Channel state information (CSI) feedback method, terminal equipment and network equipment

Abstract

A channel state information (CSI) feedback method, a terminal device and a network device, the method comprising: a terminal device processing first input information to obtain target input information (S1010), the first input information being CSI data obtained based on a target configuration, the dimension of the target input information being the same as the dimension of CSI data obtained based on a reference configuration; encoding the target input information based on an encoder to obtain a target bit stream (S1020); sending the target bit stream to the network device (S1030); wherein the encoder comprises an adaptation layer, the input dimension of the adaptation layer is the output dimension of the first network, the output dimension of the adaptation layer is determined according to the length of the bit stream needing to be fed back by the target configuration, and the first network is an upper-level network of the adaptation layer in the encoder.

Description

Channel state information (CSI) feedback method, terminal device, and network device Technical Field

[0001] The embodiments of the present application relate to the field of communications, and specifically to a method, terminal device, and network device for feedback of channel state information (CSI). Background Art

[0002] In related technologies, an artificial intelligence (AI)-based channel state information (CSI) autoencoder can be used for CSI feedback. The CSI autoencoder used in this feedback scheme needs to be trained on a large-scale dataset. During deployment, the number of antenna ports and subbands actually configured must be consistent with the dataset used when training the model. Otherwise, the CSI autoencoder cannot be used for CSI feedback, thus reducing the flexibility of the AI-based CSI feedback scheme.

[0003] Summary of the Invention

[0004] The present application provides a method, terminal device, and network device for channel state information (CSI) feedback, which can adapt to CSI feedback under various physical resource configurations at a relatively low model update cost, thereby improving the flexibility and generalization performance of AI-based feedback.

[0005] In a first aspect, a method for channel state information (CSI) feedback is provided, comprising: a terminal device processes first input information to obtain target input information, wherein the first input information is CSI data obtained based on a target configuration, and the dimension of the target input information is the same as the dimension of the CSI data obtained based on a reference configuration; encoding the target input information based on an encoder to obtain a target bit stream; and sending the target bit stream to a network device; wherein the encoder includes an adaptation layer, the input dimension of the adaptation layer is the output dimension of a first network, the output dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back for the target configuration, and the first network is the upper-level network of the adaptation layer in the encoder.

[0006] In a second aspect, a method for channel state information (CSI) feedback is provided, including: a network device receives a target bit stream sent by a terminal device, wherein the target bit stream is a bit stream that needs to be fed back under a target configuration; decoding the target bit stream based on a decoder to obtain first output information, wherein the dimension of the first output information is the same as the dimension of CSI data obtained based on a reference configuration; processing the first output information to obtain target output information, wherein the dimension of the target output information is the same as the dimension of the CSI data obtained based on the target configuration; wherein the decoder includes an adaptation layer, the input dimension of the adaptation layer is determined according to the length of the bit stream that needs to be fed back for the target configuration, and the output dimension of the adaptation layer is the input dimension of a second network, and the second network is the next level network of the adaptation layer in the decoder.

[0007] According to a third aspect, a method for channel state information (CSI) feedback is provided, comprising: a terminal device receiving fourth indication information sent by a network device, wherein the fourth indication information is used to indicate whether an encoder deployed on the terminal device supports CSI data of multiple dimensions.

[0008] In a fourth aspect, a method for channel state information (CSI) feedback is provided, comprising: a network device sends fourth indication information to a terminal device, wherein the fourth indication information is used to indicate whether an encoder deployed on the terminal device supports CSI data of multiple dimensions.

[0009] In a fifth aspect, a terminal device is provided for executing the method in the above-mentioned first aspect or its various implementation modes.

[0010] Specifically, the terminal device includes a functional module for executing the method in the above-mentioned first aspect or third aspect or its respective implementation manner.

[0011] In a sixth aspect, a network device is provided for executing the method in the above second aspect or its various implementations.

[0012] Specifically, the network device includes a functional module for executing the method in the above-mentioned second aspect or fourth aspect or its respective implementation manner.

[0013] In a seventh aspect, a terminal device is provided, comprising a processor and a memory. The memory is configured to store a computer program, and the processor is configured to call and execute the computer program stored in the memory to perform the method of the first aspect or the third aspect or any implementation thereof.

[0014] In an eighth aspect, a network device is provided, comprising a processor and a memory. The memory is configured to store a computer program, and the processor is configured to call and execute the computer program stored in the memory to perform the method of the second aspect or the fourth aspect or any implementation thereof.

[0015] In a ninth aspect, a chip is provided for implementing the method described in any one of the first to fourth aspects or their respective implementations. Specifically, the chip includes a processor configured to retrieve and execute a computer program from a memory, causing a device equipped with the chip to perform the method described in any one of the first to fourth aspects or their respective implementations.

[0016] In a tenth aspect, a computer-readable storage medium is provided for storing a computer program, which enables a computer to execute the method of any one of the first to fourth aspects or its various implementations.

[0017] In an eleventh aspect, a computer program product is provided, comprising computer program instructions, wherein the computer program instructions enable a computer to execute the method of any one of the first to fourth aspects or any of their implementations.

[0018] In the twelfth aspect, a computer program is provided, which, when executed on a computer, enables the computer to execute the method in any one of the above-mentioned first to fourth aspects or their respective implementations.

[0019] Through the above technical solution, the design of the CSI autoencoder based on the adaptation layer adapts to CSI data with variable dimensions by designing an adaptation layer. In the case of configuration updates, there is no need for a large number of model updates, and only the adaptation layer model needs to be updated. This enables the AI-based CSI feedback solution to adapt to CSI feedback under various physical resource configurations at a lower model update cost, thereby improving the flexibility and generalization performance of AI-based feedback. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] FIG1 is a schematic diagram of a communication system architecture provided in an embodiment of the present application.

[0021] FIG2 is a schematic diagram of a neuron structure.

[0022] FIG3 is a schematic diagram of a neural network.

[0023] FIG4 is a schematic diagram of a convolutional neural network.

[0024] FIG5 is a schematic diagram of an LSTM unit.

[0025] FIG6 is a schematic structural diagram of a CSI autoencoder.

[0026] FIG7 is a schematic interaction diagram of a channel state information (CSI) feedback method provided in the present application.

[0027] FIG8 is a schematic structural diagram of a CSI autoencoder provided in the present application.

[0028] FIG9 is a schematic interactive diagram of a method for updating an adaptation layer model provided in this application.

[0029] FIG10 is a schematic interaction diagram of another channel state information (CSI) feedback method provided in the present application.

[0030] FIG11 is a schematic structural diagram of another CSI autoencoder provided in the present application.

[0031] FIG12 is a schematic block diagram of a terminal device provided according to an embodiment of the present application.

[0032] FIG13 is a schematic block diagram of a network device provided according to an embodiment of the present application.

[0033] FIG14 is a schematic block diagram of another terminal device provided according to an embodiment of the present application.

[0034] Figure 15 is a schematic block diagram of another network device provided according to an embodiment of the present application.

[0035] FIG16 is a schematic block diagram of a communication device provided according to an embodiment of the present application.

[0036] FIG17 is a schematic block diagram of a chip provided according to an embodiment of the present application.

[0037] Figure 18 is a schematic block diagram of a communication system provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0038] The following will describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of the embodiments. With respect to the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0039] The technical solutions of the embodiments of the present application can be applied to various communication systems, such as: Global System of Mobile communication (GSM) system, Code Division Multiple Access (CDMA) system, Wideband Code Division Multiple Access (WCDMA) system, General Packet Radio Service (GPRS), Long Term Evolution (LTE) system, Advanced Long Term Evolution (LTE-A) system, New Radio (NR) system, NR system evolution system, LTE on unlicensed spectrum (LTE-U) system, NR on unlicensed spectrum (NR-U) system, Non-Terrestrial Networks (NTN) system, Universal Mobile Telecommunication System (UMTS), Wireless Local Area Networks (WLAN), Wireless Fidelity (Wireless Fidelity) system. Fidelity, WiFi), fifth-generation communication (5th-Generation, 5G) system or other communication systems, etc.

[0040] Generally speaking, traditional communication systems support a limited number of connections and are easy to implement. However, with the development of communication technology, mobile communication systems will not only support traditional communications, but will also support, for example, device-to-device (D2D) communication, machine-to-machine (M2M) communication, machine-type communication (MTC), vehicle-to-vehicle (V2V) communication, or vehicle-to-everything (V2X) communication, etc. The embodiments of the present application can also be applied to these communication systems.

[0041] Optionally, the communication system in the embodiment of the present application can be applied to a carrier aggregation (CA) scenario, a dual connectivity (DC) scenario, or a standalone (SA) networking scenario.

[0042] Optionally, the communication system in the embodiment of the present application can be applied to an unlicensed spectrum, where the unlicensed spectrum can also be considered as a shared spectrum; or, the communication system in the embodiment of the present application can also be applied to an authorized spectrum, where the authorized spectrum can also be considered as an unshared spectrum.

[0043] The embodiments of the present application describe various embodiments in conjunction with network devices and terminal devices, wherein the terminal device may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device, etc.

[0044] The terminal device can be a station (STA) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA) device, a handheld device with wireless communication capabilities, a computing device or other processing device connected to a wireless modem, a vehicle-mounted device, a wearable device, a terminal device in a next-generation communication system such as an NR network, or a terminal device in a future evolved Public Land Mobile Network (PLMN) network, etc.

[0045] In an embodiment of the present application, the terminal device can be deployed on land, including indoors or outdoors, handheld, wearable or vehicle-mounted; it can also be deployed on the water surface (such as ships, etc.); it can also be deployed in the air (such as airplanes, balloons and satellites, etc.).

[0046] In an embodiment of the present application, the terminal device may be a mobile phone, a tablet computer, a computer with wireless transceiver function, a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal device in industrial control, a wireless terminal device in self-driving, a wireless terminal device in remote medical, a wireless terminal device in a smart grid, a wireless terminal device in transportation safety, a wireless terminal device in a smart city, or a wireless terminal device in a smart home, etc.

[0047] As an example and not a limitation, in the embodiment of the present application, the terminal device may also be a wearable device. Wearable devices may also be called wearable smart devices, which are a general term for wearable devices that are intelligently designed and developed using wearable technology for daily wear, such as glasses, gloves, watches, clothing, and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are fully functional, large in size, and can achieve complete or partial functions without relying on smartphones, such as smart watches or smart glasses, as well as those that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0048] In an embodiment of the present application, the network device may be a device for communicating with a mobile device. The network device may be an access point (AP) in WLAN, a base station (BTS) in GSM or CDMA, a base station (NodeB, NB) in WCDMA, an evolved base station (eNB or eNodeB) in LTE, or a relay station or access point, or a vehicle-mounted device, a wearable device, and a network device (gNB) in an NR network, or a network device in a future evolved PLMN network or a network device in an NTN network, etc.

[0049] As an example and not a limitation, in an embodiment of the present application, the network device may have a mobile feature, for example, the network device may be a mobile device. Alternatively, the network device may be a satellite or a balloon station. For example, the satellite may be a low earth orbit (LEO) satellite, a medium earth orbit (MEO) satellite, a geostationary earth orbit (GEO) satellite, a high elliptical orbit (HEO) satellite, etc. Optionally, the network device may also be a base station set up in a location such as land or water.

[0050] In an embodiment of the present application, the network device can provide services for a cell, and the terminal device communicates with the network device through the transmission resources used by the cell (for example, frequency domain resources, or spectrum resources). The cell can be a cell corresponding to the network device (for example, a base station). The cell can belong to a macro base station or a base station corresponding to a small cell. The small cells here may include: metro cells, micro cells, pico cells, femto cells, etc. These small cells have the characteristics of small coverage and low transmission power, and are suitable for providing high-speed data transmission services.

[0051] For example, a communication system 100 used in an embodiment of the present application is shown in FIG1 . The communication system 100 may include a network device 110, which may be a device that communicates with a terminal device 120 (or a communication terminal or terminal). The network device 110 may provide communication coverage for a specific geographic area and may communicate with terminal devices within the coverage area.

[0052] FIG1 exemplarily shows a network device and two terminal devices. Optionally, the communication system 100 may include multiple network devices and each network device may include other numbers of terminal devices within its coverage area, which is not limited in this embodiment of the present application.

[0053] Optionally, the communication system 100 may further include other network entities such as a network controller and a mobility management entity, which is not limited in the embodiment of the present application.

[0054] It should be understood that in the embodiments of the present application, a device having a communication function in a network / system may be referred to as a communication device. Taking the communication system 100 shown in FIG1 as an example, the communication device may include a network device 110 and a terminal device 120 having a communication function. The network device 110 and the terminal device 120 may be the specific devices described above and will not be described in detail here. The communication device may also include other devices in the communication system 100, such as a network controller, a mobility management entity, and other network entities, which are not limited in the embodiments of the present application.

[0055] It should be understood that the terms "system" and "network" are often used interchangeably herein. The term "and / or" is simply a description of an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " generally indicates that the related objects are in an "or" relationship.

[0056] It should be understood that the "indication" mentioned in the embodiments of this application can be a direct indication, an indirect indication, or an indication of an association. For example, "A indicates B" can mean that A directly indicates B, for example, B can be obtained through A; it can also mean that A indirectly indicates B, for example, A indicates C, and B can be obtained through C; it can also mean that there is an association between A and B.

[0057] In the description of the embodiments of the present application, the term "corresponding" may indicate a direct or indirect correspondence between the two, or an association relationship between the two, or a relationship between indication and being indicated, configuration and being configured, etc.

[0058] In the embodiments of the present application, "pre-defined" may be implemented by pre-storing corresponding codes, tables, or other methods that can be used to indicate relevant information in a device (e.g., a terminal device or a network device). The present application does not limit the specific implementation method. For example, pre-defined may refer to information defined in a protocol.

[0059] In the embodiments of the present application, the "protocol" may refer to a standard protocol in the communication field, for example, it may include an LTE protocol, an NR protocol, and related protocols used in future communication systems, and this application does not limit this.

[0060] To facilitate a better understanding of the embodiments of the present application, the neural network and deep learning related to the present application are explained.

[0061] A neural network is a computational model consisting of multiple interconnected neuron nodes, where the connections between nodes represent weighted values ​​from input signals to output signals, called weights. Each node performs a weighted summation (SUM) on different input signals and outputs them through a specific activation function (f). Figure 2 is a schematic diagram of a neuron structure, where a1, a2, …, an represent input signals, w1, w2, …, wn represent weights, f represents the activation function, and t represents the output.

[0062] A simple neural network, shown in Figure 3, consists of an input layer, hidden layers, and an output layer. By using different connections, weights, and activation functions among multiple neurons, different outputs can be generated, thereby fitting the mapping relationship from input to output. Each node in the previous level is connected to all nodes in the next level. This neural network is a fully connected neural network, also known as a deep neural network (DNN).

[0063] Deep learning utilizes deep neural networks with multiple hidden layers, significantly improving the network's ability to learn features and fitting complex, nonlinear mappings from input to output. Consequently, it has found widespread application in speech and image processing. In addition to deep neural networks, deep learning also includes other commonly used basic structures such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) for different tasks.

[0064] The basic structure of a convolutional neural network includes an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, and an output layer, as shown in Figure 4. Each neuron in the convolution kernel of the convolutional layer is locally connected to its input, and the introduction of the pooling layer extracts the local maximum or average features of a certain layer, effectively reducing the network parameters and mining local features, enabling the convolutional neural network to converge quickly and achieve excellent performance.

[0065] RNNs are neural networks that model sequential data and have achieved remarkable success in natural language processing applications such as machine translation and speech recognition. Specifically, the network memorizes information from past moments and uses it in the calculation of current outputs. This means that nodes in the hidden layers are no longer disconnected but connected, and the input to a hidden layer includes not only the input layer but also the output of the previous hidden layer. Common RNN structures include long short-term memory (LSTM) and gated recurrent unit (GRU). Figure 5 shows a basic LSTM cell structure, which can include a tanh activation function. Unlike RNNs, which only consider the most recent state, the LSTM cell state determines which states should be retained and which should be forgotten, addressing the shortcomings of traditional RNNs in long-term memory.

[0066] To facilitate a better understanding of the embodiments of the present application, a codebook-based channel state information (CSI) feedback scheme related to the present application is described.

[0067] In the NR system, codebook-based eigenvector feedback is usually used to enable the base station to obtain downlink CSI. Specifically, the base station sends a downlink CSI-RS to the user, and the user uses the CSI-RS to estimate the CSI of the downlink channel, and performs eigenvalue decomposition on the estimated downlink channel to obtain the eigenvector corresponding to the downlink channel. Furthermore, the NR system provides two codebook design schemes: Type 1 and Type 2. Among them, the Type 1 codebook is used for CSI feedback of conventional accuracy, mainly for CSI feedback in single-user multiple input multiple output (SU-MIMO) scenarios, and the Type 2 codebook is mainly used to improve the transmission performance of multiple user multiple input multiple output (MU-MIMO).

[0068] Both Type 1 and Type 2 codebooks employ two-level codebook feedback, where W = W1W2. W1 describes the wideband, long-term characteristics of the channel, while W2 describes the subband, short-term characteristics of the channel. Generally, the Type 2 codebook utilizes a higher number of feedback bits, resulting in higher-precision CSI feedback performance.

[0069] To facilitate a better understanding of the embodiments of the present application, the CSI feedback solution based on artificial intelligence (AI) related to the present application is described.

[0070] Given the tremendous success of AI technology, especially deep learning, in computer vision and natural language processing, the communications field has begun to explore the use of deep learning to address technical challenges that are difficult to solve with traditional communications methods, such as deep learning. The neural network architecture commonly used in deep learning is nonlinear and data-driven. It can extract features from actual channel matrix data and restore the channel matrix information compressed and fed back by the terminal side as much as possible on the base station side. This ensures the restoration of channel information while also providing the possibility of reducing CSI feedback overhead on the terminal side. Deep learning-based CSI feedback treats channel information as an image to be compressed, uses a deep learning autoencoder to compress and feed back the channel information, and reconstructs the compressed channel image at the transmitter, which can preserve the channel information to a greater extent.

[0071] The structure of a typical channel information feedback system (or AI-based CSI autoencoder) is shown in Figure 6. The entire feedback system is divided into an encoder and a decoder, deployed at the transmitter and receiver, respectively. After the transmitter obtains the channel information through channel estimation, the encoder's neural network compresses and encodes the channel information matrix and feeds the compressed bit stream back to the receiver via an air interface feedback link. The receiver uses the decoder to recover the channel information based on the feedback bit stream to obtain complete feedback channel information. The encoder and decoder backbone networks shown in Figure 6 can adopt a DNN composed of multiple layers of fully connected layers, a CNN composed of multiple layers of convolutional layers, or an RNN with structures such as LSTM and GRU. Various neural network architectures such as residual and self-attention mechanisms can also be used to improve the performance of the encoder and decoder. While the encoding and decoding framework remains unchanged, the backbone network model structure of the encoder and decoder can be flexibly designed.

[0072] The following describes the technical issues existing in the above two CSI feedback solutions.

[0073] For the codebook-based CSI feedback solution, since the codebook is pre-set, the correlation between different antenna ports and subbands is not effectively utilized, resulting in large feedback overhead and poor feedback performance.

[0074] AI-based CSI feedback schemes can extract the correlation of CSI feature vectors in the time and frequency domains, thereby achieving good feedback performance with low feedback overhead. However, the encoder and decoder used in this feedback scheme both employ a neural network architecture and require training on large-scale datasets. Therefore, during deployment, the actual number of configured antenna ports and subbands must be consistent with the dataset used in the training model. If the actual number of configured antenna ports and subbands does not match the dataset used in the training model, the model cannot be used for CSI feedback, thus reducing the flexibility of the AI-based CSI feedback scheme. For example, when constructing the training set based on a dataset configured with 32 transmit antenna ports and 12 subbands, each sample in the dataset is 32*12*2=768 (where 2 represents the split between the real and imaginary parts). The CSI encoder and decoder models trained on this dataset have inputs and outputs of vectors or matrices of length 768, and cannot effectively adapt to feature vector inputs configured with other numbers of transmit antenna ports and subbands.

[0075] For example, in actual deployments, if a configuration of 16 transmit antenna ports and 12 subbands is used, the input vector is 384 in length. The spatial and frequency domain features it contains are inconsistent with those in the training set, so the model obtained from the training set is not applicable to this configuration. In actual deployments, different CSI encoder and decoder models are provided for each antenna port and subband configuration. This requires saving too many different models, resulting in excessive model storage overhead. Alternatively, when downloading models over the air, the model download overhead is excessive, which is not an effective solution. Therefore, designing a CSI autoencoder for variable-dimensional input is an urgent problem that needs to be solved.

[0076] Therefore, how to solve the problem of flexible adaptation of AI-based CSI autoencoders to different numbers of antenna ports and subband configurations during actual deployment is of great significance to the research on AI-based CSI feedback.

[0077] If, in actual deployment, a corresponding model is trained for each antenna port and subband configuration, the terminal device needs to maintain too many types of models, resulting in high storage overhead. Alternatively, if the model needs to be downloaded from the air interface, the air interface overhead will be high.

[0078] To facilitate understanding of the technical solutions of the embodiments of the present application, the technical solutions of the present application are described in detail below through specific embodiments. The above related technologies can be combined arbitrarily with the technical solutions of the embodiments of the present application as optional solutions, and all of them fall within the scope of protection of the embodiments of the present application. The embodiments of the present application include at least part of the following contents.

[0079] FIG7 is a schematic interaction diagram of a method 1000 for channel state information (CSI) feedback according to an embodiment of the present application. As shown in FIG7 , the method 1000 includes at least part of the following:

[0080] S1010: The terminal device processes the first input information to obtain target input information;

[0081] S1020, the terminal device encodes the target input information based on the encoder to obtain a target bit stream;

[0082] S1030, the terminal device sends the target bit stream to the network device;

[0083] S1040: The network device decodes the target bit stream based on the decoder to obtain first output information;

[0084] S1050: The network device processes the first output information to obtain target output information.

[0085] In some embodiments, the first input information is CSI data obtained based on the target configuration. This CSI data may be full channel information obtained by performing channel estimation on a reference signal, i.e., unquantized channel information. Alternatively, it may be eigenvector information obtained based on the full channel information. This application does not limit the specific form of this CSI data.

[0086] In some embodiments, the dimension of the target input information is the same as the dimension of the CSI data obtained based on the reference configuration.

[0087] In some embodiments, the dimension of the first output information is the same as the dimension of the CSI data obtained based on the reference configuration.

[0088] In some embodiments, the dimension of the target output information is the same as the dimension of the CSI data obtained based on the target configuration.

[0089] In some embodiments, the target configuration may be the current configuration of the terminal device, ie, the configuration indicated by the network device.

[0090] In some embodiments, the CSI data obtained under different configurations has different dimensions, and CSI data of different dimensions corresponds to different feedback overheads. In other words, different configurations can correspond to different feedback overheads. For example, when the number of antenna ports or subbands decreases, the feedback overhead also decreases accordingly. In other words, for input CSI data of different dimensions, the length of the generated target bitstream is different.

[0091] The configuration in the embodiments of the present application may include at least one of the following: the number of antenna ports, the subband configuration, or other configurations used when obtaining CSI data, which are not limited in this application. The configuration in the embodiments of the present application is also referred to as physical resource configuration or CSI configuration.

[0092] In some embodiments, the encoder may include a backbone network for extracting time domain, frequency domain, spatial domain and other features within the CSI data, such as extracting the correlation of CSI data in the time domain and frequency domain, which is conducive to reducing feedback overhead and improving feedback performance.

[0093] It should be understood that the present application does not limit the implementation method of the encoder's backbone network. For example, it can be implemented using neural network models such as DNN, CNN, RNN, or it can also be implemented using network structures such as residual, branch, and attention networks. The present application is not limited to this.

[0094] In some embodiments, the decoder may include a backbone network for recovering the received target bit stream into CSI data.

[0095] It should be understood that the present application does not limit the implementation method of the decoder's backbone network. For example, it can be implemented using neural network models such as DNN, CNN, RNN, or it can also be implemented using network structures such as residual, branch, and attention networks. The present application is not limited to this.

[0096] In an embodiment of the present application, the backbone network of the encoder is adapted to the dimension of the CSI data under the reference configuration.

[0097] In some embodiments, the encoder includes an adaptation layer, the input dimension of the adaptation layer being the output dimension of a first network, the output dimension of the adaptation layer being determined based on the bitstream length required for feedback in the target configuration, and the first network being the upper-level network of the adaptation layer in the encoder. Optionally, the first network may be a connection layer, or may be another network structure, which is not limited in this application.

[0098] In some embodiments, the decoder includes an adaptation layer, the input dimension of which is determined based on the bitstream length required for feedback in the target configuration, and the output dimension of which is the input dimension of a second network, which is the next level network after the adaptation layer in the decoder. Optionally, the second network can be a connection layer, or can be another network structure, which is not limited in this application.

[0099] In an embodiment of the present application, the encoder is deployed on the terminal device side, and the decoder is deployed on the network device side.

[0100] For ease of distinction and explanation, the adaptation layer in the encoder is referred to as the encoder adaptation layer, and the adaptation layer in the decoder is referred to as the decoder adaptation layer. It should be noted that the adaptation layer in the embodiment of the present application can also be referred to as an adaptation network.

[0101] It should be understood that the present application does not limit the specific implementation of the adaptation layer. For example, it can be implemented through one fully connected layer or multiple fully connected layers.

[0102] In some embodiments, in the network model of the encoder, the previous network of the adaptation layer is the first network. The input dimension of the adaptation layer can be the output dimension of the first network, denoted as M. The output dimension of the adaptation layer can be adjusted according to the bitstream length required for feedback under the target configuration (e.g., the number of floating-point numbers corresponding to the target bitstream length), denoted as N. For example, if the bitstream length required for feedback in the target configuration is 64 bits and 2-bit quantization is used, the output dimension of the adaptation layer is N=128.

[0103] In some embodiments, in the network model of the decoder, the input dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back by the target configuration, denoted as N, and the output dimension of the adaptation layer is the input dimension of the second network, denoted as P. The second network is the next level network of the adaptation layer in the decoder.

[0104] Figure 8 is a network structure diagram of a CSI autoencoder provided in an embodiment of the present application, but the present application is not limited thereto. For example, the network structure before the adaptation layer in the encoder can be a fully connected layer, or other network structures, and the network structure after the adaptation layer in the decoder can be a fully connected layer, or other network structures, or the fully connected layer can be considered as part of the backbone network of the model.

[0105] As shown in Figure 8, the encoder may include a backbone network, a fully connected layer, an adaptation layer, and a quantization layer. The decoder may include a dequantization layer, an adaptation layer, a fully connected layer, and a backbone network.

[0106] Based on the network structure shown in Figure 8, the input dimension of the encoder adaptation layer can be the output dimension M of the encoder's fully connected layer. The output dimension of the encoder adaptation layer can be adjusted according to the bit stream length required for feedback in the target configuration. The input dimension of the decoder adaptation layer can be the output dimension N of the dequantization layer. The output dimension of the decoder adaptation layer is the input dimension P of the decoder's fully connected layer.

[0107] In some embodiments, the reference configuration may be the highest configuration among multiple configurations, such as the configuration with the largest number of antenna ports and / or the largest number of subbands, or a high-dimensional physical resource configuration. In other words, generally, the dimension of the CSI data corresponding to the reference configuration is higher than the dimension of the CSI data corresponding to the target configuration.

[0108] In an embodiment of the present application, before inputting the first input information into the encoder, the terminal device first processes the first input information to obtain target input information having the same dimension as the CSI data under the reference configuration, and further inputs the target input information into the backbone network of the encoder. In this way, for CSI inputs of different dimensions, only the adaptation layers of the encoder and decoder need to be adjusted, without having to adjust the backbone networks of the encoder and decoder, which is beneficial to ensuring the stability of the backbone networks of the encoder and decoder.

[0109] In some embodiments, the dimension of the CSI data under the reference configuration is greater than the dimension of the CSI data corresponding to the target configuration. In this case, the terminal device may perform padding processing on the first input information to obtain the target input information. Optionally, the padding processing here may be zero-padding, which helps prevent the padded data from affecting the characteristic distribution of the original CSI data. Alternatively, other padding methods may be used, such as noise padding, cyclic shift padding, etc., but the present application is not limited thereto.

[0110] In some embodiments, the CSI data may be represented by a real part and an imaginary part. When calculating the dimension of the CSI data, the real part and the imaginary part may be treated as one dimension respectively, or the real part and the imaginary part may be combined into one dimension. This application does not limit this.

[0111] In other embodiments, the CSI data may be represented by amplitude and phase. When calculating the dimension of the CSI data, the amplitude and phase may be used as a dimension respectively, or the amplitude and phase may be combined into a dimension. This application does not limit this.

[0112] The following description takes the calculation of the dimension of the CSI data by taking the real part and the imaginary part (or the amplitude and the phase) of the CSI data as one dimension respectively as an example, but the present application is not limited to this.

[0113] For example, the reference configuration is 32-port 12-subband, that is, the dimension of the CSI data under the reference configuration is 32*12*2 (the 2 here means that the real part and the imaginary part are separated into one dimension, or the amplitude and the phase are separated into one dimension) = 768, and the target configuration is 16-port 12-subband, that is, the dimension of the CSI data under the target configuration is 16*12*2 (the 2 here means that the real part and the imaginary part are separated into one dimension, or the amplitude and the phase are separated into one dimension) = 384. In this case, padding operation is required in the port dimension.

[0114] For another example, the reference configuration is 32-port 12 sub-bands, that is, the dimension of the CSI data under the reference configuration is 32*12*2 (the 2 here means that the real part and the imaginary part are separated into one dimension, or the amplitude and the phase are separated into one dimension) = 768, and the target configuration is 16-port 6 sub-bands, that is, the dimension of the CSI data under the target configuration is 16*6*2 (the 2 here means that the real part and the imaginary part are separated into one dimension, or the amplitude and the phase are separated into one dimension) = 192. In this case, padding operations need to be performed on both the port dimension and the sub-band dimension.

[0115] In some embodiments, after receiving the target bit stream, the network device can decode the target bit stream based on the decoder to obtain first output information, and then process the first output information to obtain target output information, wherein the dimension of the target output information is the same as the dimension of the CSI data obtained based on the target configuration.

[0116] For example, the network device decodes the target bitstream using a decoder to obtain first CSI data. Since the first CSI data includes CSI data obtained by padding performed by the terminal device, the network device accordingly crops the CSI data at the padding position to obtain the target CSI data. In other words, the positions of the CSI data targeted by the padding performed by the terminal device and the cropping performed by the network device match.

[0117] The following describes the encoder and decoder model training methods in conjunction with specific embodiments.

[0118] For the convenience of distinction and explanation, the network model excluding the adaptation layer in the encoder is recorded as the basic encoder model, and the network model excluding the adaptation layer in the decoder is recorded as the basic decoder model.

[0119] In some embodiments, the basic encoder model is used to encode CSI data under a reference configuration, the basic decoder model is used to decode CSI data under a reference configuration, the combination of the basic encoder model and the encoder adaptation layer model can be used to encode CSI data under configurations other than the reference configuration, and the combination of the basic decoder model and the decoder adaptation layer model can be used to decode the target bit stream under configurations other than the reference configuration.

[0120] In some embodiments, the base encoder model and base decoder model can be trained based on a larger-dimensional CSI dataset. This allows the encoder and decoder backbone networks to extract correlations from CSI data across a larger spatial and frequency domain for compression and recovery. When configured with smaller-dimensional CSI data, the encoder's adaptation layer can be used to further extract high-dimensional latent space vector information into lower-dimensional latent space vector information while maintaining the encoder's backbone network. Correspondingly, the decoder's adaptation layer performs the corresponding inverse operation.

[0121] Therefore, in an embodiment of the present application, based on the backbone network of the encoder and decoder, the adaptation layer parameters can be adjusted according to different configurations, and the input information of different dimensions can be padded on the encoder side to obtain the input information of the target dimension, and cropped at the output end of the decoder to obtain the output information of the target dimension. This allows the adaptation layer to flexibly adapt CSI feedback of different dimensions (or under different configurations) without changing the backbone network model of the encoder and decoder.

[0122] In some embodiments, the encoder and decoder models are trained by a network device.

[0123] In some embodiments, the encoder and decoder models may be trained through offline training or online training.

[0124] In some embodiments, the method 1000 further includes:

[0125] The terminal device may report a first CSI dataset and at least one second CSI dataset to the network device. The first CSI dataset corresponds to a reference configuration, and the at least one second CSI dataset corresponds to at least one first configuration, which does not include the reference configuration. That is, the first CSI dataset is obtained based on the reference configuration, and the at least one second CSI dataset is obtained based on another configuration.

[0126] For example, the first CSI dataset corresponds to configuration 0, and the at least one second CSI dataset includes CSI dataset 2-1, CSI dataset 2-2, ..., CSI dataset 2-X, which correspond to configurations 1-1 to 1-X, respectively. That is, CSI dataset 2-i is obtained based on configuration 1-i, where i = 1, 2, ..., X, where X is a positive integer.

[0127] In some embodiments, the base encoder model, base decoder model, encoder adaptation layer model, and decoder adaptation layer model are trained serially. For example, first, training is performed based on a CSI dataset in a reference configuration to obtain a base encoder model and a base decoder model. Adaptation layer models corresponding to other configurations are then further trained based on CSI datasets in other configurations.

[0128] For example, a basic encoder model and a basic decoder model are obtained by training based on the first CSI data set.

[0129] For another example, adaptation layer parameters corresponding to configurations 1-1 through 1-X are obtained by training based on CSI dataset 2-1, CSI dataset 2-2, ..., and CSI dataset 2-X. For example, when the basic encoder model and basic decoder model are fixed, the adaptation layer model is trained based on CSI dataset 2-1 corresponding to configuration 1-1 to obtain the adaptation layer model corresponding to configuration 1-1.

[0130] The specific process of a model training method provided in an embodiment of the present application is explained with reference to specific examples.

[0131] In this example, DataSet-Config1 represents the CSI dataset corresponding to Config1 (i.e., the reference configuration), DataSet-Config2 represents the CSI dataset corresponding to Config2, and DataSet-ConfigX represents the CSI dataset corresponding to ConfigX, where Config1 is the reference configuration and Config2-ConfigX is the non-reference configuration.

[0132] Step S1: Train the network models of the encoder and decoder except the adaptation layer based on DataSet-Config1 until the models converge, and obtain the basic encoder model and basic decoder model, or in other words, the basic model parameters of the encoder and decoder;

[0133] In step S2-SX, based on the basic encoder model and basic decoder model obtained in step S1, the adaptation layers in the encoder and decoder are trained based on DataSet-Config2 to DataSet-ConfigX, respectively. That is, the model parameters other than the adaptation layer in the encoder and decoder are fixed, and retraining is performed to obtain the adaptation layer parameters corresponding to Config2 to ConfigX, respectively.

[0134] For example, in step S2, based on the basic encoder model and the basic decoder model obtained in step S1, the adaptation layer in the encoder and decoder is trained based on DataSet-Config2, that is, the other model parameters in the encoder and decoder except the adaptation layer are fixed, and training is performed to obtain the adaptation layer parameters corresponding to Config2.

[0135] For another example, in step SX, based on the basic encoder model and basic decoder model obtained in step S1, the adaptation layer in the encoder and decoder is trained based on DataSet-ConfigX, that is, the other model parameters in the encoder and decoder except the adaptation layer are fixed, and training is performed to obtain the adaptation layer parameters corresponding to ConfigX.

[0136] In some embodiments, the base encoder model, base decoder model, encoder adaptation layer model, and decoder adaptation layer model are trained using a parallel training approach. That is, the models are alternately trained based on the CSI dataset in the reference configuration and the CSI datasets in other configurations to obtain the base encoder model, base decoder model, encoder adaptation layer model, and decoder adaptation layer model.

[0137] The CSI dataset under the reference configuration is used to train the basic encoder model and the basic decoder model, while the CSI dataset under other configurations is used to train the adaptation layer models corresponding to other configurations.

[0138] The specific process of a model training method provided in an embodiment of the present application is explained with reference to specific examples.

[0139] In this example, DataSet-Config1 represents the CSI dataset corresponding to Config1 (i.e., the reference configuration), DataSet-Config2 represents the CSI dataset corresponding to Config2, and DataSet-ConfigX represents the CSI dataset corresponding to ConfigX, where Config1 is the reference configuration and Config2-ConfigX is the non-reference configuration.

[0140] Step S1: train the network model T1 in the encoder and decoder except the adaptation layer based on DataSet-Config1;

[0141] In step S2-SX, based on the basic encoder model and basic decoder model obtained in step S1, the adaptation layers in the encoder and decoder are trained for rounds T2-TX based on DataSet-Config2 to DataSet-ConfigX respectively.

[0142] For example, in step S2, based on the basic encoder model and the basic decoder model obtained in step S1, the adaptation layers in the encoder and decoder are trained for T2 rounds based on DataSet-Config2.

[0143] For another example, in step SX, based on the basic encoder model and the basic decoder model obtained in step S1, the adaptation layers in the encoder and decoder are trained for a TX round based on DataSet-ConfigX.

[0144] Alternate steps S1 and S2-SX until the model converges. Step S1 can obtain the basic encoder model and basic decoder model, and S2-SX can obtain the adaptation layer parameters corresponding to Config2-ConfigX respectively.

[0145] Optionally, the above training round number parameters T1, T2, ..., TX may be determined by the network device or predefined.

[0146] It should be understood that in the above model training method, before training the model based on DataSet-Config2 to DataSet-ConfigX, it is first necessary to fill the CSI data in DataSet-Config2 to DataSet-ConfigX so that the dimension of the filled CSI data is the same as the dimension of the CSI data in DataSet-Config1.

[0147] Optionally, in other embodiments, the encoder model and decoder model are trained by the terminal device. For example, after obtaining the first CSI dataset and at least one second CSI dataset, the terminal device may perform model training based on the first CSI dataset and the at least one second CSI dataset to obtain the encoder model and decoder model. The specific training method is similar to the training method on the network device side and is not further described here.

[0148] In some embodiments, the serial training method can be applicable to both offline training scenarios and online training scenarios.

[0149] For example, in an online training scenario, when the target configuration of the terminal device is updated to ConfigY, if there is no corresponding CSI dataset on the network device side, a serial training method can be used to first obtain the CSI dataset DataSet-ConfigY corresponding to ConfigY, and then retrain the model based on DataSet-ConfigY to obtain the adaptation layer parameters corresponding to ConfigY.

[0150] In some embodiments, the parallel training approach may be applicable to offline training scenarios.

[0151] In some scenarios, considering that terminal devices may support more ports and larger bandwidth, using an adaptation layer for all smaller-dimensional CSI data to adapt the backbone network of the model trained with a large-dimensional CSI dataset will result in performance loss.

[0152] In view of this, in an embodiment of the present application, the terminal device can adapt the backbone network model in units of configuration groups, which can more flexibly support the CSI feedback method based on the adaptation layer of an embodiment of the present application, while ensuring better CSI feedback performance.

[0153] In some embodiments, multiple configurations can be divided into multiple configuration groups, each of which includes a reference configuration and at least one non-reference configuration. The CSI data corresponding to a non-reference configuration in a configuration group only needs to adapt to the dimensions of the CSI data corresponding to the reference configuration in the configuration group to which it belongs, without having to adapt to the dimensions of the CSI data corresponding to the highest configuration. In other words, the same configuration group can correspond to the same basic encoder model and basic decoder model, while different configuration groups can correspond to different basic encoder models and basic decoder models.

[0154] In some embodiments, multiple configuration groups may be predefined or network device configured.

[0155] In some embodiments, the dimension of the CSI data corresponding to the reference configuration in a configuration group is greater than the dimension of the CSI data corresponding to other configurations in the configuration group, that is, the reference configuration in a configuration group is the configuration with the highest dimension in the configuration group.

[0156] In some embodiments, configurations in different configuration groups do not overlap.

[0157] In some embodiments, the number of configurations included in different configuration groups may be the same or different.

[0158] For example, the multiple configuration groups include G configuration groups, where G is a positive integer greater than 1. For example, G=2, that is, the multiple configuration groups include a low-dimensional configuration group and a high-dimensional configuration group. For the high-dimensional configuration group, the following configurations are included: Config0-1, Config0-2..., Config0-X0, a total of X0 configurations, where Config0-1 is the reference configuration in the high-dimensional configuration group, that is, the maximum number of ports and maximum bandwidth configuration that can be supported in the high-dimensional configuration group (for example, 64 ports, 12 subbands), the CSI data set corresponding to Config0-1 is used to determine the basic encoder model and the basic decoder model, recorded as the Config0-1 model, and the CSI data sets corresponding to Config0-2 to Config0-X0 are used to determine the corresponding adaptation layer model, wherein the adaptation layer models corresponding to Config0-2 to Config0-X0 are all adapted to the Config0-1 model.

[0159] For the low-dimensional configuration group, the following configurations are included: Config1-1, Config1-2…, Config1-X1, a total of X1 configurations, where Config1-1 is the reference configuration of the low-dimensional configuration group, that is, the maximum port and bandwidth configuration supported by the low-dimensional configuration group (for example, 8 ports, 6 subbands). The CSI data set corresponding to Config1-1 is used to determine the basic encoder model and the basic decoder model, recorded as the Config1-1 model, and the CSI data sets corresponding to Config1-2 to Config1-X1 are used to determine the corresponding adaptation layer model. The adaptation layer models corresponding to Config1-2 to Config1-X1 are all adapted to the Config1-1 model.

[0160] When the target configuration is Config0-2, the dimensions of the CSI data corresponding to Config0-2 are first padded to the dimensions of the CSI data corresponding to Config0-1, and then the CSI data is compressed and restored based on the Config0-1 model and the adaptation layer model corresponding to Config0-2. When the target configuration is Config1-2, the dimensions of the CSI data corresponding to Config1-2 are first padded to the dimensions of the CSI data corresponding to Config1-1, and then the CSI data is compressed and restored based on the Config1-1 model and the adaptation layer model corresponding to Config1-2.

[0161] In some embodiments, all configurations within the low-dimensional configuration group do not overlap with all configurations within the high-dimensional configuration group.

[0162] In some embodiments, X0 and X1 may be the same or different.

[0163] In some embodiments, the target configuration belongs to a first configuration group, and the network device may indicate the first configuration group and the target configuration to the terminal device.

[0164] In some implementations, the network device may indicate the target configuration group and the target configuration in a joint indication manner.

[0165] In some embodiments, the network device sends first indication information to the terminal device, where the first indication information is used to indicate the first configuration group and the target configuration.

[0166] In some embodiments, the first indication information is log2(G)+log2(max{X0, X1, ..., X G-1}) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

[0167] Wherein, log2(G) bits are used to indicate the first configuration group, log2(max{X0, X1, ..., X G-1}) bits are used to indicate the target configuration in the first configuration group.

[0168] It should be understood that the embodiments of the present application do not limit the sending method of the first indication information. For example, the first indication information can be sent through any downlink signaling. As an example, it is sent through the Media Access Control Control Element (MAC CE) and the downlink control information (DCI).

[0169] In some implementations, the network device may indicate the first configuration group and the target configuration in a hierarchical manner.

[0170] In some embodiments, the network device sends second indication information and third indication information to the terminal device, where the second indication information is used to indicate the first configuration group, and the third indication information is used to indicate the target configuration.

[0171] Optionally, the second indication information and the third indication information may be sent via the same signaling, or may be sent via different signaling. For example, the second indication information and the third indication information may be carried via different indication fields in the same signaling.

[0172] Optionally, the second indication information may be sent via MAC CE or DCI.

[0173] Optionally, the third indication information may be sent via MAC CE or DCI.

[0174] For example, the second indication information is sent through MAC CE, and the third indication information is sent through DCI.

[0175] For another example, the second indication information is sent via DCI, and the third indication information is sent via MAC CE.

[0176] For another example, the second indication information and the third indication information are both sent via DCI or MAC CE.

[0177] In some embodiments, the second indication information is represented by log2(G) bits, and the third indication information is represented by log2(max{X0, X1, ..., X G-1}) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

[0178] In some embodiments, when updating the adaptation layer model at the configuration group granularity, if the configuration group corresponding to the target configuration changes, and if the terminal device does not store the base encoder model corresponding to the updated configuration group, the terminal device needs to obtain the base encoder model corresponding to the updated configuration group from the network device. Alternatively, the base encoder models corresponding to multiple configuration groups can be obtained from the network device in advance.

[0179] 9 , a method for updating an adaptation layer model according to an embodiment of the present application is described.

[0180] S1001, the network device sends a configuration instruction to the terminal device.

[0181] In some embodiments, when updating the adaptation layer at a configuration group granularity, the configuration indication may be the first indication information, or may be the second indication information and the third indication information. That is, the configuration indication may be used to determine the target configuration group and the target configuration.

[0182] In some embodiments, when the adaptation layer is updated at a configuration granularity, the configuration indication may determine the target configuration.

[0183] S1002: The terminal device sends a first request message to the network device to request an adaptation layer model corresponding to a target configuration.

[0184] S1003: The network device sends the adaptation layer model corresponding to the target configuration to the terminal device.

[0185] Optionally, if the terminal device does not store the adaptation layer model corresponding to the target configuration, the terminal device sends the first request message. If the terminal device stores the adaptation layer model corresponding to the target configuration, the terminal device may not send the first request message. For example, if the terminal device has received a configuration indication of the target configuration in a cell and has the corresponding adaptation layer model stored locally, in this case, S1002 and S1003 may not be executed, and only the adaptation layer model corresponding to the target configuration needs to be selected locally.

[0186] Optionally, if the configuration group corresponding to the target configuration of the terminal device changes, and the basic encoder model corresponding to the configuration group is not stored on the terminal device, the first request message is also used to request the basic encoder model corresponding to the updated configuration group.

[0187] Therefore, in an embodiment of the present application, the design scheme of the CSI autoencoder based on the adaptation layer adapts to CSI data of variable dimensions by designing an adaptation layer. In the case of a configuration update, there is no need to perform a large number of model updates, only the adaptation layer model needs to be updated. This enables the AI-based CSI feedback scheme to adapt to CSI feedback under various physical resource configurations at a relatively low model update cost, thereby improving the flexibility and practicality of the AI-based CSI feedback method in cellular systems.

[0188] In addition, a model update method based on a configuration group is provided. For CSI data with variable dimensions, it is only necessary to adapt the dimensions of the CSI data of the reference configuration within the configuration group, without adapting the CSI data of the maximum dimension. This can more flexibly support the CSI feedback method based on the adaptation layer of the embodiment of the present application, while ensuring better CSI feedback performance.

[0189] FIG10 is a schematic interaction diagram of a method 2000 for channel state information (CSI) feedback according to an embodiment of the present application. As shown in FIG10 , the method 2000 includes at least part of the following:

[0190] S2010, the network device may send fourth indication information to the terminal device, where the fourth indication information is used to indicate whether the encoder configured on the terminal device supports CSI data of multiple dimensions, or in other words, whether it has the ability to adapt to variable-dimensional input.

[0191] It should be understood that the embodiment of the present application does not limit the sending method of the fourth indication information. For example, the first indication information can be sent through any downlink signaling, as an example, through radio resource control (RRC) signaling, MAC CE or DCI.

[0192] In some embodiments, when the terminal device is downloading the encoder model, the network device can indicate to the terminal device whether the encoder supports CSI data of multiple dimensions. Specifically, the main bottleneck of whether the AI ​​model can process CSI data with variable dimensional input is the encoder design on the terminal device side. Since the decoder is deployed on the network device side, its specific implementation method and the model adjustment performed to adapt to CSI data with different dimensional inputs are transparent to the terminal device side. Therefore, whether the encoder on the terminal device side has the ability to adapt to CSI data with variable dimensional input will determine whether the model needs to perform operations such as downloading the adaptation layer when the input dimension changes. However, since the encoder on the terminal device side is often downloaded from the network device side when the terminal device enters the cell, it is not determined by the local implementation of the terminal device, and the CSI autoencoder architecture deployed in each cell may not be consistent. Therefore, in an embodiment of the present application, when the terminal device performs the initial encoder model download, the network device can indicate to the terminal device whether the model has the ability to adapt to variable dimensional input.

[0193] In some embodiments, the fourth indication information may be carried in the model parameters of the encoder, that is, whether the model has the ability to adapt to variable-dimensional inputs may be carried in the model parameters. For example, as a characteristic capability parameter of the AI ​​model, it is sent to the terminal device along with the AI ​​model, stored locally by the terminal device, and retrieved when the terminal device uses the model.

[0194] In some embodiments, the terminal device can encode the first input information based on the first encoder model to obtain a target bit stream, wherein the first encoder model can support CSI data of multiple dimensions, and the first input information is CSI data.

[0195] In some embodiments, the first encoder model is implemented using a fully convolutional neural network, or it can also be implemented through other network structures that are insensitive to the input dimension, which is not limited in this application.

[0196] In some embodiments, the first encoder model does not include fully connected layers.

[0197] Figure 11 is a network structure diagram of a CSI autoencoder provided in an embodiment of the present application, but the present application is not limited to this. For example, the convolutional layer blocks in the encoder and decoder can be a combination of multiple convolutional layers and multiple pooling layers, and can also include a neural network architecture that does not contain trainable parameters such as residuals and branches.

[0198] As shown in Figure 11, the encoder can include a convolutional layer block, a reshaping layer, and a quantization layer. The decoder can also include a dequantization layer, a reshaping layer, and a convolutional layer block. The reshaping layer does not contain any trainable parameters, meaning it does not require model training. The reshaping layer is used to reshape the input to the output dimensions. For example, it can convert a 32x12 dimensional input into a 768x1 output.

[0199] In some embodiments, if the encoder model has the ability to adapt to variable-dimensional input, when the network device changes the port, sub-band and other physical resource configurations of the terminal device, the network device does not need to re-send the adaptation layer model to the terminal device, that is, the terminal device does not need to re-download the adaptation layer model on the air interface, which is beneficial to reducing the air interface overhead.

[0200] In some embodiments, if the encoder model does not have the ability to adapt to variable-dimensional input, for example, using the adaptation layer-based model structure in method 1000, when the network device changes the physical resource configuration of the terminal device, such as the port or subband, if the terminal device does not store the adaptation layer model corresponding to the updated configuration, the adaptation layer model needs to be downloaded over the air interface. For example, the adaptation layer model is updated using the method in the embodiment shown in FIG9 .

[0201] Therefore, in the embodiments of the present application, the terminal device can encode CSI data based on an encoder model that adapts to variable-dimensional input. This eliminates the need for frequent updates of the AI-based CSI autoencoder model as the physical resource configuration of the terminal device changes. This reduces the model update overhead during the actual deployment of the AI ​​model and enables the CSI autoencoder implemented within a single model framework to adapt to CSI data of multiple dimensions as input. This improves the flexibility and generalization performance of AI-based feedback.

[0202] The above text, in combination with Figures 7 to 11, describes in detail the method embodiment of the present application. The following text, in combination with Figures 12 to 18, describes in detail the device embodiment of the present application. It should be understood that the device embodiment and the method embodiment correspond to each other, and similar descriptions can refer to the method embodiment.

[0203] FIG12 shows a schematic block diagram of a terminal device 400 according to an embodiment of the present application. As shown in FIG12 , the terminal device 400 includes:

[0204] a processing unit 410 configured to process first input information to obtain target input information, wherein the first input information is CSI data obtained based on the target configuration, and a dimension of the target input information is the same as a dimension of the CSI data obtained based on the reference configuration;

[0205] The processing unit 410 is further configured to: encode the target input information based on the encoder to obtain a target bit stream;

[0206] The communication unit 420 is configured to send the target bit stream to a network device;

[0207] In which, the encoder includes an adaptation layer, the input dimension of the adaptation layer is the output dimension of the first network, the output dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back according to the target configuration, and the first network is the upper-level network of the adaptation layer in the encoder.

[0208] In some embodiments, the target configuration includes at least one of the following configurations:

[0209] Number of antenna ports and subbands.

[0210] In some embodiments, the dimension of the first input information is smaller than the dimension of the CSI data obtained based on the reference configuration, and the processing unit 410 is further configured to:

[0211] Filling processing is performed on the first input information to obtain the target input information.

[0212] In some embodiments, the communication unit 420 is further configured to:

[0213] A first CSI data set and at least one second CSI data set are sent to the network device, wherein the first CSI data set is obtained based on the reference configuration, the at least one second CSI data set corresponds to at least one first configuration, the at least one first configuration does not include the reference configuration, and each second CSI data set is obtained based on the corresponding first configuration.

[0214] In some embodiments, the first CSI data set is used to train a model in the encoder excluding the adaptation layer and a model in the decoder excluding the adaptation layer;

[0215] The at least one second CSI data set is used to train a model of the adaptation layer in the encoder and a model of the adaptation layer in the decoder;

[0216] Wherein, the decoder is deployed on the network device.

[0217] In some embodiments, the target configuration belongs to a first configuration group, which is one of a plurality of configuration groups, each configuration group including a reference configuration and at least one other configuration.

[0218] In some embodiments, the plurality of configuration groups are predefined or configured by the network device.

[0219] In some embodiments, the communication unit 420 is further configured to:

[0220] First indication information sent by a network device is received, where the first indication information is used to indicate the target configuration in the first configuration group.

[0221] In some embodiments, the first indication information indicates the configuration group to which the target configuration belongs and the target configuration in a joint indication manner.

[0222] In some embodiments, the first indication information is log2(G)+log2(max{X0, X1, ..., X G-1}) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

[0223] In some embodiments, the first indication information is sent via at least one of the following signaling:

[0224] Media Access Control Element MAC CE, Downlink Control Information DCI.

[0225] In some embodiments, the communication unit 420 is further configured to:

[0226] Second indication information and third indication information sent by a network device are received, where the second indication information is used to indicate the configuration group to which the target configuration belongs, and the third indication information is used to indicate index information of the target configuration in the first configuration group.

[0227] In some embodiments, the second indication information is represented by log2(G) bits, and the third indication information is represented by log2(max{X0, X1, ..., X G-1}) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

[0228] In some embodiments, the second indication information is sent via at least one of the following signaling: MAC CE, DCI.

[0229] In some embodiments, the third indication information is sent via at least one of the following signaling: MAC CE, DCI.

[0230] In some embodiments, the communication unit 420 is further configured to:

[0231] Sending a first request message to the network device to request an adaptation layer model corresponding to the target configuration;

[0232] Receive the adaptation layer model corresponding to the target configuration sent by the network device.

[0233] In some embodiments, the communication unit 420 is further configured to:

[0234] In a case where the adaptation layer model corresponding to the target configuration is not saved on the terminal device, the first request message is sent to the network device.

[0235] Alternatively, in some embodiments, the communication unit may be a communication interface or a transceiver, or an input / output interface of a communication chip or a system on chip. The processing unit may be one or more processors.

[0236] It should be understood that the terminal device 400 according to the embodiment of the present application may correspond to the terminal device in the embodiment of the method of the present application, and the above-mentioned and other operations and / or functions of each unit in the terminal device 400 are respectively for realizing the corresponding processes of the terminal device in the method 1000 shown in Figures 7 to 9. For the sake of brevity, they will not be repeated here.

[0237] FIG13 is a schematic block diagram of a network device according to an embodiment of the present application. The network device 500 of FIG13 includes:

[0238] The communication unit 510 is configured to receive a target bit stream sent by a terminal device, wherein the target bit stream is a bit stream that requires feedback under a target configuration;

[0239] a processing unit 520, configured to decode the target bitstream based on a decoder to obtain first output information, wherein a dimension of the first output information is the same as a dimension of CSI data obtained based on a reference configuration; and

[0240] processing the first output information to obtain target output information, wherein a dimension of the target output information is the same as a dimension of CSI data obtained based on the target configuration;

[0241] In which, the decoder includes an adaptation layer, the input dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back by the target configuration, the output dimension of the adaptation layer is the input dimension of the second network, and the second network is the next level network of the adaptation layer in the decoder.

[0242] In some embodiments, the target configuration includes at least one of the following configurations:

[0243] Number of antenna ports and subbands.

[0244] In some embodiments, the dimension of the first output information is smaller than the dimension of the CSI data obtained based on the reference configuration, and the processing unit 520 is further configured to:

[0245] The first output information is clipped to obtain the target output information.

[0246] In some embodiments, the communication unit 510 is further configured to:

[0247] Receive a first CSI data set and at least one second CSI data set sent by the terminal device, wherein the first CSI data set is obtained based on the reference configuration, the at least one second CSI data set corresponds to at least one first configuration, the at least one first configuration does not include the reference configuration, and each second CSI data set is obtained based on the corresponding first configuration.

[0248] In some embodiments, the first CSI data set is used to train a model other than the adaptation layer in the decoder and a model other than the adaptation layer in the encoder;

[0249] The at least one second CSI data set is used to train a model of the adaptation layer in the decoder and a model of the adaptation layer in the encoder;

[0250] Wherein, the encoder is deployed on the terminal device.

[0251] In some embodiments, the processing unit 520 is further configured to: train the models of the encoder and the decoder except the adaptation layer based on the first CSI data set to obtain basic model parameters of the encoder and the decoder.

[0252] In some embodiments, the processing unit 520 is also used to: in the case of basic model parameters of the encoder and the decoder, train the model of the adaptation layer in the encoder and the decoder based on each second CSI data set in the at least one second CSI data set to obtain adaptation layer parameters corresponding to at least one first configuration respectively.

[0253] In some embodiments, the processing unit 520 is further configured to perform the following steps:

[0254] S1: training the models of the encoder and the decoder except the adaptation layer based on a first CSI data set, wherein the first CSI data set is obtained based on the reference configuration;

[0255] S2: Based on the model parameters obtained in S1, train the models of the adaptation layers in the encoder and the decoder in sequence according to each CSI data set in at least one second CSI data set;

[0256] S1 and S2 are executed in a loop until the models of the encoder and the decoder converge.

[0257] In some embodiments, the target configuration belongs to a first configuration group, which is one of a plurality of configuration groups, each configuration group including a reference configuration and at least one other configuration.

[0258] In some embodiments, the plurality of configuration groups are predefined or configured by the network device.

[0259] In some embodiments, the communication unit 510 is further configured to:

[0260] Sending first indication information to the terminal device, where the first indication information is used to indicate the target configuration in the first configuration group.

[0261] In some embodiments, the first indication information indicates the configuration group to which the target configuration belongs and the target configuration in a joint indication manner.

[0262] In some embodiments, the first indication information is log2(G)+log2(max{X0, X1, ..., X G-1}) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

[0263] In some embodiments, the first indication information is sent via at least one of the following signaling:

[0264] Media Access Control Element MAC CE, Downlink Control Information DCI.

[0265] In some embodiments, the communication unit 510 is further configured to:

[0266] Second indication information and third indication information are sent to the terminal device, where the second indication information is used to indicate the configuration group to which the target configuration belongs, and the third indication information is used to indicate index information of the target configuration in the first configuration group.

[0267] In some embodiments, the second indication information is represented by log2(G) bits, and the third indication information is represented by log2(max{X0, X1, ..., X G-1}) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

[0268] In some embodiments, the second indication information is sent via at least one of the following signaling: MAC CE, DCI.

[0269] In some embodiments, the third indication information is sent via at least one of the following signaling: MAC CE, DCI.

[0270] In some embodiments, the communication unit 510 is further configured to:

[0271] receiving a first request message sent by the terminal device, where the first request message is used to request an adaptation layer model corresponding to the target configuration;

[0272] Sending the adaptation layer model corresponding to the target configuration to the terminal device.

[0273] Alternatively, in some embodiments, the communication unit may be a communication interface or a transceiver, or an input / output interface of a communication chip or a system on chip. The processing unit may be one or more processors.

[0274] It should be understood that the network device 500 according to the embodiment of the present application may correspond to the network device in the embodiment of the method of the present application, and the above-mentioned and other operations and / or functions of each unit in the network device 500 are respectively for implementing the corresponding processes of the network device in the method 1000 shown in Figures 7 to 9. For the sake of brevity, they will not be repeated here.

[0275] FIG14 shows a schematic block diagram of a terminal device 600 according to an embodiment of the present application. As shown in FIG14 , the terminal device 600 includes:

[0276] The communication unit 610 is used to receive fourth indication information sent by the network device, where the fourth indication information is used to indicate whether the encoder deployed on the terminal device supports CSI data of multiple dimensions.

[0277] In some embodiments, the fourth indication information is sent through at least one of the following signaling:

[0278] Radio resource control RRC signaling, media access control element MAC CE, downlink control information DCI.

[0279] In some embodiments, the fourth indication information is carried in a model parameter of the encoder.

[0280] In some embodiments, the encoder includes an adaptation layer, the input dimension of the adaptation layer is the output dimension of the first network, the output dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back according to the target configuration, and the first network is the upper-level network of the adaptation layer in the encoder.

[0281] In some embodiments, the communication unit 610 is further configured to:

[0282] In the case where the network device changes the target configuration, a second request message is sent to the network device, where the second request message is used to request an adaptation layer model corresponding to the changed target configuration.

[0283] In some embodiments, the encoder deployed on the terminal device supports CSI data of multiple dimensions, and the encoder is implemented using a fully convolutional neural network.

[0284] Optionally, in some embodiments, the communication unit may be a communication interface or a transceiver, or an input / output interface of a communication chip or a system on chip.

[0285] It should be understood that the terminal device 600 according to the embodiment of the present application may correspond to the terminal device in the embodiment of the method of the present application, and the above-mentioned and other operations and / or functions of each unit in the terminal device 600 are respectively for realizing the corresponding processes of the terminal device in the method 2000 shown in Figures 10 to 11. For the sake of brevity, they will not be repeated here.

[0286] FIG15 is a schematic block diagram of a network device according to an embodiment of the present application. The network device 700 of FIG15 includes:

[0287] The communication unit 710 is used to send fourth indication information to the terminal device, where the fourth indication information is used to indicate whether the encoder deployed on the terminal device supports CSI data of multiple dimensions.

[0288] In some embodiments, the fourth indication information is sent through at least one of the following signaling:

[0289] Radio resource control RRC signaling, media access control element MAC CE, downlink control information DCI.

[0290] In some embodiments, the fourth indication information is carried in a model parameter of the encoder.

[0291] In some embodiments, the encoder includes an adaptation layer, the input dimension of the adaptation layer is the output dimension of the first network, the output dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back according to the target configuration, and the first network is the upper-level network of the adaptation layer in the encoder.

[0292] In some embodiments, the communication unit 710 is further configured to:

[0293] In the case where the network device changes the target configuration, a second request message sent by the network device is received, where the second request message is used to request an adaptation layer model corresponding to the changed target configuration.

[0294] In some embodiments, the encoder deployed on the terminal device supports CSI data of multiple dimensions, and the encoder is implemented using a fully convolutional neural network.

[0295] Optionally, in some embodiments, the communication unit may be a communication interface or a transceiver, or an input / output interface of a communication chip or a system on chip.

[0296] It should be understood that the network device 700 according to the embodiment of the present application may correspond to the network device in the embodiment of the method of the present application, and the above-mentioned and other operations and / or functions of each unit in the network device 700 are respectively for implementing the corresponding processes of the network device in the method 2000 shown in Figures 10 to 11. For the sake of brevity, they will not be repeated here.

[0297] Figure 16 is a schematic structural diagram of a communication device 800 provided in an embodiment of the present application. The communication device 800 shown in Figure 16 includes a processor 810, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0298] Optionally, as shown in FIG16 , the communication device 800 may further include a memory 820. The processor 810 may call and execute a computer program from the memory 820 to implement the method in the embodiment of the present application.

[0299] The memory 820 may be a separate device independent of the processor 810 , or may be integrated into the processor 810 .

[0300] Optionally, as shown in FIG8 , the communication device 800 may further include a transceiver 830 , and the processor 810 may control the transceiver 830 to communicate with other devices, specifically, to send information or data to other devices, or to receive information or data sent by other devices.

[0301] The transceiver 830 may include a transmitter and a receiver. The transceiver 830 may further include an antenna, and the number of antennas may be one or more.

[0302] Optionally, the communication device 800 may specifically be a network device in an embodiment of the present application, and the communication device 800 may implement the corresponding processes implemented by the network device in each method in the embodiment of the present application. For the sake of brevity, they will not be repeated here.

[0303] Optionally, the communication device 800 may specifically be a mobile terminal / terminal device in an embodiment of the present application, and the communication device 800 may implement the corresponding processes implemented by the mobile terminal / terminal device in each method in the embodiment of the present application. For the sake of brevity, they will not be repeated here.

[0304] Figure 17 is a schematic structural diagram of a chip according to an embodiment of the present application. The chip 900 shown in Figure 17 includes a processor 910, which can call and run a computer program from a memory to implement the method according to the embodiment of the present application.

[0305] Optionally, as shown in FIG9 , the chip 900 may further include a memory 920 , wherein the processor 910 may call and execute a computer program from the memory 920 to implement the method in the embodiment of the present application.

[0306] The memory 920 may be a separate device independent of the processor 910 , or may be integrated into the processor 910 .

[0307] Optionally, the chip 900 may further include an input interface 930. The processor 910 may control the input interface 930 to communicate with other devices or chips, and specifically, may obtain information or data sent by other devices or chips.

[0308] Optionally, the chip 900 may further include an output interface 940. The processor 910 may control the output interface 940 to communicate with other devices or chips, and specifically, may output information or data to other devices or chips.

[0309] Optionally, the chip can be applied to the network device in the embodiments of the present application, and the chip can implement the corresponding processes implemented by the network device in each method of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0310] Optionally, the chip can be applied to the mobile terminal / terminal device in the embodiments of the present application, and the chip can implement the corresponding processes implemented by the mobile terminal / terminal device in the various methods of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0311] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0312] FIG18 is a schematic block diagram of a communication system 300 provided in an embodiment of the present application. As shown in FIG18 , the communication system 300 includes a terminal device 310 and a network device 320 .

[0313] Among them, the terminal device 310 can be used to implement the corresponding functions implemented by the terminal device in the above method, and the network device 320 can be used to implement the corresponding functions implemented by the network device in the above method. For the sake of brevity, they will not be repeated here.

[0314] It should be understood that the processor of the embodiments of the present application may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiment can be completed by hardware integrated logic circuits in the processor or software instructions. The above processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0315] It is understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0316] It should be understood that the above-mentioned memories are exemplary but not restrictive. For example, the memories in the embodiments of the present application may also be static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM RAM (DR RAM), etc. In other words, the memories in the embodiments of the present application are intended to include, but are not limited to, these and any other suitable types of memories.

[0317] An embodiment of the present application also provides a computer-readable storage medium for storing a computer program.

[0318] Optionally, the computer-readable storage medium can be applied to the network device in the embodiments of the present application, and the computer program enables the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiments of the present application. For the sake of brevity, they are not repeated here.

[0319] Optionally, the computer-readable storage medium can be applied to the mobile terminal / terminal device in the embodiments of the present application, and the computer program enables the computer to execute the corresponding processes implemented by the mobile terminal / terminal device in the various methods of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0320] An embodiment of the present application also provides a computer program product, including computer program instructions.

[0321] Optionally, the computer program product can be applied to the network device in the embodiments of the present application, and the computer program instructions enable the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiments of the present application. For the sake of brevity, they are not repeated here.

[0322] Optionally, the computer program product can be applied to the mobile terminal / terminal device in the embodiments of the present application, and the computer program instructions enable the computer to execute the corresponding processes implemented by the mobile terminal / terminal device in the various methods of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0323] The embodiment of the present application also provides a computer program.

[0324] Optionally, the computer program can be applied to the network device in the embodiments of the present application. When the computer program runs on a computer, the computer executes the corresponding processes implemented by the network device in the various methods of the embodiments of the present application. For the sake of brevity, they are not described here.

[0325] Optionally, the computer program can be applied to the mobile terminal / terminal device in the embodiments of the present application. When the computer program runs on the computer, the computer executes the corresponding processes implemented by the mobile terminal / terminal device in the various methods of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0326] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0327] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0328] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0329] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0330] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0331] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0332] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for channel state information (CSI) feedback, characterized in that: Applied to terminal equipment, including: Processing the first input information to obtain target input information, wherein the first input information is CSI data obtained based on the target configuration, and a dimension of the target input information is the same as a dimension of the CSI data obtained based on the reference configuration; Encode the target input information based on the encoder to obtain a target bit stream; Sending the target bitstream to a network device; In which, the encoder includes an adaptation layer, the input dimension of the adaptation layer is the output dimension of the first network, the output dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back according to the target configuration, and the first network is the upper-level network of the adaptation layer in the encoder.

2. The method according to claim 1, characterized in that The target configuration includes at least one of the following configurations: Number of antenna ports and subbands.

3. The method according to claim 1 or 2, characterized in that The dimension of the first input information is smaller than the dimension of CSI data obtained based on the reference configuration, and the processing the first input information to obtain the target input information includes: Filling processing is performed on the first input information to obtain the target input information.

4. The method according to any one of claims 1 to 3, characterized in that The method further comprises: A first CSI data set and at least one second CSI data set are sent to the network device, wherein the first CSI data set is obtained based on the reference configuration, the at least one second CSI data set corresponds to at least one first configuration, the at least one first configuration does not include the reference configuration, and each second CSI data set is obtained based on the corresponding first configuration.

5. The method according to claim 4, characterized in that The first CSI data set is used to train a model other than the adaptation layer in the encoder and a model other than the adaptation layer in the decoder; The at least one second CSI data set is used to train a model of the adaptation layer in the encoder and a model of the adaptation layer in the decoder; Wherein, the decoder is deployed on the network device.

6. The method according to any one of claims 1 to 5, characterized in that The target configuration belongs to a first configuration group, which is one of a plurality of configuration groups, each configuration group including a reference configuration and at least one other configuration.

7. The method according to claim 6, characterized in that The plurality of configuration groups are predefined or configured by the network device.

8. The method according to claim 6 or 7, characterized in that The method further comprises: First indication information sent by a network device is received, where the first indication information is used to indicate the target configuration in the first configuration group.

9. The method according to claim 8, characterized in that The first indication information indicates the configuration group to which the target configuration belongs and the target configuration in a joint indication manner.

10. The method according to claim 9, characterized in that The first indication information is log2(G)+log2(max{X0, X1, ..., X G-1 }) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

11. The method according to any one of claims 8 to 10, characterized in that The first indication information is sent via at least one of the following signaling: Media Access Control Element MAC CE, Downlink Control Information DCI.

12. The method according to claim 6 or 7, characterized in that The method further comprises: Second indication information and third indication information sent by a network device are received, where the second indication information is used to indicate the configuration group to which the target configuration belongs, and the third indication information is used to indicate index information of the target configuration in the first configuration group.

13. The method according to claim 12, characterized in that The second indication information is represented by log2(G) bits, and the third indication information is represented by log2(max{X0, X1, ..., X G-1 }) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

14. The method according to claim 12 or 13, characterized in that The second indication information is sent via at least one of the following signalings: MAC CE, DCI.

15. The method according to any one of claims 12 to 14, characterized in that The third indication information is sent via at least one of the following signaling: MAC CE, DCI.

16. The method according to any one of claims 1 to 15, characterized in that The method further comprises: Sending a first request message to the network device, where the first request message is used to request an adaptation layer model corresponding to the target configuration; Receive the adaptation layer model corresponding to the target configuration sent by the network device.

17. The method according to claim 16, characterized in that The sending a first request message to the network device includes: In a case where the adaptation layer model corresponding to the target configuration is not saved on the terminal device, the first request message is sent to the network device.

18. A method for channel state information (CSI) feedback, characterized in that: Applicable to network equipment, including: Receiving a target bit stream sent by a terminal device, wherein the target bit stream is a bit stream that requires feedback under a target configuration; Decoding the target bitstream based on a decoder to obtain first output information, wherein a dimension of the first output information is the same as a dimension of CSI data obtained based on a reference configuration; processing the first output information to obtain target output information, wherein a dimension of the target output information is the same as a dimension of CSI data obtained based on the target configuration; In which, the decoder includes an adaptation layer, the input dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back by the target configuration, the output dimension of the adaptation layer is the input dimension of the second network, and the second network is the next level network of the adaptation layer in the decoder.

19. The method according to claim 18, characterized in that The target configuration includes at least one of the following configurations: Number of antenna ports and subbands.

20. The method according to claim 18 or 19, characterized in that The dimension of the first output information is smaller than the dimension of CSI data obtained based on the reference configuration, and the processing the first output information to obtain target output information includes: The first output information is clipped to obtain the target output information.

21. The method according to any one of claims 18 to 20, characterized in that The method further comprises: Receive a first CSI data set and at least one second CSI data set sent by the terminal device, wherein the first CSI data set is obtained based on the reference configuration, the at least one second CSI data set corresponds to at least one first configuration, the at least one first configuration does not include the reference configuration, and each second CSI data set is obtained based on the corresponding first configuration.

22. The method according to claim 21, characterized in that The first CSI data set is used to train a model other than the adaptation layer in the decoder and a model other than the adaptation layer in the encoder; The at least one second CSI data set is used to train a model of the adaptation layer in the decoder and a model of the adaptation layer in the encoder; Wherein, the encoder is deployed on the terminal device.

23. The method according to claim 21 or 22, characterized in that The method further comprises: Based on the first CSI data set, the models of the encoder and the decoder except the adaptation layer are trained to obtain basic model parameters of the encoder and the decoder.

24. The method according to claim 23, wherein The method further comprises: In the case of basic model parameters of the encoder and the decoder, the models of the adaptation layers in the encoder and the decoder are trained based on each second CSI data set in the at least one second CSI data set to obtain adaptation layer parameters corresponding to at least one first configuration respectively.

25. The method according to claim 21 or 22, characterized in that The method further comprises: S1: training a model other than the adaptation layer in the encoder and the decoder based on a first CSI data set, wherein the first CSI data set is obtained based on the reference configuration; S2: Based on the model parameters obtained in S1, train the models of the adaptation layers in the encoder and the decoder in sequence according to each CSI data set in at least one second CSI data set; S1 and S2 are executed in a loop until the models of the encoder and the decoder converge.

26. The method according to any one of claims 18 to 25, characterized in that The target configuration belongs to a first configuration group, which is one of a plurality of configuration groups, each configuration group including a reference configuration and at least one other configuration.

27. The method according to claim 26, characterized in that The plurality of configuration groups are predefined or configured by the network device.

28. The method according to claim 26 or 27, characterized in that The method further comprises: Sending first indication information to the terminal device, where the first indication information is used to indicate the target configuration in the first configuration group.

29. The method according to claim 28, characterized in that The first indication information indicates the configuration group to which the target configuration belongs and the target configuration in a joint indication manner.

30. The method according to claim 29, wherein The first indication information is log2(G)+log2(max{X0, X1, ..., X G-1 }) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

31. The method according to any one of claims 28 to 30, characterized in that The first indication information is sent via at least one of the following signaling: Media Access Control Element MAC CE, Downlink Control Information DCI.

32. The method according to claim 26 or 27, characterized in that The method further comprises: Second indication information and third indication information are sent to the terminal device, where the second indication information is used to indicate the configuration group to which the target configuration belongs, and the third indication information is used to indicate index information of the target configuration in the first configuration group.

33. The method according to claim 32, characterized in that The second indication information is represented by log2(G) bits, and the third indication information is represented by log2(max{X0, X1, ..., X G-1 }) bits, where G represents the total number of configuration groups, X i Indicates the number of configurations included in configuration group i, i = 0, 1, ..., G-1, and max indicates the maximum value.

34. The method according to claim 32 or 33, characterized in that The second indication information is sent via at least one of the following signalings: MAC CE, DCI.

35. The method according to any one of claims 32 to 34, characterized in that The third indication information is sent via at least one of the following signaling: MAC CE, DCI.

36. The method according to any one of claims 18 to 35, characterized in that The method further comprises: receiving a first request message sent by the terminal device, where the first request message is used to request an adaptation layer model corresponding to the target configuration; Sending the adaptation layer model corresponding to the target configuration to the terminal device.

37. A method for channel state information (CSI) feedback, characterized in that: Applied to terminal equipment, including: Receive fourth indication information sent by the network device, where the fourth indication information is used to indicate whether the encoder deployed on the terminal device supports CSI data of multiple dimensions.

38. The method according to claim 37, wherein The fourth indication information is sent through at least one of the following signaling: Radio resource control RRC signaling, media access control element MAC CE, downlink control information DCI.

39. The method according to claim 37, wherein The fourth indication information is carried in the model parameters of the encoder.

40. The method according to any one of claims 37 to 39, wherein The encoder includes an adaptation layer, the input dimension of the adaptation layer is the output dimension of the first network, the output dimension of the adaptation layer is determined according to the bit stream length required to be fed back according to the target configuration, and the first network is the upper-level network of the adaptation layer in the encoder.

41. The method according to claim 40, wherein The method further comprises: In the case where the network device changes the target configuration, a second request message is sent to the network device, where the second request message is used to request an adaptation layer model corresponding to the changed target configuration.

42. The method according to any one of claims 37 to 39, wherein: The encoder deployed on the terminal device supports CSI data of multiple dimensions, and the encoder is implemented using a fully convolutional neural network.

43. A method for channel state information (CSI) feedback, characterized in that: Applicable to network equipment, including: Send fourth indication information to the terminal device, where the fourth indication information is used to indicate whether the encoder deployed on the terminal device supports CSI data of multiple dimensions.

44. The method according to claim 43, wherein The fourth indication information is sent through at least one of the following signaling: Radio resource control RRC signaling, media access control element MAC CE, downlink control information DCI.

45. The method according to claim 43, wherein The fourth indication information is carried in the model parameters of the encoder.

46. ​​The method according to any one of claims 43 to 45, wherein: The encoder includes an adaptation layer, the input dimension of the adaptation layer is the output dimension of the first network, the output dimension of the adaptation layer is determined according to the bit stream length required to be fed back according to the target configuration, and the first network is the upper-level network of the adaptation layer in the encoder.

47. The method according to claim 46, wherein The method further comprises: In the case where the network device changes the target configuration, a second request message sent by the network device is received, where the second request message is used to request an adaptation layer model corresponding to the changed target configuration.

48. The method according to any one of claims 43 to 45, wherein: The encoder deployed on the terminal device supports CSI data of multiple dimensions, and the encoder is implemented using a fully convolutional neural network.

49. A terminal device, characterized in that: include: a processing unit, configured to process first input information to obtain target input information, wherein the first input information is CSI data obtained based on the target configuration, and a dimension of the target input information is the same as a dimension of the CSI data obtained based on the reference configuration; The processing unit is further configured to: encode the target input information based on the encoder to obtain a target bit stream; A communication unit, configured to send the target bit stream to a network device; In which, the encoder includes an adaptation layer, the input dimension of the adaptation layer is the output dimension of the first network, the output dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back according to the target configuration, and the first network is the upper-level network of the adaptation layer in the encoder.

50. A network device, characterized in that include: A communication unit, configured to receive a target bit stream sent by a terminal device, wherein the target bit stream is a bit stream requiring feedback under a target configuration; a processing unit, configured to decode the target bitstream based on a decoder to obtain first output information, wherein a dimension of the first output information is the same as a dimension of CSI data obtained based on a reference configuration; and processing the first output information to obtain target output information, wherein a dimension of the target output information is the same as a dimension of CSI data obtained based on the target configuration; In which, the decoder includes an adaptation layer, the input dimension of the adaptation layer is determined according to the bit stream length that needs to be fed back by the target configuration, the output dimension of the adaptation layer is the input dimension of the second network, and the second network is the next level network of the adaptation layer in the decoder.

51. A terminal device, characterized in that: include: A communication unit is used to receive fourth indication information sent by a network device, where the fourth indication information is used to indicate whether the encoder deployed on the terminal device supports CSI data of multiple dimensions.

52. A network device, characterized in that: include: A communication unit is used to send fourth indication information to a terminal device, where the fourth indication information is used to indicate whether an encoder deployed on the terminal device supports CSI data of multiple dimensions.

53. A terminal device, characterized in that: include: A processor and a memory, the memory being used to store a computer program, the processor being used to call and run the computer program stored in the memory to perform the method according to any one of claims 1 to 17, or the method according to any one of claims 37 to 42.

54. A network device, characterized in that include: A processor and a memory, the memory being used to store a computer program, the processor being used to call and run the computer program stored in the memory to perform the method according to any one of claims 18 to 36, or the method according to any one of claims 43 to 48.

55. A chip, characterized in that: include: A processor for calling and running a computer program from a memory so that a device equipped with the chip performs the method according to any one of claims 1 to 17, or the method according to any one of claims 18 to 36, or the method according to any one of claims 37 to 42, or the method according to any one of claims 43 to 48.

56. A computer-readable storage medium, characterized in that Used to store a computer program, the computer program causing a computer to execute the method according to any one of claims 1 to 17, or the method according to any one of claims 18 to 36, or the method according to any one of claims 37 to 42, or the method according to any one of claims 43 to 48.

57. A computer program product, characterized in that Comprising computer program instructions which cause a computer to perform the method of any one of claims 1 to 17, or the method of any one of claims 18 to 36, or the method of any one of claims 37 to 42, or the method of any one of claims 43 to 48.

58. A computer program, characterized in that The computer program causes a computer to perform the method of any one of claims 1 to 17, or the method of any one of claims 18 to 36, or the method of any one of claims 37 to 42, or the method of any one of claims 43 to 48.

Images