second information based on the third information processing unit to obtain fourth information, where the second information is obtained by the first node processing the first information based on the first information processing unit; Determine the performance evaluation index between the first information and the fourth information; Determine the target first information processing unit from the first information processing units based on the performance evaluation index between the first information and the fourth information.

12. The method according to claim 11, wherein The method further includes: Send the identifier of the target first information processing unit or the performance evaluation index between the first information and the fourth information to the second node.

13. The method according to claim 1, characterized in that, The first signaling further includes performance index information, and the performance index information includes the performance evaluation index between the first information and the fifth information, where the fifth information is obtained by the first node processing the second information based on the fourth information processing unit, and the second information is obtained by the first node processing the first information based on the first information processing unit; The second signaling includes the identifier of the target first information processing unit, and the target first information processing unit is determined by the second node from the first information processing units based on the performance index information.

14. The method according to claim 13, wherein The determining of the target first information processing unit from the first information processing units based on the second signaling includes: Determine the target first information processing unit from the first information processing units based on the identifier of the target first information processing unit included in the second signaling.

15. The method according to claim 1, characterized in that The first signaling further includes performance index information, and the performance index information includes the performance evaluation index between the first information and the fifth information, where the fifth information is obtained by the first node processing the second information based on the fourth information processing unit, and the second information is obtained by the first node processing the first information based on the first information processing unit; The second signaling includes the identifier of the candidate first information processing unit, and the candidate first information processing unit is determined by the second node from the first information processing units based on the performance index information.

16. The method according to claim 15, wherein Determining the target first information processing unit from the first information processing units based on the second signaling includes: Determining the target first information processing unit from the first information processing units based on the attribute information of the first node and the identifier of the candidate first information processing unit.

17. The method according to claim 16, characterized in that, The method further includes: Sending the identifier of the target first information processing unit to the second node.

18. The method according to claim 1, characterized in that The first signaling further includes first information and second information, where the second information is obtained by the first node processing the first information based on the first information processing unit; The second signaling includes the identifier of the candidate first information processing unit, and the candidate first information processing unit is determined by the second node from the first information processing units based on the performance evaluation index between the third information and the first information, and the third information is obtained by the second node processing the second information based on the second information processing unit.

19. The method according to claim 18, wherein Determining the target first information processing unit from the first information processing units based on the second signaling includes: Determining the target first information processing unit from the first information processing units based on the attribute information of the first node and the identifier of the candidate first information processing unit.

20. The method according to claim 19, wherein The method further includes: Sending the identifier of the target first information processing unit to the second node.

21. The method according to claim 1, characterized in that, The identifier of the first information processing unit is of a single-layer structure, and the identifier of the first information processing unit is indicated, assigned, or configured by the second node.

22. The method according to claim 1, characterized in that, The sorting method of the identifier of the first information processing unit is related to the generation order of the first information processing unit.

23. The method according to claim 1, wherein The identifier of the first information processing unit is of a multi-layer structure, and the identifier of the first information processing unit is determined based on the corresponding identifier set information of the first information processing unit and the identifier information of the first information processing unit.

24. The method according to claim 1, wherein The method further includes: Receiving first indication information sent by the second node, where the first indication information is used to indicate the fifth information processing unit; Determining a sixth information processing unit that matches the fifth information processing unit based on the first indication information.

25. The method according to claim 24, wherein The method further includes: Sending the identifier of the sixth information processing unit to the second node.

26. The method according to claim 1, wherein The method further includes: Receiving second indication information sent by the second node, where the second indication information is used for the first node to select an information processing unit.

27. The method according to claim 26, wherein The second indication information includes at least one of the following: The structure information of the information processing unit that the first node should use; The structure information of the information processing unit used by the second node.

28. The method according to claim 27, wherein The second indication information further includes at least one of the following: The information processing unit complexity of the information processing unit; The computing complexity of the information processing unit; The inference time range of the information processing unit; The storage overhead range of the information processing unit.

29. The method according to claim 1, wherein The method further includes: Receiving third indication information sent by the second node, where the third indication information is used for at least one of the following: Activation of information processing unit group, deactivation of information processing unit group, switching of information processing unit group, activation of information processing unit, deactivation of information processing unit, switching of information processing unit, update of information processing unit, fallback to default information processing mode, update of information processing unit group.

30. The method according to claim 1, wherein The method further includes: Receiving fourth indication information sent by a second node, where the fourth indication information is used for at least one of the following: Activation of some information processing units, deactivation of some information processing units, switching of some information processing units; In response to the fourth indication information, sending decision information to the second node, where the decision information is used to represent the decisions made by the first node within a preset time period.

31. The method according to claim 1, characterized in that, The method further includes: Receiving fifth indication information sent by a second node, where the fifth indication information is used to indicate a switching duration threshold when the first node performs an information processing unit switching.

32. The method according to claim 1, characterized in that, The method further includes: Receiving sixth indication information sent by a second node, where the sixth indication information includes a threshold value or a threshold condition, and the sixth indication information is used to indicate determining a target first information processing unit based on the threshold value or the threshold condition.

33. The method according to claim 1, characterized in that, The method further includes: Sending information processing unit status information to the second node, where the information processing unit status information includes at least one of the following: An identifier of a seventh information processing unit; Indication information used to indicate whether transmission of an eighth information processing unit is supported; An identifier of the eighth information processing unit; Time information, where the time information is used to represent at least one of the following: Time required to switch to the eighth information processing unit, time required to activate the eighth information processing unit, time required to deactivate the seventh information processing unit.

34. The method according to claim 33, wherein The method further includes: In the case of hearing response information of the second node within a specific time window, performing a switching or deactivation operation of the information processing unit after a preset time period.

35. A communication method, characterized in that, Applied to a second node, the method includes: Receiving a first signaling sent by a first node, where the first signaling includes an identifier of a first information processing unit; Sending a second signaling to the first node, where the second signaling is used to determine a target first information processing unit.

36. The method according to claim 35, wherein The first signaling further includes second information, where the second information is obtained by the first node after processing first information based on the first information processing unit; The second signaling includes third information, where the third information is obtained by the second node after processing the second information based on a second information processing unit.

37. The method according to claim 35, wherein The second signaling includes a third information processing unit, where the third information processing unit is an information processing unit determined by the second node based on the identifier of the first information processing unit and matching the first information processing unit.

38. A communication device, characterized in that, It includes a memory, a processor, and computer program instructions stored on the memory and executable on the processor. When the processor executes the computer program instructions, the method according to any one of claims 1 to 34, or, according to any one of claims 35 to 37 is implemented.

39. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes computer program instructions; wherein, when the computer program instructions run on a computer, the computer is caused to execute the method according to any one of claims 1 to 34, or claims 35 to 37.

Images