CSI feedback method, apparatus and device, and readable storage medium

Abstract

The invention discloses a CSI feedback method, device and equipment and a readable storage medium, and belongs to the technical field of communication, and the method comprises the steps that network side equipment carries out training to obtain a target layer in an update coding AI unit; the terminal receives the first data from the network side device; the terminal updates a target layer in the coding AI unit according to the first data to obtain an updated coding AI unit; the terminal performs coding processing on the CSI through the update coding AI unit to obtain CSI feedback information; the terminal sends CSI feedback information to the network side device; wherein the coding AI unit is an AI unit for coding the CSI by the terminal, and the first data comprises information associated with updating a target layer in the coding AI unit.

Description

Technical Field

[0001] The present application belongs to the field of communication technology, and specifically relates to a CSI feedback method, device, equipment and readable storage medium. Background Art

[0002] Artificial Intelligence (AI) technology is applied to the CSI coding process in the channel state information (CSI) feedback process. When the configuration related to the CSI feedback changes, and the changed configuration is a configuration that the AI ​​unit has not considered during the training process, the AI ​​unit used for CSI coding also needs to be updated. The existing methods for updating the unit are: 1) Generally, the complete encoder needs to be re-transmitted to the terminal side (or the decoder needs to be re-transmitted to the network side); 2) The training process needs to be re-performed to update the unit; 3) The relevant unit information (such as the input and output data of the unit, etc.) needs to be re-transmitted to update the unit. Generally speaking, the overhead of re-transmitting the unit is lower than the overhead of re-training the unit, so method 1) has an advantage over method 2) and method 3) in flexible unit updates. However, when the structure of the encoder is more complex or the parameter scale is larger, the overhead of unit transmission is still high. Summary of the invention

[0003] The embodiments of the present application provide a CSI feedback method, apparatus, device and readable storage medium, which can solve the problem of high overhead when updating the AI ​​unit during the CSI feedback process.

[0004] In a first aspect, a CSI feedback method is provided, including:

[0005] The terminal receives first data from the network side device;

[0006] The terminal updates the target layer in the coding artificial intelligence AI unit according to the first data to obtain an updated coding AI unit;

[0007] The terminal encodes the CSI through the update encoding AI unit to obtain CSI feedback information;

[0008] The terminal sends the CSI feedback information to the network side device;

[0009] The coding AI unit is an AI unit for encoding CSI by the terminal, and the first data includes information associated with a target layer in the updated coding AI unit.

[0010] In a second aspect, a CSI feedback method is provided, including:

[0011] The network-side device training obtains the target layer in the updated coding AI unit;

[0012] The network side device sends first data to the terminal;

[0013] The network side device receives CSI feedback information from the terminal;

[0014] Among them, the first data is used by the terminal to update the target layer in the coding AI unit to obtain an updated coding AI unit, and the coding AI unit is an AI unit for the terminal to encode CSI, and the first data contains information associated with the target layer in the updated coding AI unit.

[0015] In a third aspect, a CSI feedback device is provided, including:

[0016] A first receiving module, used for a terminal to receive first data from a network side device;

[0017] An updating module, configured for the terminal to update a target layer in a coding AI unit according to the first data to obtain an updated coding AI unit;

[0018] A first processing module, configured for the terminal to encode the CSI through the update encoding AI unit to obtain CSI feedback information;

[0019] A first sending module, configured for the terminal to send the CSI feedback information to the network side device;

[0020] The coding AI unit is an AI unit for encoding CSI by the terminal, and the first data includes information associated with a target layer in the updated coding AI unit.

[0021] In a fourth aspect, a CSI feedback device is provided, including:

[0022] The training module is used for training network-side devices to obtain the target layer in the updated coding AI unit;

[0023] A second sending module, used for the network side device to send first data to the terminal;

[0024] A second receiving module, the network side device receives CSI feedback information from the terminal;

[0025] Among them, the first data is used by the terminal to update the target layer in the coding AI unit to obtain an updated coding AI unit, and the coding AI unit is an AI unit for the terminal to encode CSI, and the first data contains information associated with the target layer in the updated coding AI unit.

[0026] In a fifth aspect, a terminal is provided, comprising a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the method described in the first aspect are implemented.

[0027] In a sixth aspect, a terminal is provided, comprising a processor and a communication interface, wherein the communication interface is used for the terminal to receive first data from a network side device; the processor is used for the terminal to update the target layer in the coding artificial intelligence AI unit according to the first data to obtain an updated coding AI unit; the terminal encodes CSI through the updated coding AI unit to obtain CSI feedback information; the communication interface is used for the terminal to send the CSI feedback information to the network side device; wherein the coding AI unit is an AI unit for the terminal to encode CSI, and the first data contains information associated with the target layer in the updated coding AI unit.

[0028] In the seventh aspect, a network side device is provided, which includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the method described in the second aspect are implemented.

[0029] In an eighth aspect, a network side device is provided, comprising a processor and a communication interface, wherein the processor is used for the network side device to train and obtain a target layer in an updated coding AI unit; the communication interface is used for the network side device to send first data to a terminal; the communication interface is used for the network side device to receive CSI feedback information from the terminal; wherein the first data is used by the terminal to update the target layer in the coding AI unit to obtain an updated coding AI unit, the coding AI unit being an AI unit for the terminal to encode CSI, and the first data contains information associated with the target layer in the updated coding AI unit.

[0030] In a ninth aspect, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented, or the steps of the method described in the second aspect are implemented.

[0031] In the tenth aspect, a wireless communication system is provided, including: a terminal and a network side device, wherein the terminal can be used to execute the steps of the method described in the first aspect, and the network side device can be used to execute the steps of the method described in the second aspect.

[0032] In the eleventh aspect, a chip is provided, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the method described in the first aspect, or to implement the method described in the second aspect.

[0033] In the twelfth aspect, a computer program / program product is provided, wherein the computer program / program product is stored in a storage medium, and the program / program product is executed by at least one processor to implement the method as described in the first aspect, or to implement the method as described in the second aspect.

[0034] In the embodiment of the present application, the terminal updates the target layer in the coding AI unit according to the data received from the network side, and encodes the CSI by updating the coding AI unit to obtain CSI feedback information and feeds it back to the network side. In this way, when the configuration related to CSI feedback changes and the AI ​​unit needs to be updated, only the target layer in the coding AI unit needs to be updated, so as to achieve the update of the AI ​​unit with lower overhead, thereby improving the flexibility of CSI feedback processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1a is a block diagram of a wireless communication system to which the embodiments of the present application can be applied;

[0036] Figure 1b It is a schematic diagram of the structure of a neural network;

[0037] Figure 1c It is a schematic diagram of the structure of a neuron;

[0038] Figure 2 This is one of the flow charts of the CSI feedback method provided in the embodiment of the present application;

[0039] Figure 3 This is the second flow chart of the CSI feedback method provided in the embodiment of the present application;

[0040] Figure 4a It is one of the schematic diagrams of the application embodiment provided by the embodiment of the present application;

[0041] Figure 4b This is the second schematic diagram of the application embodiment provided by the embodiment of the present application;

[0042] Figure 4c This is the third schematic diagram of the application embodiment provided by the embodiment of the present application;

[0043] Figure 5 This is one of the structural diagrams of the CSI feedback device provided in the embodiment of the present application;

[0044] Figure 6This is the second structural diagram of the CSI feedback device provided in the embodiment of the present application;

[0045] Figure 7 is a schematic diagram of the structure of a communication device provided in an embodiment of the present application;

[0046] Figure 8 is a schematic diagram of the structure of a terminal provided in an embodiment of the present application;

[0047] Fig. 9 This is one of the structural diagrams of the network side device provided in the embodiment of the present application;

[0048] Fig.10 This is the second structural diagram of the network side device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of this application.

[0050] The terms "first", "second", etc. of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way are interchangeable where appropriate, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of one type, and the number of objects is not limited, for example, the first object can be one or more. In addition, "or" in the present application represents at least one of the connected objects. For example, "A or B" covers three schemes, namely, Scheme 1: including A but not including B; Scheme 2: including B but not including A; Scheme 3: including both A and B. The character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0051] The term "indication" in this application can be a direct indication (or explicit indication) or an indirect indication (or implicit indication). A direct indication can be understood as the sender explicitly informing the receiver of specific information, operations to be performed, or request results in the sent indication; an indirect indication can be understood as the receiver determining the corresponding information according to the indication sent by the sender, or making a judgment and determining the operation to be performed or the request result according to the judgment result.

[0052] It is worth noting that the technology described in the embodiments of the present application is not limited to the Long Term Evolution (LTE) / LTE-Advanced (LTE-A) system, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA) or other systems. The terms "system" and "network" in the embodiments of the present application are often used interchangeably, and the described technology can be used for the systems and radio technologies mentioned above as well as for other systems and radio technologies. The following description describes a New Radio (NR) system for illustrative purposes, and NR terminology is used in most of the following descriptions, but these technologies can also be applied to systems other than NR systems, such as the 6th generation (6 th Generation, 6G) communication system.

[0053] Figure 1aA block diagram of a wireless communication system applicable to an embodiment of the present application is shown. The wireless communication system includes a terminal 11 and a network side device 12. Among them, the terminal 11 can be a mobile phone, a tablet computer (Tablet Personal Computer), a laptop computer (Laptop Computer), a notebook computer, a personal digital assistant (Personal Digital Assistant, PDA), a handheld computer, a netbook, an ultra-mobile personal computer (Ultra-mobile Personal Computer, UMPC), a mobile Internet device (Mobile Internet Device, MID), an augmented reality (Augmented Reality, AR), a virtual reality (Virtual Reality, VR) device, a robot, a wearable device (Wearable Device), an aircraft (flight vehicle), a vehicle-mounted device (Vehicle User Equipment, VUE), a ship-mounted device, a pedestrian terminal (Pedestrian User Equipment, PUE), a smart home (home appliances with wireless communication functions, such as refrigerators, televisions, washing machines or furniture, etc.), a game console, a personal computer (Personal Computer, PC), a teller machine or a self-service machine and other terminal side devices. Wearable devices include: smart watches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart bracelets, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, etc. Among them, the vehicle-mounted device can also be called a vehicle-mounted terminal, a vehicle-mounted controller, a vehicle-mounted module, a vehicle-mounted component, a vehicle-mounted chip or a vehicle-mounted unit, etc. It should be noted that the specific type of the terminal 11 is not limited in the embodiment of the present application. The network side device 12 may include an access network device or a core network device, wherein the access network device may also be referred to as a radio access network (Radio Access Network, RAN) device, a radio access network function or a radio access network unit. The access network device may include a base station, a wireless local area network (Wireless Local Area Network, WLAN) access point (Access Point, AP) or a wireless fidelity (Wireless Fidelity, WiFi) node, etc.Among them, the base station may be referred to as a Node B (NB), an evolved Node B (eNB), a next generation Node B (gNB), a New Radio Node B (NR Node B), an access point, a Relay Base Station (RBS), a Serving Base Station (SBS), a Base Transceiver Station (BTS), a radio base station, a radio transceiver, a Basic Service Set (BSS), an Extended Service Set (ESS), a Home Node B (HNB), a Home Evolved Node B, a Transmission Reception Point (TRP) or other appropriate terms in the field. As long as the same technical effect is achieved, the base station is not limited to specific technical terms. It should be noted that in the embodiments of the present application, only the base station in the NR system is used as an example for introduction, and the specific type of the base station is not limited.

[0054] The core network device may include, but is not limited to, at least one of the following: a core network node, a core network function, a mobility management entity (Mobility Management Entity, MME), an access mobility management function (Access and Mobility Management Function, AMF), a session management function (Session Management Function, SMF), a user plane function (User Plane Function, UPF), a policy control function (Policy Control Function, PCF), a policy and charging rules function unit (Policy and Charging Rules Function, PCRF), an edge application service discovery function (Edge Application Server Discovery Function, EASDF), a unified data management (Unified Data Management, UDM), a unified data storage (Unified Data Repository, UDR), a home user server (Home Subscriber Server, HSS), a centralized network configuration (CNC), a network storage function (Network Repository Function, NRF), a network exposure function (Network Exposure Function, NEF), a local NEF (Local NEF, or L-NEF), and a binding support function (Binding Support Function, BSF), application function (Application Function, AF), etc. It should be noted that in the embodiment of the present application, only the core network device in the NR system is introduced as an example, and the specific type of the core network device is not limited.

[0055] In order to better understand the technical solution of this application, some contents are first introduced:

[0056] CSI Compression Based on AI / Machine Learning (ML)

[0057] From information theory, we know that accurate CSI is crucial to channel capacity. Especially for multi-antenna systems, the transmitter can optimize the signal transmission based on CSI to make it more compatible with the channel state. For example, the channel quality indicator (CQI) can be used to select the appropriate modulation and coding scheme (MCS) to achieve link adaptation; the precoding matrix indicator (PMI) can be used to achieve eigen beamforming to maximize the strength of the received signal, or to suppress interference (such as interference between cells, interference between multiple users, etc.). Therefore, since the multi-input multi-output (MIMO) technology was proposed, CSI acquisition has always been a research hotspot.

[0058] Usually, the base station sends a CSI reference signal (CSI-ReferenceSignal) CSI-RS on certain time-frequency resources in a certain slot. The terminal performs channel estimation based on the CSI-RS, calculates the channel information on this time slot, and feeds back the PMI to the base station through the codebook. The base station combines the channel information based on the codebook information fed back by the terminal. Before the next CSI report, the base station uses this to perform data precoding and multi-user scheduling.

[0059] In order to further reduce the CSI feedback overhead, the terminal can change the PMI reported for each subband to reporting PMI according to delay. Since the channels in the delay domain are more concentrated, PMIs with fewer delays can approximately represent the PMIs of all subbands, that is, the delay domain information is compressed before reporting. Similarly, in order to reduce overhead, the base station can pre-code the CSI-RS in advance and send the coded CSI-RS to the terminal. The terminal sees the channel corresponding to the coded CSI-RS. The terminal only needs to select several ports with higher strength from the ports indicated by the network side and report the coefficients corresponding to these ports.

[0060] Furthermore, in order to better compress channel information, neural network or machine learning methods can be used.

[0061] Artificial intelligence has been widely used in various fields. There are many ways to implement AI modules, such as neural networks, decision trees, support vector machines, Bayesian classifiers, etc. This application uses neural networks as an example, but does not limit the specific type of AI modules. A simple neural network structure diagram is shown in Figure 1. Figure 1b As shown:

[0062] Among them, the neural network is composed of neurons, and the schematic diagram of neurons is as follows Figure 1c . Where a1, a2, ... aK are inputs, w is the weight (multiplicative coefficient), b is the bias (additive coefficient), and σ(.) is the activation function. Common activation functions include Sigmoid, tanh, ReLU (Rectified Linear Unit), etc.

[0063] The parameters of the neural network are optimized through optimization algorithms. Optimization algorithms are a type of algorithm that can help us minimize or maximize the objective function (sometimes called loss function). The objective function is often a mathematical combination of model parameters and data. For example, given data X and its corresponding label Y, we build a neural network model f(.). With the model, we can get the predicted output f(x) based on the input x, and we can calculate the difference between the predicted value and the true value (f(x)-Y), which is the loss function. Our goal is to find the right w and b to minimize the value of the above loss function. The smaller the loss value, the closer our model is to the actual situation.

[0064] At present, the common optimization algorithms are basically based on the error back propagation (BP) algorithm. The basic idea of ​​the BP algorithm is that the learning process consists of two processes: the forward propagation of the signal and the back propagation of the error. During the forward propagation, the input sample is transmitted from the input layer, processed by each hidden layer layer by layer, and then transmitted to the output layer. If the actual output of the output layer does not match the expected output, it will enter the error back propagation stage. Error back propagation is to propagate the output error layer by layer through the hidden layer to the input layer in some form, and distribute the error to all units in each layer, so as to obtain the error signal of each layer unit, and this error signal is used as the basis for correcting the weights of each unit. This process of adjusting the weights of each layer of the signal forward propagation and error back propagation is repeated. The process of continuous adjustment of weights is the learning and training process of the network. This process continues until the error of the network output is reduced to an acceptable level, or until the pre-set number of learning times is reached.

[0065] Common optimization algorithms include Gradient Descent, Stochastic Gradient Descent (SGD), mini-batch gradient descent, Momentum, Nesterov (name of the inventor, specifically stochastic gradient descent with momentum), Adaptive Gradient Descent (Adagrad), Adadelta, root mean square prop (RMSprop), Adaptive Moment Estimation (Adam), etc. When the error is back-propagated, these optimization algorithms all obtain the derivative / partial derivative of the current neuron based on the error / loss obtained by the loss function, add the learning rate, the previous gradient / derivative / partial derivative, etc., to obtain the gradient, and pass the gradient to the previous layer.

[0066] Specifically, the channel information is compressed and encoded at the terminal, and the compressed content is decoded at the base station to restore the channel information. In AI-based CSI compression, the decoding network deployed at the base station and the encoding network deployed at the terminal need to be used together to effectively restore the original channel information. Once the decoding network and the encoding network do not match, the accuracy of the entire CSI feedback will be significantly reduced. In order to ensure that the decoding network and the encoding network match each other, we can use several training methods, or so-called training collaboration types. For a detailed introduction to the training collaboration types, please see the corresponding section below.

[0067] After the terminal and the network have deployed the corresponding models respectively, the terminal estimates the CSI-RS, calculates the channel information, obtains the encoding result through the calculated channel information or the original estimated channel information through the encoding network, and sends the encoding result to the base station. The base station receives the encoded result and inputs it into the decoding network to restore the channel information.

[0068] AI Unit / AI Model

[0069] The AI ​​unit / AI model described in this application may also be referred to as an AI unit, an AI model, an ML model, an ML unit, an AI structure, an AI function, an AI feature, a machine learning model, a neural network, a neural network function, a neural network function, etc., or the AI ​​unit / AI model may also refer to a processing unit that can implement specific algorithms, formulas, processing procedures, capabilities, etc. related to AI, or the AI ​​unit / AI model may be a processing method, algorithm, function, module or unit for a specific data set, or the AI ​​unit / AI model may be a processing method, algorithm, function, module or unit running on AI / ML related hardware such as a graphics processor (GPU), a network processing unit (NPU), a tensor processor (TPU), an application-specific integrated circuit (ASIC), etc., and the present invention does not specifically limit this. Optionally, the specific data set includes the input and / or output of the AI ​​unit / AI model.

[0070] Optionally, the identifier of the AI ​​unit / AI model may be an AI model identifier, an AI structure identifier, an AI algorithm identifier, or an identifier of a specific data set associated with the AI ​​unit / AI model, or an identifier of a specific scenario, environment, channel feature, or device related to the AI / ML, or an identifier of a function, feature, capability, or module related to the AI / ML, and the present invention does not specifically limit this.

[0071] Scalability of AI / ML CSI compression models

[0072] The scalability of a model refers to the ability of a single model to adapt to multiple input / output configurations at the same time. In CSI compression, the scalability of the model mainly considers the following indicators: the number of subbands of the (encoder) input, the number of ports of the (encoder) input, and the length of the (encoder) output payload. That is, we hope that a CSI compression model can support as many of the above configuration combinations as possible at the same time (for example, a model can support both 64-bit and 116-bit payloads. When the network needs to change the payload from 64-bit to 116-bit, it does not need to retrain the model but rely on the existing model implementation), thereby reducing the cost of replacing models and\or maintaining multiple models at the same time. Generally speaking, in order to achieve a certain amount of scalability of the CSI compression model, it is necessary to consider multiple situations in the model training stage and combine it with a special model structure to achieve this.

[0073] Theoretically, all possible scalability of CSI compression models can be considered during the model training phase. However, if all possible scalability of CSI compression models are considered during the model training phase, the difficulty of model training will be greatly increased and the model structure will become very complex. Even so, it is impossible to use the existing compression model to process when receiving parameter configurations that are not known during the training phase.

[0074] Training collaboration types for AI / ML CSI compression models

[0075] The CSI compression use case is a typical two-end model use case, that is, the complete CSI compression model needs to be deployed on different network nodes. Currently, most of the cases considered are to deploy the encoder on the UE side and the decoder on the NW side. The (sub) models deployed on multiple nodes need to be paired with each other to work properly. Considering the above characteristics of the two-end model, 3GPP has identified several basic types of training collaboration:

[0076] 1) Joint training at single entity (also called type 1)

[0077] The training framework refers to training a complete encoder-decoder model on a network node (UE or NW or a third-party server node, etc.), and then deploying the corresponding model module to the target node through model transfer and other methods (for example, transferring the encoder part to the UE, or transferring the decoder part to the NW). The node that completes the training can be the UE (or the server on the UE side), the NW (or the server on the NW side), or a third-party server node.

[0078] 2) Joint training at multiple entities (also called type 2)

[0079] The training framework refers to the joint participation of multiple nodes in the training process, and each node independently calculates the forward / backward propagation information required for local model training and updates the model parameters of its own node. Since the training process requires forward / backward propagation of the entire model (including encoder and decoder), the corresponding forward / backward propagation information needs to be transmitted between the nodes participating in the training. After the training is completed, the model no longer needs to be transmitted between the nodes.

[0080] 3) Separate training on multiple nodes (or type 3)

[0081] The training framework refers to first training a model used as a reference on a certain node, then sending the relevant information of the reference model to the target node, and finally the target node trains the model required by the node based on the information, so as to ensure that each node (sub) model can be paired with each other. For example, the NW side first trains a set of complete models of encoders and decoders and determines that the obtained decoder is the decoder actually used in the future, and then sends the relevant information of the encoder corresponding to (the decoder) (generally the input and output data of the encoder) to the UE side, and the UE side trains the encoder used by itself based on the information. This training framework can be further subdivided into two cases: UE-first training and NW-first training. UE-first training refers to first training a complete model on the UE side, then sending the information required for NW training to match the model (generally the input and output data of the model to be trained on the NW side) to the NW side, and finally the NW side trains the required model (generally the decoding model) on this side. In contrast, NW training first means training the complete model on the NW side first, then sending the information required for UE to train a matching model (generally the input and output data of the model to be trained on the UE side) to the UE side, and finally the UE trains the required model (generally the coding model) on its side.

[0082] When CSI compression requires a payload configuration that was not considered in the previous model training, it is often necessary to update the model to meet this requirement. In the existing AI CSI compression training collaboration types, the method of updating the model is: 1) For type1, it is generally necessary to re-transmit the complete encoder to the UE side (or re-transmit the decoder to the NW side); 2) For type2, it is necessary to re-train the process to update the model; 3) For type3, it is necessary to re-transmit the relevant model information (such as the input and output data of the model, etc.) to update the model. Generally speaking, the overhead of re-transmitting the model is lower than the overhead of retraining the model, so type1 has an advantage over type2 and type3 in flexible model updates. However, when the structure of the encoder is more complex or the parameter scale is large, the overhead of model transmission is still high, which brings obvious challenges to model updates.

[0083] The CSI feedback method provided in the embodiment of the present application is described in detail below through some embodiments and their application scenarios in combination with the accompanying drawings.

[0084] See also Figure 2 The embodiment of the present application provides a CSI feedback method, the execution subject of the method is a terminal, and includes:

[0085] Step 201: The terminal receives first data from a network side device;

[0086] Step 202: The terminal updates the target layer in the coding AI unit according to the first data to obtain an updated coding AI unit;

[0087] Step 203: The terminal encodes the CSI by updating the encoding AI unit to obtain CSI feedback information;

[0088] Step 204: The terminal sends CSI feedback information to the network side device;

[0089] The coding AI unit is an AI unit used by the terminal to encode the CSI, and the first data includes information associated with updating the target layer in the coding AI unit.

[0090] In the embodiment of the present application, the terminal updates the target layer in the coding AI unit according to the data received from the network side, and encodes the CSI by updating the coding AI unit to obtain CSI feedback information and feeds it back to the network side. In this way, when the configuration related to CSI feedback changes and the AI ​​unit needs to be updated, only the target layer in the coding AI unit needs to be updated, so as to achieve the update of the AI ​​unit with lower overhead, thereby improving the flexibility of CSI feedback processing.

[0091] It should be noted that the encoding AI unit for encoding CSI in the terminal involved in the embodiments of the present application, and the decoding AI unit for decoding CSI feedback information in the network side device involved below, can be understood as the AI ​​units currently used by the terminal side and the network side. In the CSI feedback process, the terminal side encodes the CSI through the encoding AI unit to achieve CSI compression to obtain CSI feedback information. After the terminal side sends the CSI feedback information to the network side, the network side decodes the CSI feedback information through the decoding AI unit to obtain specific CSI.

[0092] In one possible implementation, the encoding AI unit and the decoding AI unit are pre-trained AI units.

[0093] It can be understood that the pre-training of the above-mentioned encoding AI unit and decoding AI unit can be implemented by adopting several types of training collaboration of the AI / ML CSI compression model described above, so that the AI ​​units on both sides can match.

[0094] It should be noted that, for the sake of convenience of description, in the description of the technical solution of the present application, the encoding AI unit may be referred to as an encoder, and the corresponding decoding AI unit may be referred to as a decoder.

[0095] In a possible implementation manner, the target layer satisfies at least one of the following:

[0096] (1) The output of the target layer is the output of the encoding AI unit;

[0097] (2) The input of the target layer is the output of other parts of the encoding AI unit except the target layer.

[0098] In the examples of this application, see Figure 4a , the structure of the coding AI unit is improved, and the coding AI unit is divided into a target layer and a part other than the target layer, and the target layer is specifically located at the tail of the coding AI unit, that is, the output of the target layer is the output of the coding AI unit, and the input of the target layer is the output of other parts.

[0099] Optionally, the target layer is called an adaptation layer, and the other parts are called fixed layers. When the configuration related to CSI feedback changes and the AI ​​unit needs to be updated, the target layer in the coding AI unit is updated to achieve updating of the AI ​​unit with lower overhead, thereby improving the flexibility of CSI feedback processing.

[0100] In a possible implementation, the target layer is used to determine the payload of the CSI feedback. The specific determination process may be that the target layer output is quantized to obtain the payload, and the length of the payload is related to the length of the target layer output, wherein the target layer output is generally a sequence of floating-point numbers, and the quantization process is specifically a process of mapping the floating-point format sequence to a binary sequence, and finally obtaining the binary sequence, i.e., the payload.

[0101] In an embodiment of the present application, a change in the configuration related to CSI feedback, specifically a change in the payload of the CSI feedback, is used as an example for explanation. Accordingly, the target layer is used to determine the payload of the CSI feedback. In this way, when the payload of the CSI feedback changes and the AI ​​unit needs to be updated, the target layer can be updated to apply the changed payload.

[0102] Specifically, in the technical solution of the present application, the decoder is consistent with the conventional AI CSI compression model, and the encoder is further divided into two parts: a fixed layer and an adaptation layer. The frozen layer generally refers to the part from the input of the encoder to a certain intermediate node, while the adaptation layer refers to the part from the end of the frozen layer to the encoded output. When the model is initially deployed, it is still deployed to each target node using conventional methods (i.e., the three types of training collaboration mentioned in the background part). When the model needs to be updated to change the supported payload, only the adaptation layer is updated, and the frozen layer remains unchanged. Generally speaking, the adaptation layer part will be designed to be relatively simple (for example, only a few layers of MLP structure are used and the parameter scale is small), so the overhead of updating the adaptation layer (compared to updating the entire encoder part) is also lower, thereby obtaining the benefit of reducing the model update overhead.

[0103] Optionally, the structure of the target layer is defined by a protocol or determined by negotiation between the UE and the NW, and may specifically satisfy at least one of the following:

[0104] (1) The number of target layers is 1, 2, or several layers;

[0105] (2) The structure of the target layer is multilayer perceptron (MLP), convolutional neural network (CNN), etc.

[0106] In a possible implementation manner, the first data includes at least one of the following:

[0107] (1) At least part of the model structure information of the target layer; for example, the number of layers included in the model, the number of neurons included in each layer, etc. The embodiments of the present application do not limit the specific content of the model structure information.

[0108] (2) At least part of the model parameters of the target layer; for example, the functions and weights used in the neurons. The embodiments of the present application do not limit the specific content of the model parameters.

[0109] The above (1) and (2) are aimed at the network side directly providing the model structure information or model parameters of the target layer in the updated AI unit to the terminal, and the terminal directly updates the target layer according to these information parameters; specifically, the network side can provide the model structure information and model parameters of the target layer at the same time, or the network side can also only provide the model parameters of the target layer.

[0110] (3) A first training data set, where the first training data set is used for terminal training to obtain a target layer in an updated coding AI unit.

[0111] The above (3) is aimed at providing training data from the network side to the terminal. The training data is specifically based on the relevant training data of the target layer in the updated coding AI unit obtained through training on the network side. The network side provides the training data to the terminal, and the terminal itself trains to obtain the specific parameters of the target layer in the updated coding AI unit.

[0112] In a possible implementation manner, before the terminal receives the first data from the network side device, the method further includes:

[0113] The terminal sends second data to the network side device;

[0114] Among them, the second data is used for network side equipment training to obtain the target layer in the updated coding AI unit.

[0115] In an embodiment of the present application, the target layer in the coding AI unit is updated for training on the network side, and specifically the network side performs training based on the data set reported by the UE side.

[0116] In a possible implementation manner, the second data includes at least one of the following:

[0117] (1) a second training data set, where the second training data set includes an input of a target layer in a coding AI unit and a target output of a decoding AI unit. The decoding AI unit is an AI unit that performs decoding processing on the CSI feedback information on a network side device, and the input of the target layer corresponds to the output of the decoding AI unit in a one-to-one manner;

[0118] It should be noted that the second training data set is a different training data set from the above-mentioned first training data set. On the one hand, the CSI feedback configurations for the two may be different. For example, when the payload supported by CSI feedback needs to be changed from 64 bits to 116 bits, the second training data set pre-provided by the terminal to the network side is a training data set with a payload of 64 bits, and the network side is trained based on a payload of 116 bits during the training process, so that the first training data set provided by the network side to the terminal is a training data set with a payload of 116 bits; on the other hand, the range, dimension, and amount of data contained in the data may also be different. The second training data set provided by the terminal to the network side is the input of the target layer and the output of the decoding AI unit, while the first training data set provided by the network side to the terminal is the input of the target layer and the output of the target layer, and there is no necessary connection between the amount of data contained in the first data set and the amount of data contained in the second data set.

[0119] (2) Encode all model information of the AI ​​unit;

[0120] (3) Encode model information of other parts of the AI ​​unit except the target layer.

[0121] The above (2) and (3) are aimed at the terminal directly providing specific model-related information of the encoding AI unit to the network side.

[0122] In a possible implementation, the method further includes at least one of the following:

[0123] (1) The terminal periodically sends relevant information about the coding AI unit currently used by the terminal to the network side device according to the configuration of the network side device;

[0124] (2) the terminal sends, to the network side device, information related to the coding AI unit currently used by the terminal according to the trigger instruction of the network side device based on the first event;

[0125] (3) The terminal actively sends relevant information of the encoding AI unit currently used by the terminal to the network side device based on the second event;

[0126] (4) The terminal sends the input of the currently used coding AI unit and the output of the currently used coding AI unit to the network side device;

[0127] The first event and the second event are associated with CSI feedback.

[0128] In an embodiment of the present application, the network side needs to monitor whether the encoder model on the terminal side has changed. Once a change occurs, the terminal side needs to re-report the data set or model used to train the adaptation layer on the network side. The specific monitoring method can be as described in (1) to (4) above. In this way, the network side can promptly learn that the encoder model on the terminal side has changed, and perform model training updates based on the data set or model for training the adaptation layer on the network side re-reported by the terminal side, ensuring that the models used by the terminal side and the network side match each other.

[0129] It should be noted that the first event may include at least one of the following:

[0130] (a) The performance of the CSI feedback unit monitored by the network side is lower than the target indicator;

[0131] (b) The network side monitors the change of the wireless environment, and the degree of change exceeds the target indicator or the applicable scope or applicable conditions / scenario of the AI ​​unit currently used for CSI feedback;

[0132] (c) The network side monitors the AI ​​computing power, and the currently available AI computing power or the remaining AI computing power is lower than the usage condition of the AI ​​unit currently used for CSI feedback;

[0133] The second event may include at least one of the following:

[0134] (a) The UE monitors the performance of the CSI feedback unit, which is lower than the target indicator;

[0135] (b) The UE monitors changes in the wireless environment, and the degree of change exceeds the target indicator or the applicable scope or applicable conditions / scenario of the AI ​​unit currently used for CSI feedback;

[0136] (c) The UE monitors the AI ​​computing power, and the currently available AI computing power or the remaining AI computing power is lower than the usage condition of the AI ​​unit currently used for CSI feedback;

[0137] In a possible implementation, the method further includes any one of the following:

[0138] (1) In the terminal capability reporting stage, the terminal sends first information to the network side device based on the instruction of the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit including the target layer;

[0139] (2) During the terminal capability reporting phase, the terminal actively sends first information to the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit containing the target layer.

[0140] See also Figure 3 The embodiment of the present application provides a CSI feedback method, the execution subject of the method is a network side device, including:

[0141] Step 301: The network side device trains to obtain the target layer in the updated coding AI unit;

[0142] Step 302: The network side device sends first data to the terminal;

[0143] Step 303: The network side device receives CSI feedback information from the terminal;

[0144] Among them, the first data is used by the terminal to update the target layer in the coding AI unit to obtain an updated coding AI unit. The coding AI unit is an AI unit for the terminal to encode CSI, and the first data contains information associated with the target layer in the updated coding AI unit.

[0145] It should be noted that the network side solution is the counterpart of the terminal side solution, and the corresponding content can be understood by referring to the corresponding content on the terminal side.

[0146] In the embodiment of the present application, the terminal updates the target layer in the coding AI unit according to the data received from the network side, and encodes the CSI by updating the coding AI unit to obtain CSI feedback information and feeds it back to the network side. In this way, when the configuration related to CSI feedback changes and the AI ​​unit needs to be updated, only the target layer in the coding AI unit needs to be updated, so as to achieve the update of the AI ​​unit with lower overhead, thereby improving the flexibility of CSI feedback processing.

[0147] In a possible implementation manner, the target layer satisfies at least one of the following:

[0148] The output of the target layer is the output of the encoding AI unit;

[0149] The input of the target layer is the output of other parts of the encoding AI unit except the target layer.

[0150] In a possible implementation manner, the target layer is used to determine the payload of the CSI feedback.

[0151] In a possible implementation manner, the first data includes at least one of the following:

[0152] At least part of the model structure information of the target layer;

[0153] at least some of the model parameters of the target layer;

[0154] The first training data set is used for terminal training to obtain a target layer in an updated coding AI unit.

[0155] In a possible implementation, the network side device trains and updates the target layer in the coding AI unit, including:

[0156] The network side device receives second data from the terminal;

[0157] The network side device trains and updates the target layer in the coding AI unit based on the second data.

[0158] In a possible implementation manner, the second data includes at least one of the following:

[0159] A second training data set, where the second training data set includes an input of a target layer in an encoding AI unit and an output of a decoding AI unit. The decoding AI unit is an AI unit that performs decoding processing on CSI feedback information on a network side device. The input of the target layer corresponds to the output of the decoding AI unit in a one-to-one manner.

[0160] Encodes all model information of the AI ​​unit;

[0161] Encodes model information of other parts of the AI ​​unit except the target layer.

[0162] In a possible implementation, the method further includes:

[0163] The network side device determines whether the coding AI unit used by the terminal has changed according to at least one of the following:

[0164] The terminal periodically sends to the network side device information related to the coding AI unit currently used by the terminal according to the configuration of the network side device;

[0165] The terminal sends, to the network side device, information related to the coding AI unit currently used by the terminal according to the trigger indication of the network side device based on the first event;

[0166] The terminal actively sends the relevant information of the encoding AI unit currently used by the terminal to the network side device based on the second event;

[0167] The input of the currently used coding AI unit and the output of the currently used coding AI unit sent by the terminal to the network side device;

[0168] The first event and the second event are associated with CSI feedback.

[0169] In a possible implementation, the method further includes any one of the following:

[0170] In the terminal capability reporting stage, the network side device instructs the terminal to send first information to the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit including the target layer;

[0171] During the terminal capability reporting phase, the network side device receives first information actively sent by the terminal, where the first information is used to indicate whether the terminal supports the coding AI unit containing the target layer.

[0172] The technical solution of the present application is described below in conjunction with specific application embodiments. It should be noted that in the following embodiments, the training collaboration type of the AI / ML CSI compression model is taken as an example of separate (or step-by-step) training on multiple nodes, where Embodiment 1 corresponds to the case of UE-first separate training, and Embodiment 2 corresponds to the case of NW-first separate training.

[0173] Embodiment 1: This embodiment considers the process of adopting the technical solution of this application in the case of UE-first training, and the flow chart is as follows Figure 4b As shown:

[0174] The specific steps are as follows:

[0175] 1. Data collection is performed on the UE side.

[0176] 2. The UE side trains the local encoder and decoder, where the encoder has a freezing layer and an adaptation layer in structure. The specific adaptation layer design can be implemented according to: i) protocol regulations; ii) negotiation between the NW side and the UE side. For example, the MLP layer at the end of the encoder can be set as the adaptation layer.

[0177] 3. The UE side reports the data set used for training the decoder and adaptation layer on the NW side to the NW, where the data set includes a one-to-one correspondence between the input of the adaptation layer and the output of the decoder.

[0178] 4. The NW side trains the adaptation layer and decoder based on the data set reported by the UE (the feedback payload used during training can be different from the payload used by the UE to train the local encoder in step 2, for example Figure 4b After the training is completed, the NW sends the adaptation layer to the UE side. The sending method is model transfer. You can use the method of sending only model parameters or sending the entire model.

[0179] 5. The UE uses the local frozen layer and the adaptation layer received from the NW side to form a complete encoder for CSI compression feedback.

[0180] 6. When the NW side needs to change the payload of CSI feedback, it will be based on the new payload in step 3 (for example, Figure 4b The adaptation layer and the decoder are retrained (116 bits in the middle), and the updated adaptation layer is sent to the UE side in the same way as step 4.

[0181] 7. The UE side receives the updated adaptation layer and combines it with the local frozen layer into a complete encoder, thereby providing CSI feedback according to the updated payload.

[0182] Embodiment 2: This embodiment considers the process of using the technical solution of this application in the case of NW-first training. The flow chart is as follows: Figure 4c As shown:

[0183] The specific steps are:

[0184] 1. Data collection is performed on the NW side.

[0185] 2. The NW side trains the local encoder and decoder, where the encoder has a freezing layer and an adaptation layer in structure. The specific adaptation layer design can be implemented according to: i) protocol regulations; ii) negotiation between the NW side and the UE side. For example, the MLP layer at the end of the encoder can be set as the adaptation layer.

[0186] 3. The NW side sends the first data set to the UE side, which includes the encoder input (also the frozen layer input), frozen layer output, and corresponding decoder output obtained by training in step 2. The above three have a one-to-one correspondence, and the sending method can be to send a triplet of (encoder input, encoder output, decoder output), or to send a certain binary group (such as encoder input, frozen layer output) and then specify the corresponding relationship between the remaining data (such as decoder output) and the aforementioned binary group.

[0187] 4. The UE side trains the frozen layer of the local encoder based on the first data set (actually based on the frozen layer input and frozen layer output in the first data set).

[0188] 5. The UE side reports the second data set to the network side, which includes the frozen layer output and the decoder output obtained by training in step 4. There is a one-to-one correspondence between the above two, and the correspondence can be obtained based on the first data set (for example, the encoder input in the first data set is passed through the frozen layer model in step 4 to obtain the frozen layer output, and then the decoder output corresponding to the encoder input is found according to the first data set). Note that the frozen layer output here is different from the frozen layer output in the first data set. The frozen layer output in this step is obtained based on the frozen layer model trained in step 4.

[0189] 6. The NW side updates the adaptation layer and the decoder according to the second data set (for example, by fine-tuning), and transmits the updated adaptation layer to the UE side.

[0190] 7. The UE side combines the adaptation layer updated in step 6 with the local frozen layer to form a complete encoder for CSI compression feedback.

[0191] 8. When the NW side needs to change the payload of CSI feedback, it will be based on the new payload in step 3 (for example Figure 4b The adaptation layer and the decoder are retrained (116 bits in the middle), and the updated adaptation layer is sent to the UE side in the same way as step 4.

[0192] 9. The UE receives the updated adaptation layer and combines it with the local frozen layer into a complete encoder, thereby providing CSI feedback according to the updated payload.

[0193] The CSI feedback method provided in the embodiment of the present application may be executed by a CSI feedback device. In the embodiment of the present application, the CSI feedback device performing the CSI feedback method is taken as an example to illustrate the CSI feedback device provided in the embodiment of the present application.

[0194] See also Figure 5 The embodiment of the present application provides a CSI feedback device, which can be applied to a terminal, and includes:

[0195] A first receiving module 501 is used for a terminal to receive first data from a network side device;

[0196] An updating module 502, configured for the terminal to update a target layer in a coding AI unit according to the first data to obtain an updated coding AI unit;

[0197] A first processing module 503 is configured for the terminal to encode the CSI through the update encoding AI unit to obtain CSI feedback information;

[0198] A first sending module 504, configured for the terminal to send the CSI feedback information to the network side device;

[0199] The coding AI unit is an AI unit for encoding CSI by the terminal, and the first data includes information associated with a target layer in the updated coding AI unit.

[0200] Optionally, the target layer satisfies at least one of the following:

[0201] The output of the target layer is the output of the encoding AI unit;

[0202] The input of the target layer is the output of other parts of the encoding AI unit except the target layer.

[0203] Optionally, the target layer is used to determine a payload of CSI feedback.

[0204] Optionally, the first data includes at least one of the following:

[0205] At least part of the model structure information of the target layer;

[0206] at least some model parameters of the target layer;

[0207] A first training data set, wherein the first training data set is used for the terminal training to obtain a target layer in the update coding AI unit.

[0208] Optionally, the device further comprises:

[0209] A third sending module, configured to, before the terminal receives the first data from the network side device, send the second data to the network side device by the terminal;

[0210] Among them, the second data is used for training the network side device to obtain the target layer in the updated coding AI unit.

[0211] Optionally, the second data includes at least one of the following:

[0212] A second training data set, wherein the second training data set includes an input of a target layer in the encoding AI unit and an output of a decoding AI unit, wherein the decoding AI unit is an AI unit that performs decoding processing on the CSI feedback information by the network side device, and the input of the target layer corresponds to the output of the decoding AI unit in a one-to-one manner;

[0213] All model information of the encoding AI unit;

[0214] Model information of other parts in the encoding AI unit except the target layer.

[0215] Optionally, the encoding AI unit and the decoding AI unit are pre-trained AI units.

[0216] Optionally, the device further comprises:

[0217] The fourth sending module is used for at least one of the following:

[0218] The terminal periodically sends relevant information of the encoding AI unit currently used by the terminal to the network side device according to the configuration of the network side device;

[0219] The terminal sends, to the network side device, relevant information of the encoding AI unit currently used by the terminal according to the trigger indication of the network side device based on the first event;

[0220] The terminal actively sends, based on the second event, to the network side device, relevant information of the encoding AI unit currently used by the terminal;

[0221] The terminal sends the input of the currently used coding AI unit and the output of the currently used coding AI unit to the network side device;

[0222] The first event and the second event are associated with CSI feedback.

[0223] Optionally, the device further comprises:

[0224] The fifth sending module is used for at least one of the following:

[0225] In the terminal capability reporting stage, the terminal sends first information to the network side device based on the instruction of the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit including the target layer;

[0226] During the terminal capability reporting stage, the terminal actively sends first information to the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit containing the target layer.

[0227] See also Figure 6 The embodiment of the present application provides a CSI feedback device, which can be applied to a network side device, and the device includes:

[0228] Training module 601, used for training the network side device to obtain the target layer in the updated coding AI unit;

[0229] A second sending module 602, configured for the network side device to send first data to the terminal;

[0230] A second receiving module 603, the network side device receives CSI feedback information from the terminal;

[0231] Among them, the first data is used by the terminal to update the target layer in the coding AI unit to obtain an updated coding AI unit, and the coding AI unit is an AI unit for the terminal to encode CSI, and the first data contains information associated with the target layer in the updated coding AI unit.

[0232] Optionally, the target layer satisfies at least one of the following:

[0233] The output of the target layer is the output of the encoding AI unit;

[0234] The input of the target layer is the output of other parts of the encoding AI unit except the target layer.

[0235] Optionally, the target layer is used to determine a payload of CSI feedback.

[0236] Optionally, the first data includes at least one of the following:

[0237] At least part of the model structure information of the target layer;

[0238] at least some model parameters of the target layer;

[0239] A first training data set, wherein the first training data set is used for the terminal training to obtain a target layer in the update coding AI unit.

[0240] Optionally, the training module is specifically used to:

[0241] The network side device receives second data from the terminal;

[0242] The network side device trains and updates the target layer in the coding AI unit based on the second data.

[0243] Optionally, the second data includes at least one of the following:

[0244] A second training data set, wherein the second training data set includes an input of a target layer in the encoding AI unit and an output of a decoding AI unit, wherein the decoding AI unit is an AI unit that performs decoding processing on the CSI feedback information by the network side device, and the input of the target layer corresponds to the output of the decoding AI unit in a one-to-one manner;

[0245] All model information of the encoding AI unit;

[0246] Model information of other parts in the encoding AI unit except the target layer.

[0247] Optionally, the device further comprises:

[0248] A determination module, configured for the network side device to determine whether the coding AI unit used by the terminal has changed according to at least one of the following:

[0249] The terminal periodically sends to the network side device, according to the configuration of the network side device, information related to the coding AI unit currently used by the terminal;

[0250] The terminal sends, to the network side device, information related to the encoding AI unit currently used by the terminal according to the trigger indication of the network side device based on the first event;

[0251] The terminal actively sends the relevant information of the encoding AI unit currently used by the terminal to the network side device based on the second event;

[0252] The input of the currently used coding AI unit and the output of the currently used coding AI unit sent by the terminal to the network side device;

[0253] The first event and the second event are associated with CSI feedback.

[0254] Optionally, the device further comprises

[0255] The second processing module is used for any of the following:

[0256] In the terminal capability reporting stage, the network side device instructs the terminal to send first information to the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit including the target layer;

[0257] During the terminal capability reporting phase, the network side device receives first information actively sent by the terminal, where the first information is used to indicate whether the terminal supports the coding AI unit containing the target layer.

[0258] The CSI feedback device in the embodiment of the present application may be an electronic device, such as an electronic device with an operating system, or a component in an electronic device, such as an integrated circuit or a chip. The electronic device may be a terminal, or may be another device other than a terminal. For example, the terminal may include but is not limited to the types of terminals 11 listed above, and other devices may be servers, network attached storage (NAS), etc., which are not specifically limited in the embodiment of the present application.

[0259] The CSI feedback device provided in the embodiment of the present application can achieve Figures 2 to 4c The various processes implemented by the method embodiment and achieving the same technical effect are not described here to avoid repetition.

[0260] like Figure 7 As shown, the embodiment of the present application further provides a communication device 700, including a processor 701 and a memory 702, and the memory 702 stores a program or instruction that can be run on the processor 701. For example, when the communication device 700 is a terminal, the program or instruction is executed by the processor 701 to implement the various steps of the above method embodiment, and can achieve the same technical effect. When the communication device 700 is a network side device, the program or instruction is executed by the processor 701 to implement the various steps of the above method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0261] The embodiment of the present application also provides a terminal, including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the steps in the above method embodiment. The terminal embodiment corresponds to the above terminal side method embodiment, and each implementation process and implementation method of the above method embodiment can be applied to the terminal embodiment and can achieve the same technical effect. Specifically, Figure 8 A schematic diagram of the hardware structure of a terminal for implementing an embodiment of the present application.

[0262] The terminal 800 includes but is not limited to: a radio frequency unit 801, a network module 802, an audio output unit 803, an input unit 804, a sensor 805, a display unit 806, a user input unit 807, an interface unit 808, a memory 809 and at least some of the components of a processor 810.

[0263] Those skilled in the art will appreciate that the terminal 800 may also include a power source (such as a battery) for supplying power to various components, and the power source may be logically connected to the processor 810 through a power management system, thereby implementing functions such as managing charging, discharging, and power consumption management through the power management system. Figure 8 The terminal structure shown in the figure does not constitute a limitation on the terminal. The terminal may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently, which will not be described in detail here.

[0264] It should be understood that in the embodiment of the present application, the input unit 804 may include a graphics processing unit (GPU) 8041 and a microphone 8042, and the graphics processor 8041 processes the image data of a static picture or video obtained by an image capture device (such as a camera) in a video capture mode or an image capture mode. The display unit 806 may include a display panel 8061, and the display panel 8061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc. The user input unit 807 includes a touch panel 8071 and at least one of other input devices 8072. The touch panel 8071 is also called a touch screen. The touch panel 8071 may include two parts: a touch detection device and a touch controller. Other input devices 8072 may include, but are not limited to, a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, and a joystick, which will not be repeated here.

[0265] In the embodiment of the present application, after receiving downlink data from the network side device, the radio frequency unit 801 can transmit the data to the processor 810 for processing; in addition, the radio frequency unit 801 can send uplink data to the network side device. Generally, the radio frequency unit 801 includes but is not limited to an antenna, an amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, etc.

[0266] The memory 809 can be used to store software programs or instructions and various data. The memory 809 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instruction required for at least one function (such as a sound playback function, an image playback function, etc.), etc. In addition, the memory 809 may include a volatile memory or a non-volatile memory. 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), a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDRSDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synchronous link dynamic random access memory (SLDRAM) and a direct memory bus random access memory (DRRAM). The memory 809 in the embodiment of the present application includes but is not limited to these and any other suitable types of memory.

[0267] The processor 810 may include one or more processing units; optionally, the processor 810 integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to an operating system, a user interface, and application programs, and the modem processor mainly processes wireless communication signals, such as a baseband processor. It is understandable that the modem processor may not be integrated into the processor 810.

[0268] The radio frequency unit 801 is used for the terminal to receive first data from the network side device;

[0269] Processor 810, configured for the terminal to update a target layer in a coding AI unit according to the first data to obtain an updated coding AI unit;

[0270] Processor 810, configured for the terminal to encode the CSI through the update encoding AI unit to obtain CSI feedback information;

[0271] A radio frequency unit 801, configured for the terminal to send the CSI feedback information to the network side device;

[0272] The coding AI unit is an AI unit for encoding CSI by the terminal, and the first data includes information associated with a target layer in the updated coding AI unit.

[0273] Optionally, the target layer satisfies at least one of the following:

[0274] The output of the target layer is the output of the encoding AI unit;

[0275] The input of the target layer is the output of other parts of the encoding AI unit except the target layer.

[0276] Optionally, the target layer is used to determine a payload of CSI feedback.

[0277] Optionally, the first data includes at least one of the following:

[0278] At least part of the model structure information of the target layer;

[0279] at least some model parameters of the target layer;

[0280] A first training data set, wherein the first training data set is used for the terminal training to obtain a target layer in the update coding AI unit.

[0281] Optionally, the device further comprises:

[0282] The radio frequency unit 801 is configured to send second data to the network side device by the terminal before the terminal receives first data from the network side device;

[0283] Among them, the second data is used for training the network side device to obtain the target layer in the updated coding AI unit.

[0284] Optionally, the second data includes at least one of the following:

[0285] A second training data set, wherein the second training data set includes an input of a target layer in the encoding AI unit and an output of a decoding AI unit, wherein the decoding AI unit is an AI unit that performs decoding processing on the CSI feedback information by the network side device, and the input of the target layer corresponds to the output of the decoding AI unit in a one-to-one manner;

[0286] All model information of the encoding AI unit;

[0287] Model information of other parts in the encoding AI unit except the target layer.

[0288] Optionally, the encoding AI unit and the decoding AI unit are pre-trained AI units.

[0289] Optionally, the device further comprises:

[0290] The radio frequency unit 801 is configured to:

[0291] The terminal periodically sends relevant information of the encoding AI unit currently used by the terminal to the network side device according to the configuration of the network side device;

[0292] The terminal sends, to the network side device, relevant information of the encoding AI unit currently used by the terminal according to the trigger indication of the network side device based on the first event;

[0293] The terminal actively sends, based on the second event, to the network side device, relevant information of the encoding AI unit currently used by the terminal;

[0294] The terminal sends the input of the currently used coding AI unit and the output of the currently used coding AI unit to the network side device;

[0295] The first event and the second event are associated with CSI feedback.

[0296] Optionally, the device further comprises:

[0297] The radio frequency unit 801 is configured to:

[0298] In the terminal capability reporting stage, the terminal sends first information to the network side device based on the instruction of the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit including the target layer;

[0299] During the terminal capability reporting stage, the terminal actively sends first information to the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit containing the target layer.

[0300] It can be understood that the implementation process of each implementation method mentioned in this embodiment can refer to the relevant description of the method embodiment and achieve the same or corresponding technical effect. To avoid repetition, it will not be repeated here.

[0301] The embodiment of the present application also provides a network side device, including a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the steps of the method embodiment. The network side device embodiment corresponds to the above-mentioned network side device method embodiment, and each implementation process and implementation method of the above-mentioned method embodiment can be applied to the network side device embodiment and can achieve the same technical effect.

[0302] Specifically, the embodiment of the present application also provides a network side device. Fig. 9 As shown, the network side device 900 includes: an antenna 91, a radio frequency device 92, a baseband device 93, a processor 94 and a memory 95. The antenna 91 is connected to the radio frequency device 92. In the uplink direction, the radio frequency device 92 receives information through the antenna 91 and sends the received information to the baseband device 93 for processing. In the downlink direction, the baseband device 93 processes the information to be sent and sends it to the radio frequency device 92. The radio frequency device 92 processes the received information and sends it out through the antenna 91.

[0303] The method executed by the network-side device in the above embodiment may be implemented in the baseband device 93, which includes a baseband processor.

[0304] The baseband device 93 may include, for example, at least one baseband board on which a plurality of chips are arranged. Fig. 9 As shown, one of the chips is, for example, a baseband processor, which is connected to the memory 95 through a bus interface to call the program in the memory 95 to execute the network device operations shown in the above method embodiment.

[0305] The network side device may further include a network interface 96, which is, for example, a Common Public Radio Interface (CPRI).

[0306] Specifically, the network side device 900 of the embodiment of the present application further includes: instructions or programs stored in the memory 95 and executable on the processor 94, and the processor 94 calls the instructions or programs in the memory 95 to execute. Figure 6 The methods executed by the modules shown achieve the same technical effects, and therefore will not be described here in detail to avoid repetition.

[0308] Specifically, the embodiment of the present application also provides a network side device. Fig.10 As shown, the network side device 1000 includes: a processor 1001, a network interface 1002 and a memory 1003. The network interface 1002 is, for example, a common public radio interface (CPRI).

[0309] Specifically, the network side device 1000 of the embodiment of the present application further includes: an instruction or program stored in the memory 1003 and executable on the processor 1001, and the processor 1001 calls the instruction or program in the memory 1003 to execute Figure 6 The methods executed by the modules shown achieve the same technical effects, and therefore will not be described here in detail to avoid repetition.

[0310] An embodiment of the present application further provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the above-mentioned CSI feedback method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0311] The processor is the processor in the terminal described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk. In some examples, the readable storage medium may be a non-transient readable storage medium.

[0312] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement each process of the above-mentioned CSI feedback method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0313] 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.

[0314] An embodiment of the present application further provides a computer program / program product, which is stored in a storage medium, and is executed by at least one processor to implement the various processes of the above-mentioned CSI feedback method embodiment, and can achieve the same technical effect. To avoid repetition, it is not repeated here.

[0315] An embodiment of the present application also provides a wireless communication system, including: a terminal and a network side device, wherein the terminal can be used to execute the steps of the terminal side method as described above, and the network side device can be used to execute the steps of the network side method as described above.

[0316] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises one..." does not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the method and device in the embodiment of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0317] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of a computer software product plus a necessary general hardware platform, and of course, can also be implemented by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, disk, CD, etc.), including several instructions to enable a terminal or a network-side device to execute the methods described in each embodiment of the present application.

[0318] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms of implementation methods without departing from the purpose of the present application and the scope of protection of the claims, and these implementation methods are all within the protection of the present application.

Claims

1. A channel state information (CSI) feedback method, characterized in that: include: The terminal receives first data from the network side device; The terminal updates the target layer in the coding artificial intelligence AI unit according to the first data to obtain an updated coding AI unit; The terminal encodes the CSI through the update encoding AI unit to obtain CSI feedback information; The terminal sends the CSI feedback information to the network side device; The coding AI unit is an AI unit for encoding CSI by the terminal, and the first data includes information associated with a target layer in the updated coding AI unit.

2. The method according to claim 1, characterized in that The target layer satisfies at least one of the following: The output of the target layer is the output of the encoding AI unit; The input of the target layer is the output of other parts of the encoding AI unit except the target layer.

3. The method according to claim 1, characterized in that The target layer is used to determine the effective load payload of the CSI feedback.

4. The method according to claim 1, characterized in that: The first data includes at least one of the following: At least part of the model structure information of the target layer; at least some model parameters of the target layer; A first training data set, wherein the first training data set is used for the terminal training to obtain a target layer in the update coding AI unit.

5. The method according to claim 1, characterized in that: Before the terminal receives the first data from the network side device, the method further includes: The terminal sends second data to the network side device; Among them, the second data is used for training the network side device to obtain the target layer in the updated coding AI unit.

6. The method according to claim 5, characterized in that The second data includes at least one of the following: A second training data set, wherein the second training data set includes an input of a target layer in the encoding AI unit and an output of a decoding AI unit, wherein the decoding AI unit is an AI unit that performs decoding processing on the CSI feedback information by the network side device, and the input of the target layer corresponds to the output of the decoding AI unit in a one-to-one manner; All model information of the encoding AI unit; Model information of other parts in the encoding AI unit except the target layer.

7. The method according to claim 6, characterized in that The encoding AI unit and the decoding AI unit are pre-trained AI units.

8. The method according to claim 1, characterized in that The method further comprises at least one of the following: The terminal periodically sends relevant information of the encoding AI unit currently used by the terminal to the network side device according to the configuration of the network side device; The terminal sends, to the network side device, relevant information of the encoding AI unit currently used by the terminal according to the trigger indication of the network side device based on the first event; The terminal actively sends, based on the second event, to the network side device, relevant information of the encoding AI unit currently used by the terminal; The terminal sends the input of the currently used coding AI unit and the output of the currently used coding AI unit to the network side device; The first event and the second event are associated with CSI feedback.

9. The method according to claim 1, characterized in that: The method further includes any of the following: In the terminal capability reporting stage, the terminal sends first information to the network side device based on the instruction of the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit including the target layer; During the terminal capability reporting stage, the terminal actively sends first information to the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit containing the target layer.

10. A CSI feedback method, characterized in that: include: The network-side device training obtains the target layer in the updated coding AI unit; The network side device sends first data to the terminal; The network side device receives CSI feedback information from the terminal; Among them, the first data is used by the terminal to update the target layer in the coding AI unit to obtain an updated coding AI unit, and the coding AI unit is an AI unit for the terminal to encode CSI, and the first data contains information associated with the target layer in the updated coding AI unit.

11. The method according to claim 10, characterized in that The target layer satisfies at least one of the following: The output of the target layer is the output of the encoding AI unit; The input of the target layer is the output of other parts of the encoding AI unit except the target layer.

12. The method according to claim 10, characterized in that The target layer is used to determine the payload of the CSI feedback.

13. The method according to claim 10, characterized in that The first data includes at least one of the following: At least part of the model structure information of the target layer; at least some model parameters of the target layer; A first training data set, wherein the first training data set is used for the terminal training to obtain a target layer in the update coding AI unit.

14. The method according to claim 10, characterized in that The network side device trains and updates the target layer in the coding AI unit, including: The network side device receives second data from the terminal; The network side device trains and updates the target layer in the coding AI unit based on the second data.

15. The method according to claim 14, characterized in that The second data includes at least one of the following: A second training data set, wherein the second training data set includes an input of a target layer in the encoding AI unit and an output of a decoding AI unit, wherein the decoding AI unit is an AI unit that performs decoding processing on the CSI feedback information by the network side device, and the input of the target layer corresponds to the output of the decoding AI unit in a one-to-one manner; All model information of the encoding AI unit; Model information of other parts in the encoding AI unit except the target layer.

16. The method according to claim 10, characterized in that The method further comprises: The network side device determines whether the coding AI unit used by the terminal changes according to at least one of the following: The terminal periodically sends to the network side device, according to the configuration of the network side device, information related to the coding AI unit currently used by the terminal; The terminal sends, to the network side device, information related to the encoding AI unit currently used by the terminal according to the trigger indication of the network side device based on the first event; The terminal actively sends the relevant information of the encoding AI unit currently used by the terminal to the network side device based on the second event; The input of the currently used coding AI unit and the output of the currently used coding AI unit sent by the terminal to the network side device; The first event and the second event are associated with CSI feedback.

17. The method according to claim 10, characterized in that The method further includes any of the following: In the terminal capability reporting stage, the network side device instructs the terminal to send first information to the network side device, where the first information is used to indicate whether the terminal supports the coding AI unit including the target layer; During the terminal capability reporting phase, the network side device receives first information actively sent by the terminal, where the first information is used to indicate whether the terminal supports the coding AI unit containing the target layer.

18. A CSI feedback device, characterized in that: include: A first receiving module, used for a terminal to receive first data from a network side device; An updating module, configured for the terminal to update a target layer in a coding AI unit according to the first data to obtain an updated coding AI unit; A first processing module, configured for the terminal to encode the CSI through the update encoding AI unit to obtain CSI feedback information; A first sending module, configured for the terminal to send the CSI feedback information to the network side device; The coding AI unit is an AI unit for encoding CSI by the terminal, and the first data includes information associated with a target layer in the updated coding AI unit.

19. A CSI feedback device, characterized in that: include: The training module is used for training network-side devices to obtain the target layer in the updated coding AI unit; A second sending module, used for the network side device to send first data to the terminal; A second receiving module, the network side device receives CSI feedback information from the terminal; Among them, the first data is used by the terminal to update the target layer in the coding AI unit to obtain an updated coding AI unit, and the coding AI unit is an AI unit for the terminal to encode CSI, and the first data contains information associated with the target layer in the updated coding AI unit.

20. A terminal, characterized in that: The method comprises a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the CSI feedback method according to any one of claims 1 to 9 are implemented.

21. A network side device, characterized in that: It includes a processor and a memory, the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the CSI feedback method according to any one of claims 10 to 18 are implemented.

22. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by the processor, the steps of the CSI feedback method according to any one of claims 1 to 9 are implemented, or the steps of the CSI feedback method according to any one of claims 10 to 18 are implemented.

Images