Ai unit processing method and apparatus, and terminal and first device

Abstract

The present application belongs to the technical field of communications. Disclosed are an AI unit processing method and apparatus, and a terminal and a first device. The AI unit processing method in the embodiments of the present application comprises: a terminal executing at least one of a first operation and a second operation. The first operation comprises: acquiring a first data set, wherein the first data set comprises first target CSI information; and on the basis of first target CSI information and a target AI unit, acquiring first information, wherein the first information comprises at least one of the following: first CSI feedback information; first indicator information; first test performance information of the target AI unit; and first failure cause information. The second operation comprises: acquiring a second data set, wherein the second data set comprises first channel measurement information; and on the basis of the first channel measurement information and the target AI unit, acquiring second information, wherein the second information comprises at least one of the following: second CSI feedback information; first monitoring result information of the target AI unit; second indicator information; second test performance information of the target AI unit; and second failure cause information.

Description

AI unit processing method, device, terminal and first device

[0001] Cross-reference to related applications

[0002] This application claims priority to Chinese Patent Application No. 202411486715.2, filed in China on October 23, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application belongs to the field of communication technology, and specifically relates to an AI unit processing method, apparatus, terminal and first device. Background Technology

[0004] Currently, in mobile communication networks, tasks or services can be performed based on Artificial Intelligence (AI). For example, terminals can compress Channel State Information (CSI) based on pre-trained AI units, and network-side devices can decompress CSI based on pre-trained AI units. However, because the data distribution, processing methods, or acquisition scenarios of the data used in actual CSI processing based on AI units may differ from those of the data used during the training phase of the AI ​​unit, this can easily lead to poor performance in CSI processing based on that AI unit. Summary of the Invention

[0005] This application provides an AI unit processing method, apparatus, terminal, and first device, which can solve the problem of poor performance of AI units in CSI processing.

[0006] Firstly, an AI unit processing method is provided, the method comprising:

[0007] The terminal performs at least one of the first and second operations;

[0008] The first operation includes:

[0009] Obtain a first dataset, which includes first target channel state information (CSI) information;

[0010] First information is obtained based on the first target CSI information and the target artificial intelligence (AI) unit;

[0011] The first information includes at least one of the following:

[0012] First CSI feedback information;

[0013] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0014] The first test performance information of the target AI unit;

[0015] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0016] The second operation includes:

[0017] Obtain a second dataset, which includes the first channel measurement information;

[0018] The second information is obtained based on the first channel measurement information and the target AI unit;

[0019] The second information includes at least one of the following:

[0020] Second CSI feedback information;

[0021] The first monitoring result information of the target AI unit;

[0022] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0023] The second test performance information of the target AI unit;

[0024] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0025] Secondly, an AI unit processing device is provided, the device comprising:

[0026] A first processing module is configured to perform at least one of a first operation and a second operation;

[0027] The first operation includes:

[0028] Obtain a first dataset, which includes first target channel state information (CSI) information;

[0029] First information is obtained based on the first target CSI information and the target artificial intelligence (AI) unit;

[0030] The first information includes at least one of the following:

[0031] First CSI feedback information;

[0032] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0033] The first test performance information of the target AI unit;

[0034] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0035] The second operation includes:

[0036] Obtain a second dataset, which includes the first channel measurement information;

[0037] The second information is obtained based on the first channel measurement information and the target AI unit;

[0038] The second information includes at least one of the following:

[0039] Second CSI feedback information;

[0040] The first monitoring result information of the target AI unit;

[0041] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0042] The second test performance information of the target AI unit;

[0043] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0044] Thirdly, an AI unit processing method is provided, the method comprising:

[0045] The first device performs at least one of the fourth and fifth operations;

[0046] The fourth operation includes at least one of the following:

[0047] Send a first dataset, which includes first target CSI information;

[0048] Send a second dataset, which includes the first channel measurement information;

[0049] The fifth operation includes at least one of the following: receiving fifth information, receiving sixth information;

[0050] The fifth piece of information includes at least one of the following:

[0051] First CSI feedback information;

[0052] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0053] The first test performance information of the target AI unit;

[0054] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0055] The sixth piece of information includes at least one of the following:

[0056] Second CSI feedback information;

[0057] The first monitoring result information of the target AI unit;

[0058] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0059] The second test performance information of the target AI unit;

[0060] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0061] Fourthly, an AI unit processing device is provided, the device comprising:

[0062] The second processing module is used to perform at least one of the fourth and fifth operations;

[0063] The fourth operation includes at least one of the following:

[0064] Send a first dataset, which includes first target CSI information;

[0065] Send a second dataset, which includes the first channel measurement information;

[0066] The fifth operation includes at least one of the following: receiving fifth information, receiving sixth information;

[0067] The fifth piece of information includes at least one of the following:

[0068] First CSI feedback information;

[0069] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0070] The first test performance information of the target AI unit;

[0071] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0072] The sixth piece of information includes at least one of the following:

[0073] Second CSI feedback information;

[0074] The first monitoring result information of the target AI unit;

[0075] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0076] The second test performance information of the target AI unit;

[0077] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0078] Fifthly, an apparatus for AI unit processing is provided, the apparatus being configured to perform the steps of the method described in the first aspect, or to implement the steps of the method described in the third aspect.

[0079] In a sixth aspect, a terminal is provided, the terminal including a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the method as described in the first aspect.

[0080] In a seventh aspect, a terminal is provided, including a processor and a communication interface, wherein the processor is configured to perform at least one of a first operation and a second operation;

[0081] The first operation includes:

[0082] Obtain a first dataset, which includes first target channel state information (CSI) information;

[0083] First information is obtained based on the first target CSI information and the target artificial intelligence (AI) unit;

[0084] The first information includes at least one of the following:

[0085] First CSI feedback information;

[0086] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0087] The first test performance information of the target AI unit;

[0088] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0089] The second operation includes:

[0090] Obtain a second dataset, which includes the first channel measurement information;

[0091] The second information is obtained based on the first channel measurement information and the target AI unit;

[0092] The second information includes at least one of the following:

[0093] Second CSI feedback information;

[0094] The first monitoring result information of the target AI unit;

[0095] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0096] The second test performance information of the target AI unit;

[0097] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0098] In an eighth aspect, a first device is provided, the first device including a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the method as described in the third aspect.

[0099] In a ninth aspect, a first device is provided, including a processor and a communication interface, wherein the processor is configured to perform at least one of a fourth operation and a fifth operation;

[0100] The fourth operation includes at least one of the following:

[0101] Send a first dataset, which includes first target CSI information;

[0102] Send a second dataset, which includes the first channel measurement information;

[0103] The fifth operation includes at least one of the following: receiving fifth information, receiving sixth information;

[0104] The fifth piece of information includes at least one of the following:

[0105] First CSI feedback information;

[0106] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0107] The first test performance information of the target AI unit;

[0108] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0109] The sixth piece of information includes at least one of the following:

[0110] Second CSI feedback information;

[0111] The first monitoring result information of the target AI unit;

[0112] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0113] The second test performance information of the target AI unit;

[0114] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0115] In a tenth aspect, a readable storage medium is provided, on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect, or implement the steps of the method described in the third aspect.

[0116] Eleventhly, a wireless communication system is provided, comprising: a terminal and a first device, wherein the terminal is configured to perform the steps of the AI ​​unit processing method as described in the first aspect, and the first device is configured to perform the steps of the AI ​​unit processing method as described in the third aspect.

[0117] In a twelfth aspect, a chip is provided, the chip including a processor and a communication interface coupled to the processor, the processor being configured to run a program or instructions to implement the steps of the method described in the first aspect, or to implement the steps of the method described in the third aspect.

[0118] In a thirteenth aspect, a computer program / program product is provided, which is stored in a storage medium and is executed by at least one processor to implement the steps of the method as described in the first aspect, or to implement the steps of the method as described in the third aspect.

[0119] In this embodiment, the terminal acquires a first dataset, which includes first target CSI information, and acquires first information based on the first target CSI information and a target AI unit. The first information includes at least one of the following: first CSI feedback information; first indicator information, which is the performance indicator information of the target AI unit on the first dataset; first test performance information of the target AI unit; and first failure reason information, which indicates the reason for the failure of the target AI unit. Since in this embodiment, the terminal processes the first target CSI information of the first dataset based on the target AI unit to acquire the first information to test the performance of the target AI unit, this helps to ensure the performance of CSI processing based on the target AI unit; and / or, the terminal... The terminal acquires a second dataset, which includes first channel measurement information; it acquires second information based on the first channel measurement information and the target AI unit; the second information includes at least one of the following: second CSI feedback information; first monitoring result information of the target AI unit; second indicator information, which is the performance indicator information of the target AI unit on the second dataset; second test performance information of the target AI unit; and second failure reason information, which is used to indicate the reason for the failure of the target AI unit. Since in this embodiment, the terminal processes the first channel measurement information of the second dataset based on the target AI unit to acquire the second information in order to test the performance of the target AI unit, it is beneficial to ensure the performance of CSI processing based on the target AI unit. Attached Figure Description

[0120] Figure 1 is a block diagram of a wireless communication system applicable to an embodiment of this application;

[0121] Figure 2a is a schematic diagram of a neural network structure provided in an embodiment of this application;

[0122] Figure 2b is a schematic diagram of the structure of a neuron provided in an embodiment of this application;

[0123] Figure 3a is a schematic diagram of AI-based CSI encoding and decoding provided in an embodiment of this application;

[0124] Figure 3b is a schematic diagram of time-frequency spatial domain CSI encoding and decoding provided in an embodiment of this application;

[0125] Figure 3c is one of the schematic diagrams of the packaged CSI compression provided in the embodiments of this application;

[0126] Figure 3d is a second schematic diagram of the packaged CSI compression provided in the embodiment of this application;

[0127] Figure 4 is a flowchart of an AI unit processing method provided in an embodiment of this application;

[0128] Figure 5 is a flowchart of another AI unit processing method provided in an embodiment of this application;

[0129] Figure 6 is a flowchart of another AI unit processing method provided in an embodiment of this application;

[0130] Figure 7 is a flowchart of another AI unit processing method provided in an embodiment of this application;

[0131] Figure 8 is a flowchart of an AI-based communication method provided in an embodiment of this application;

[0132] Figure 9 is a structural diagram of an AI unit processing device provided in an embodiment of this application;

[0133] Figure 10 is a structural diagram of another AI unit processing device provided in an embodiment of this application;

[0134] Figure 11 is a structural diagram of the communication device provided in an embodiment of this application;

[0135] Figure 12 is a structural diagram of the terminal provided in an embodiment of this application;

[0136] Figure 13 is a structural diagram of the network-side device provided in an embodiment of this application. Detailed Implementation

[0137] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0138] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, not limited in number; for example, the first object can be one or more. Furthermore, "or" in this application indicates at least one of the connected objects. For example, the scope of protection for "A or B" covers at least three scenarios: Scenario 1: including A but not B; Scenario 2: including B but not A; Scenario 3: including both A and B. In addition, the terms "A and / or B," "at least one of A and B," and "at least one of A or B" also cover at least the above three scenarios. The character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0139] The term "instruction" in this application can be either a direct instruction (or explicit instruction) or an indirect instruction (or implicit instruction). A direct instruction can be understood as one in which the sender explicitly informs the receiver of specific information, the operation to be performed, or the requested result, etc., in the instruction sent. An indirect instruction can be understood as one in which the receiver determines the corresponding information based on the instruction sent by the sender, or makes a judgment and determines the operation to be performed or the requested result, etc., based on the judgment result.

[0140] It is worth noting that the technologies described in this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, 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 this application are often used interchangeably, and the described technologies can be used with the systems and radio technologies mentioned above, as well as with other systems and radio technologies. The following description describes New Radio (NR) systems for illustrative purposes, and the term NR is used in most of the following description; however, these technologies can also be applied to systems other than NR systems, such as 6th generation (6G) radio systems. th Generation 6G communication system.

[0141] Figure 1 shows a block diagram of a wireless communication system applicable to an embodiment of this application. The wireless communication system includes a terminal 11 and a network-side device 12. The terminal 11 can be a mobile phone, tablet computer, laptop computer, notebook computer, personal digital assistant (PDA), handheld computer, netbook, ultra-mobile personal computer (UMPC), mobile internet device (MID), augmented reality (AR), virtual reality (VR) device, robot, wearable device, flight vehicle, vehicle user equipment (VUE), shipboard equipment, pedestrian user equipment (PUE), smart home (home devices with wireless communication capabilities, such as refrigerators, televisions, washing machines, or furniture), game console, personal computer (PC), ATM, or self-service machine, etc. Wearable devices include: smartwatches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart chains, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, etc. Among these, in-vehicle devices can also be referred to as in-vehicle terminals, in-vehicle controllers, in-vehicle modules, in-vehicle components, in-vehicle chips, or in-vehicle units, etc. It should be noted that the specific type of terminal 11 is not limited in this application embodiment. Network-side equipment 12 may include access network equipment or core network equipment, wherein access network equipment may also be referred to as Radio Access Network (RAN) equipment, radio access network function, or radio access network unit. Access network equipment may include base stations, Wireless Local Area Network (WLAN) access points (APs), or Wireless Fidelity (WiFi) nodes, etc.The term "base station" can be referred to as Node B (NB), Evolved Node B (eNB), Next Generation Node B (gNB), New Radio Node B (NR Node B), Access Point, Relay Base Station (RBS), Serving Base Station (SBS), Base Transceiver Station (BTS), Radio Base Station, Radio Transceiver, Basic Service Set (BSS), Extended Service Set (ESS), Home Node B (HNB), Home Evolved Node B, Transmit / Receive Point (TRP), or any other suitable term in the relevant field, as long as the same technical effect is achieved. The term "base station" is not limited to any specific technical terminology. It should be noted that this application embodiment only uses a base station in an NR system as an example for description and does not limit the specific type of base station.

[0142] Core network equipment, also known as core network nodes, core network functions, or core network elements, includes, but is not limited to, at least one of the following: Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (or L-NEF), and Binding Support. The core network functions include: BSF (Block Network Function), Application Function (AF), Location Management Function (LMF), Gateway Mobile Location Centre (GMLC), and Network Data Analytics Function (NWDAF). It should be noted that this application embodiment only uses core network equipment in the NR system as an example and does not limit the specific type of core network equipment. If the name of the core network equipment mentioned in this application embodiment changes in subsequent protocol versions (e.g., 6G), it will still be within the scope of protection of this application.

[0143] Optionally, the core network equipment can be implemented by one or more functional modules in a single device, or by multiple devices working together; this application does not specifically limit this. It is understood that the aforementioned functional modules can be network elements in hardware devices, software functional modules running on dedicated hardware, or virtualized functional modules instantiated on a platform (e.g., a cloud platform).

[0144] For ease of understanding, the following describes some aspects of the embodiments of this application:

[0145] I. Artificial Intelligence (AI)

[0146] Artificial intelligence (AI) has been widely applied in various fields. Integrating AI into wireless communication networks to significantly improve technical indicators such as throughput, latency, and user capacity is an important task for future wireless communication networks. AI modules can be implemented in various ways, such as neural networks, decision trees, support vector machines, and Bayesian classifiers. This application uses neural networks as an example for illustration, but does not limit the specific type of AI module.

[0147] For example, a neural network can be shown in Figure 2a. The neural network is composed of neurons, and each neuron can be shown in Figure 2b. Here, a1, a2, ..., aK are the 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, and the Rectified Linear Unit (ReLU), etc.

[0148] The parameters of a neural network are optimized using gradient optimization algorithms. Gradient optimization algorithms are a class of algorithms that minimize or maximize an objective function (also known as a loss function), which is often a mathematical combination of model parameters and data. For example, given data X and its corresponding label Y, we construct a neural network model f(.). With the model, we can obtain the predicted output f(x) based on the input x, and calculate the difference between the predicted value and the true value (f(x) - Y), which is the loss function. The goal is to find suitable W and b that minimize the value of the above loss function, where a smaller loss value indicates that the model is closer to the reality.

[0149] Most common optimization algorithms are based on the error back propagation (BP) algorithm. The basic idea of ​​the BP algorithm is that the learning process consists of two parts: forward propagation of the signal and backward propagation of the error. During forward propagation, the input sample is introduced from the input layer, processed layer by layer by the hidden layers, and then propagated to the output layer. If the actual output of the output layer does not match the expected output, the process transitions to the error back propagation stage. Error back propagation involves propagating the output error back to the input layer layer by layer through the hidden layers, distributing the error to all units in each layer, thus obtaining the error signal of each unit. This error signal serves as the basis for adjusting the weights of each unit. This process of adjusting the weights through forward and backward propagation is repeated continuously. This 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 predetermined number of learning iterations is reached.

[0150] Common optimization algorithms include gradient descent, stochastic gradient descent (SGD), mini-batch gradient descent, momentum method, Nesterov (named after the inventor, specifically stochastic gradient descent with momentum), adaptive gradient descent (Adagrad), Adadelta, root mean square prop (RMSprop), and adaptive momentum estimation (Adam).

[0151] During error backpropagation, these optimization algorithms calculate the gradient based on the error / loss obtained from the loss function with respect to the current neuron, add the learning rate, previous gradients / derivatives / partial derivatives, etc., and then pass the gradient to the previous layer.

[0152] II. AI Units / AI Models

[0153] The AI ​​unit / AI model in this application embodiment can also be referred to as a Machine Learning (ML) model, ML unit, AI structure, AI function, AI characteristic, machine learning model, neural network, neural network function, neural network functionality, etc. Alternatively, the aforementioned AI unit / AI model can refer to a processing unit capable of implementing specific algorithms, formulas, processing flows, capabilities, etc., related to AI. Or, the AI ​​unit / AI model can be a processing method, algorithm, function, module, or unit for a specific dataset. Alternatively, the AI ​​unit / AI model can be a processing method, algorithm, function, module, or unit running on AI / ML related hardware such as GPU, NPU, TPU, and ASIC. This application embodiment does not specifically limit these aspects. Optionally, the specific dataset includes at least one of the inputs and outputs of the AI ​​unit / AI model.

[0154] 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 dataset associated with the AI ​​unit / AI model, or an identifier of a specific scenario, environment, channel characteristics, or device related to the AI / ML, or an identifier of a function, feature, capability, or module related to the AI / ML. This application embodiment does not specifically limit this.

[0155] III. Model Lifecycle

[0156] This describes the lifecycle management (LCM) of AI / ML models (e.g., model training, model deployment, model inference, model monitoring, and model updating) and AI / ML functions. The model lifecycle includes at least one or more of the following: model training, model deployment, model inference, model monitoring, and model updating.

[0157] IV. Combining Channel State Information (CSI) Compression with AI

[0158] 1. Training collaboration types of AI / ML CSI compressed models

[0159] CSI compression is a typical two-end model use case, meaning the complete CSI compression model needs to be deployed on different network nodes. Currently, most considerations involve deploying the encoder on the User Equipment (UE) side and the decoder on the Network (NW) side. Models or sub-models deployed on multiple nodes need to be paired with each other to function properly. Considering the above characteristics of two-end models, 3GPP has identified several basic training collaboration types:

[0160] 1) Joint training at a single entity, also known as type 1.

[0161] This training framework refers to training a complete encoder and decoder model on a network node (UE, NW, or third-party server node, etc.), and then deploying the corresponding model modules to the target node through methods such as model transfer, for example, transferring the encoder part to the UE and the decoder part to the NW.

[0162] 2) Joint training at multiple entities, also known as type 2.

[0163] This training framework involves multiple nodes collaboratively participating in the training process, with each node independently calculating the forward / backward propagation information required for its local model training and updating its own model parameters. Since the training process requires forward / backward propagation of the entire model (including the encoder and decoder), participating nodes need to exchange the corresponding forward / backward propagation information. Once training is complete, model transfer between nodes is no longer necessary.

[0164] 3) Separate training on multiple nodes, also known as type 3.

[0165] This training framework involves first training a reference model on a specific node, then sending the reference model's information to the target node. Finally, the target node uses this information to train its own required model, ensuring that each node's model or sub-model can be paired and used interchangeably. For example, the NW side first trains a complete encoder-decoder model and determines that the resulting decoder is the one to be used in the future. Then, it sends the encoder's information (usually the encoder's input and output data) to the UE side, which then trains its own encoder based on this information. This training framework can be further subdivided into UE-first training and NW-first training. UE-first training means training the complete model on the UE side first, then sending the information needed for the NW to train its matching model (usually the NW side's input and output data) to the NW side. Conversely, NW-first training means training the complete model on the NW side first, then sending the information needed for the UE to train its matching model (usually the UE side's input and output data) to the UE side.

[0166] 2. AI-based CSI / Precoding Matrix Indicator (PMI) Compression

[0167] In Release 18 (R18), the UE's expected or target CSI or codebook W N*B Compression is performed using AI, such as compressing it into an AI-based PMI value, and then reported to the network-side device. The network-side device performs decompression to obtain W′. N*B As shown in Figure 3a.

[0168] Building upon the spatial-frequency domain CSI compression studied in R18 SI, this paper introduces the utilization of time-domain CSI correlation, that is, jointly compressing CSI from multiple time slots, thereby further reducing the overhead of CSI reporting or improving the accuracy of CSI reporting. Figure 3b shows a schematic diagram of time-frequency spatial-domain CSI compression, which illustrates the joint compression and reporting of CSI from four slots, while the CSI from each slot can be regarded as a separate spatial-frequency domain CSI report.

[0169] Furthermore, based on the reporting method of CSI on multiple slots, time-frequency spatial domain CSI compression can be further divided into two types: packaged reporting and progressive reporting. Packaged reporting is to report CSI on multiple slots at once, as shown in Figures 3c and 3d, while progressive reporting is to report CSI on each slot sequentially in an autoregressive manner, as shown in Figure 3b.

[0170] It should be noted that if there is an inconsistency between the UE-side data (such as CSI or codebook matrix) actually used for CSI processing based on the AI ​​unit and the training data of the AI ​​unit, performance degradation in CSI processing may occur, such as decreased system throughput and poor AI unit performance. For example, the following problems may exist:

[0171] Question 1: When the terminal receives the encoder (i.e. AI unit) sent by the network side, if the data distribution of the CSI or codebook matrix (which can be understood as the input information input from the terminal to the encoder) obtained by the terminal side is inconsistent with the data distribution of the CSI or codebook matrix used by the network side to train the encoder, it will lead to performance loss, such as reduced system throughput and poor performance of AI unit type.

[0172] Question 2: When the terminal receives the encoder sent by the network side, if the UE-side processing method of the CSI or codebook matrix (which can be understood as the input information input from the terminal side to the encoder) obtained by the terminal side is inconsistent with the processing method used by the network side in the dataset used to train the encoder, it will lead to performance loss, such as reduced system throughput and poor AI unit performance.

[0173] Question 3: When the terminal receives the encoder sent by the network side, if the channel or scene of the CSI or codebook matrix (which can be understood as the input information input from the terminal to the encoder) obtained by the terminal side is inconsistent with the channel conditions and scene distribution of the CSI or codebook matrix used by the network side to train the encoder, it will lead to performance loss, such as reduced system throughput and poor AI unit performance.

[0174] It should also be noted that the test in this embodiment can also be described as verification, etc. Furthermore, the AI ​​unit processing method in this embodiment is not limited to AI units used for CSI processing, but can also be used for AI units used in other services, such as video services, voice services, etc. That is to say, for AI units in other services, the AI ​​unit processing method provided in this application embodiment can also be used to test the effectiveness of the AI ​​unit.

[0175] The AI ​​unit processing method provided in this application will be described in detail below with reference to the accompanying drawings, through some embodiments and application scenarios.

[0176] Please refer to Figure 4, which is a flowchart of an AI unit processing method provided in an embodiment of this application. This method can be executed by a terminal, and as shown in Figure 4, it includes the following steps:

[0177] Step 401: The terminal performs at least one of the first operation and the second operation;

[0178] The first operation includes:

[0179] Obtain a first dataset, which includes first target CSI information;

[0180] First information is obtained based on the first target CSI information and the target AI unit;

[0181] The first information includes at least one of the following:

[0182] First CSI feedback information;

[0183] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0184] The first test performance information of the target AI unit;

[0185] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0186] The second operation includes:

[0187] Obtain a second dataset, which includes the first channel measurement information;

[0188] The second information is obtained based on the first channel measurement information and the target AI unit;

[0189] The second information includes at least one of the following:

[0190] Second CSI feedback information;

[0191] The first monitoring result information of the target AI unit;

[0192] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0193] The second test performance information of the target AI unit;

[0194] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0195] In this embodiment, the first dataset can be predefined by the protocol or sent to the terminal by a network-side device or server. In some embodiments, the first dataset can be consistent with the training dataset used for the target AI model; for example, the first dataset and the training dataset used for the target AI model may have the same data distribution, processing method, or acquisition scenario. It should be noted that the first dataset can also be called the first testing dataset, test dataset 1, first performance testing dataset, or performance testing dataset 1, etc.

[0196] The aforementioned first dataset includes at least first target CSI information, such as codebook information. This codebook information may include a codebook matrix, for example, W. N*B Where N is the number of ports and B is the number of frequency domain units, or the codebook matrix in the angle-delay domain; or, the above codebook information may include a compressed codebook matrix, such as type II codebook information, such as weighted amplitude coefficients, phase information, orthogonal beam vectors, etc.

[0197] The aforementioned target AI unit may include, but is not limited to, at least one of an AI unit for acquiring CSI feedback information and an AI unit for acquiring reconstructed CSI information. The AI ​​unit for acquiring CSI feedback information may include an AI unit trained by the terminal or an AI unit trained by a network-side device or server. Similarly, the AI ​​unit for acquiring reconstructed CSI information may also include an AI unit trained by the terminal or an AI unit trained by a network-side device or server. It should be noted that the AI ​​unit for acquiring CSI feedback information may also be called an encoder, and the AI ​​unit for acquiring reconstructed CSI information may also be called a decoder. It is understood that, when the target AI unit is trained by a network-side device or server, the method further includes the terminal receiving the target unit from the network-side device or server.

[0198] The aforementioned CSI feedback information can be understood as compressed or encoded CSI information, such as CSI information obtained by compressing or encoding the target CSI information based on the AI ​​unit.

[0199] The aforementioned first CSI feedback information can be understood as CSI feedback information obtained based on the first target CSI information. For example, the terminal can input the first target CSI information into the target AI unit to obtain the first CSI feedback information. In some optional embodiments, the aforementioned first CSI feedback information can be used to test the performance of the target AI unit. For example, the aforementioned first CSI feedback information can be used to obtain performance indicator information of the target AI unit on a first dataset or to obtain test performance information of the target AI unit on the first dataset, etc. For example, the terminal can obtain the test performance information of the target AI unit from the expected CSI feedback information corresponding to the first CSI feedback information and the first target CSI information. The expected CSI feedback information can be understood as the CSI feedback information corresponding to the first target CSI information obtained by the network-side device based on the network-side AI unit. Alternatively, the terminal can report the first CSI feedback information, and the network-side device can reconstruct the CSI information based on the first CSI feedback information, and then obtain the test performance information of the target AI unit based on the reconstructed CSI information and the first target CSI information. Or, the terminal can report the first CSI feedback information, and the network-side device can reconstruct the CSI information based on the first CSI feedback information, and then obtain the test performance information of the target AI unit by comparing the reconstructed CSI information with the reconstructed CSI information obtained by the network-side device based on the expected CSI feedback information. For example, the expected CSI feedback information can be the CSI feedback information obtained by the terminal's peer device based on the first target CSI information and the reference AI unit. The peer device of the terminal could be, for example, a network-side device, a base station, or a test device.

[0200] The aforementioned first indicator information is the performance indicator information of the target AI unit on the first dataset, that is, the aforementioned first indicator information can be used to reflect the performance indicators of the target AI unit tested based on the first dataset. In one optional embodiment, the first indicator information may be indicator information obtained based on the first CSI feedback information and the expected CSI feedback information sent by the network side or determined by the protocol; in another optional embodiment, the first indicator information may be indicator information obtained based on the first CSI reconstruction information and the first target CSI information determined by the network side device; in yet another optional embodiment, the first indicator information may be indicator information obtained based on the first target CSI information, the first CSI feedback information, and combined with the first AI model; wherein, the first AI model is used to infer the indicator information of the target AI unit of the terminal.

[0201] The first test performance information of the target AI unit mentioned above can be understood as the test performance information of the target AI unit on the first dataset. For example, the above test performance information can be used to reflect the effectiveness of the target AI unit, etc.

[0202] The aforementioned first failure reason information is used to indicate the reason for the failure of the target AI unit. For example, the reason for the failure of the target AI unit may include, but is not limited to, inconsistent data distribution, inconsistent data processing methods, or inconsistent scene distribution. In some optional embodiments, the aforementioned first failure reason information may be used to indicate the reason for the failure of the target AI unit under a first dataset.

[0203] In some optional implementations, after obtaining the first information, the terminal may report part or all of the information in the first information. Different information items of the first information may be carried in the same information or signaling, or they may be carried in different information or signaling. For example, the first CSI feedback information may be carried in uplink control information (UCI), while the first test performance information or the first failure reason information of the target AI unit may be carried in higher-layer signaling.

[0204] In some embodiments, where the first dataset includes multiple first target CSI information, the first information may include first information obtained based on the multiple first target CSI information.

[0205] The aforementioned second dataset can be predefined by the protocol, or it can be sent to the terminal by network-side devices or servers. It should be noted that this second dataset can also be called the second testing dataset, test dataset 2, performance testing dataset, or performance testing dataset 2, etc.

[0206] The aforementioned second dataset includes at least the first channel measurement information. The first channel measurement information can be understood as raw channel information, such as measurement information obtained based on Channel State Information Reference Signal (CSI-RS) measurements.

[0207] The first channel measurement information may include a raw channel matrix. For example, this raw channel matrix may include a first raw channel matrix, the dimension of which may be TX*RX*FRU, where TX represents the number of TX antennas, RX represents the number of RX antennas, and FRU represents the number of frequency domain units. It is understood that the three dimensions (TX, RX, and FRU) of the first raw channel matrix can be ordered in any order, and this embodiment does not limit this. For example, the dimension of the first raw channel matrix may also be RX*TX*FRU or FRU*TX*RX, etc. Alternatively, the raw channel matrix may include a second raw channel matrix, the dimension of which may be TX*RX*delay, where TX represents the number of TX antennas, RX represents the number of RX antennas, and delay represents the number of delay units. It is understood that the three dimensions (TX, RX, and delay) of the second raw channel matrix can be ordered in any order, and this embodiment does not limit this. For example, the dimension of the second raw channel matrix may also be RX*TX*delay or delay*TX*RX, etc.

[0208] The aforementioned second CSI feedback information can be understood as CSI feedback information obtained based on the first channel measurement information. For example, target CSI information can be obtained based on the first channel measurement information, and this target CSI information can be input into the target AI unit to obtain the second CSI feedback information. In some optional embodiments, the aforementioned second CSI feedback information can be used to test the performance of the target AI unit. For example, the aforementioned second CSI feedback information can be used to obtain performance index information of the target AI unit on a second dataset or to obtain test performance information of the target AI unit on a second dataset. Exemplarily, the terminal can report the second CSI feedback information, and the network-side device can obtain reconstructed CSI information based on the second CSI feedback information, and then obtain test performance information of the target AI unit based on the reconstructed CSI information and the real CSI information corresponding to the second CSI feedback information. In other optional embodiments, the aforementioned second CSI feedback information can be used to test terminal-side condition information or the characteristics of terminal-side processing measurement information. That is, if the validity of the target AI unit is determined based on the first dataset, and the test performance or indicator information does not meet the requirements based on the second dataset, then it can be determined that the characteristics of the terminal-side condition information or terminal-side processing measurement information are different from those of the network-side dataset or the terminal-side condition information or terminal-side processing measurement information of the dataset used to obtain the target AI unit.

[0209] The first monitoring result information of the target AI unit mentioned above can be understood as the monitoring result information obtained by monitoring the target AI unit based on the second dataset. For example, the first monitoring result information of the target AI unit can be obtained based on the real CSI information obtained from the first channel measurement information and the reconstructed CSI information obtained based on the second CSI feedback information.

[0210] The aforementioned second indicator information is the performance indicator information of the target AI unit on the second dataset, that is, the aforementioned second indicator information can be used to reflect the performance indicators of the target AI unit tested based on the second dataset. In one optional embodiment, the second indicator information may be indicator information obtained based on the second CSI feedback information and the expected CSI feedback information sent by the network side or determined by the protocol; in another optional embodiment, the second indicator information may be indicator information obtained based on the fourth reconstructed CSI information and the second target CSI information determined by the network-side device, wherein the second target CSI information is the target CSI information obtained by the network-side device based on the first channel measurement information; in yet another optional embodiment, the second indicator information may be indicator information obtained based on the second target CSI information, the second CSI feedback information, and combined with the second AI model; wherein, the second AI model is used to infer the indicator information of the target AI unit of the terminal.

[0211] The second test performance information of the target AI unit mentioned above can be understood as the test performance information of the target AI unit on the second dataset. For example, the above test performance information can be used to reflect the effectiveness of the target AI unit, etc.

[0212] The aforementioned second failure reason information is used to indicate the reason for the failure of the target AI unit. For example, the reasons for the failure of the target AI unit may include, but are not limited to, inconsistent data distribution, inconsistent terminal-side condition information or terminal-side processing measurement information, inconsistent data processing methods, or inconsistent scene distribution. In some optional embodiments, the aforementioned second failure reason information may be used to indicate the reason for the failure of the target AI unit under a second dataset.

[0213] In some optional implementations, after obtaining the second information, the terminal can report part or all of the second information. Different information items of the second information can be carried in the same information or in different information, such as the second CSI feedback information being carried in UCI, and the second test performance information or the second failure reason information of the target AI unit being carried in higher-layer signaling.

[0214] In some embodiments, where the second dataset includes multiple first channel measurement information, the first information includes second information obtained based on the multiple first channel measurement information.

[0215] In this embodiment, the terminal acquires a first dataset, which includes first target CSI information, and acquires first information based on the first target CSI information and a target AI unit. The first information includes at least one of the following: first CSI feedback information; first indicator information, which is the performance indicator information of the target AI unit on the first dataset; first test performance information of the target AI unit; and first failure reason information, which indicates the reason for the failure of the target AI unit. Since in this embodiment, the terminal processes the first target CSI information of the first dataset based on the target AI unit to acquire the first information to test the performance of the target AI unit, this helps to ensure the performance of CSI processing based on the target AI unit; and / or, the terminal... The terminal acquires a second dataset, which includes first channel measurement information; it acquires second information based on the first channel measurement information and the target AI unit; the second information includes at least one of the following: second CSI feedback information; first monitoring result information of the target AI unit; second indicator information, which is the performance indicator information of the target AI unit on the second dataset; second test performance information of the target AI unit; and second failure reason information, which is used to indicate the reason for the failure of the target AI unit. Since in this embodiment, the terminal processes the first channel measurement information of the second dataset based on the target AI unit to acquire the second information in order to test the performance of the target AI unit, it is beneficial to ensure the performance of CSI processing based on the target AI unit.

[0216] It is understood that the target AI unit may include an encoder and / or a decoder. In one optional embodiment, if the terminal side only has an encoder, the first information only includes the first CSI feedback information; or, if the terminal side only has an encoder, the first indicator information can only be obtained based on the first CSI feedback information and the CSI feedback information expected by the network-side device (i.e., the network-side device obtains the CSI feedback information based on the first target CSI information and the target AI unit of the network-side device). In another optional embodiment, the target AI unit may include an encoder and a decoder, and the first information may include at least one of the first indicator information, the first test performance information of the target AI unit, and the first failure reason. This is because the terminal side can obtain the reconstructed CSI information based on the encoder and decoder, and can obtain the first indicator information, the first test performance information of the target AI unit, and at least one of the first failure reasons by comparing the first target CSI information and the reconstructed CSI information.

[0217] Optionally, the model structure of the target AI unit may be indicated by the network-side device or agreed upon by the protocol. If there are multiple model structures, the network-side device may need to further indicate one of the multiple model structures.

[0218] Optionally, the model parameters of the target AI unit can be sent by the network-side device, agreed upon by the protocol, or determined by the terminal itself during training.

[0219] Optionally, the first dataset further includes at least one of the following:

[0220] First dataset identification information;

[0221] The indication information or type information of the target CSI information;

[0222] Third CSI feedback information.

[0223] The aforementioned first dataset identifier (Dataset id) information can be used to identify the first dataset. In some optional embodiments, the aforementioned first dataset identifier information can also be used to indicate other information about the device acquiring the first dataset, such as scene information, area information, cell information, etc. In some optional embodiments, the aforementioned first dataset identifier information can be used to identify network-side condition information or additional condition information for acquiring the dataset, such as an associated identifier (ID), used to represent the identification information of network-side condition information or additional condition information.

[0224] The aforementioned indication information of the target CSI information can be used to indicate the first target CSI information. The aforementioned type information of the target CSI information can be used to indicate the type of the first target CSI information. For example, it can be further indicated that the first target CSI information is type 2 CSI information, or CSI information in the angle and time delay domain, or channel measurement information, etc.

[0225] The aforementioned third CSI feedback information can be understood as the expected CSI feedback information, such as codebook information compressed by the AI ​​unit. In some optional embodiments, the aforementioned third CSI feedback information can be CSI feedback information obtained by the peer device of the terminal (e.g., network-side device or server) based on the first target CSI information and the AI ​​unit on the peer device side. In this embodiment, the first target CSI information and the third CSI feedback information can be used to test the performance of the encoder on the terminal side, such as determining the first indicator information as the indicator information obtained based on the first CSI feedback information and the third feedback information; determining the first test performance information as the test performance information in the first dataset obtained based on the first CSI feedback information and the third feedback information. Furthermore, it is understood that if the first dataset includes the first target CSI information and the third CSI feedback information, the first dataset can also be used to determine the target AI unit, such as training or fine-tuning the target AI unit on the terminal side based on the first target CSI information and the third CSI feedback information.

[0226] In this embodiment, the first dataset also includes at least one of the following: first dataset identification information, indication information of target CSI information or type information of target CSI information, and third CSI feedback information. This first dataset is conducive to more accurate and convenient performance testing of the target AI unit.

[0227] Optionally, the first dataset is associated with third information, which includes at least one of the following:

[0228] The third indicator information is a performance indicator information calculated based on the first target CSI information and the first reconstructed CSI information, wherein the first reconstructed CSI information is the reconstructed CSI information calculated by the network-side device.

[0229] The first indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit;

[0230] The first target indicator information is the expected performance indicator information of the target AI unit;

[0231] First threshold information.

[0232] The aforementioned third indicator information can be understood as expected performance indicator information. In some optional embodiments, the aforementioned third indicator information is the third indicator information obtained by the peer device (e.g., network-side device or server) based on the first target CSI information. It can be understood that the relationship between the third indicator information and the first target CSI information can be one-to-one (e.g., each first target CSI information is associated with a corresponding third indicator information) or many-to-one (e.g., multiple first target CSI information are associated with a corresponding third indicator information, or all first target CSI information are associated with a third indicator information).

[0233] In some optional embodiments, the aforementioned third indicator information may be performance indicator information calculated by the terminal's peer device (e.g., a network-side device or server) based on the first target CSI information and the first reconstruction CSI information. The aforementioned first reconstruction CSI information is the reconstruction CSI information calculated by the network-side device, which can be understood as the first reconstruction CSI information being the reconstruction CSI information obtained by the network-side AI unit.

[0234] The performance metric indication information of the aforementioned target AI unit is used to indicate the performance metric of the target AI unit. The type information of the aforementioned performance metric of the target AI unit is used to indicate the type of performance metric of the target AI unit. It is understood that the aforementioned first metric indication information can be used to indicate the performance metric of the target AI unit or the type of performance metric of the target AI unit under the first dataset. For example, the aforementioned first metric indication information can be used to indicate an end-to-end performance metric, such as performance metric information obtained based on the first target CSI information or reconstructed CSI information; or, a non-end-to-end performance metric, such as performance metric information obtained based on the first CSI feedback information and the third CSI feedback information. As another example, the aforementioned first metric indication information can be used to indicate that the performance metric is for a preset percentage or a preset amount of data, or, performance metric information obtained by averaging data on the first dataset.

[0235] The aforementioned first threshold information is used to indicate the lower limit of the expected indicator obtained based on the first dataset, or, the first threshold information is used to indicate the lower limit of the indicator that the target AI unit needs to meet, or, the first threshold information is used to indicate the lower limit of the indicator, or, the first threshold information is used to indicate the lower limit of the performance obtained based on the first dataset, or, the first threshold information is used to indicate the lower limit of the performance that the target AI unit needs to meet, or, the first threshold information is used to indicate the lower limit of the performance, etc.

[0236] Optionally, the first threshold information may represent a lower limit of performance that a preset percentage of data needs to meet, or it may be a lower limit of performance that needs to be met when the first indicator information obtained from the first dataset is the average of multiple data. In an optional embodiment, if the first indicator information obtained by the terminal performing the first operation is lower than the first threshold information, then the target AI unit fails, or there is a problem with the target AI unit on the terminal side, or the terminal needs to report the target AI unit failure indication information of the network-side device, or indicate the reason for the failure of the target AI unit of the network-side device.

[0237] In an optional embodiment, the first threshold information is used to indicate the upper limit of the difference between the first indicator information and the third indicator information obtained based on the first dataset, or the first threshold information is used to indicate the upper limit of the difference between the first indicator information and the first target indicator information obtained based on the first dataset. In an optional embodiment, if the difference between the first indicator information obtained by the terminal performing the first operation and the received third indicator information is higher than the first threshold information, then the target AI unit fails, or the target AI unit on the terminal side has a problem, or the terminal needs to report the target AI unit failure indication information of the network-side device, or indicate the reason for the failure of the target AI unit of the network-side device.

[0238] The aforementioned first target indicator information is the expected performance indicator information of the target AI unit. Optionally, the first target indicator information can indirectly indicate the type of the first indicator information, and the first indicator information needs to be of the same type as the first target indicator information. In an optional embodiment, if the difference between the first indicator information and the first target indicator information is greater than a first threshold information, then the target AI unit fails, or the terminal-side target AI unit has a problem, or the terminal needs to report network-side device target AI unit failure indication information, or indicate the reason for the network-side device target AI unit failure.

[0239] In some optional embodiments, the method further includes:

[0240] The terminal receives third information associated with the first dataset, the third information including at least one of the following: the third indicator information, the first indicator indication information, the first target indicator information, and the first threshold information.

[0241] In some optional embodiments, the first dataset further includes third information, which includes at least one of the following: the third indicator information, the first indicator indication information, the first target indicator information, and the first threshold information.

[0242] Optionally, the first operation further includes at least one of the following:

[0243] Report the first CSI feedback information;

[0244] The first indicator information is reported, which is the performance indicator information of the target AI unit estimated by the terminal on the first dataset, or the first indicator information is the performance indicator information calculated based on the first target CSI information and the second reconstructed CSI information, where the second reconstructed CSI information is the reconstructed CSI information calculated by the terminal or the reconstructed CSI information calculated by the network side device.

[0245] Obtain the first test performance information of the target AI unit;

[0246] Report the first test performance information of the target AI unit.

[0247] The aforementioned first indicator information is the performance indicator information of the target AI unit on the first dataset estimated by the terminal. For example, the terminal can estimate the performance indicator information of the target AI unit based on a performance estimation algorithm or an AI unit. Alternatively, the first indicator information is the performance indicator information calculated based on the first target CSI information and the second reconstructed CSI information, wherein the second reconstructed CSI information is the reconstructed CSI information calculated by the terminal, for example, the reconstructed CSI information obtained by the terminal-side AI unit and the first CSI feedback information, or the second reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, for example, the network-side device obtains the reconstructed CSI information based on the network-side AI unit and the first CSI feedback information and sends it to the terminal.

[0248] It is understood that, if the first operation further includes acquiring the first test performance information of the target AI unit, the aforementioned first information does not include the first test performance information of the target AI unit. That is, if the first information includes at least one of the first CSI feedback information, the first indicator information, and the first failure reason information, the first operation may further include acquiring the first test performance information of the target AI unit.

[0249] In some optional embodiments, the terminal may obtain first test performance information of the target AI unit based on at least one of the first information, second information, and first dataset.

[0250] In this embodiment, the terminal acquires and / or reports at least one of the first indicator information and the first test performance information. This enables the network-side device to determine the performance of the target AI unit of the terminal on the first dataset, thereby determining the validity of the target AI unit or the problem of failure. This is beneficial for the network-side device to configure the CSI reporting method, update the CSI reporting configuration information, or update and match the target AI unit.

[0251] Optionally, obtaining the first test performance information of the target AI unit includes:

[0252] The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or...

[0253] The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the third CSI feedback information; or...

[0254] Obtain the first test performance information of the target AI unit based on the first indicator information; or...

[0255] The first test performance information of the target AI unit is obtained based on the first indicator information and the third indicator information; or...

[0256] The first test performance information of the target AI unit is obtained based on the first indicator information and the first target indicator information; or...

[0257] The first test performance information of the target AI unit is obtained based on the first threshold information; or...

[0258] The first test performance information of the target AI unit is obtained based on the first indicator information and the first threshold information; or...

[0259] The first test performance information of the target AI unit is obtained based on the first indicator information, the first threshold information, and the first target indicator information.

[0260] In some implementations, the terminal can obtain the first test performance information of the target AI unit based on the first CSI feedback information and the second CSI feedback information. For example, when the first target CSI information is target CSI information obtained based on the first channel measurement information, the terminal can obtain the similarity index information, difference, or statistical error between the first CSI feedback information and the second CSI feedback information, and obtain the first test performance information of the target AI unit based on the similarity index information, difference, or statistical error. The first test performance information can reflect whether there is a performance degradation problem caused by the data processing method not matching the target AI unit.

[0261] In other implementations, the terminal can obtain the first test performance information of the target AI unit based on the first CSI feedback information and the third CSI feedback information. For example, the terminal can obtain the similarity index information, difference, or statistical error between the first CSI feedback information and the third CSI feedback information, and obtain the first test performance information of the target AI unit based on the similarity index information, difference, or statistical error. The first test performance information can reflect whether there are any problems in the deployment of the AI ​​unit.

[0262] In some other implementations, the terminal can obtain the first test performance information of the target AI unit based on the first indicator information. For example, the terminal can compare the first indicator information with a threshold predefined by the protocol or a threshold configured by the network-side device or server to obtain the first test performance information of the target AI unit. Alternatively, the terminal can compare the first indicator information with the third indicator information to obtain the first test performance information of the target AI unit. The first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the validity or failure of the target AI unit.

[0263] In some other implementations, the terminal can obtain the first test performance information of the target AI unit based on the first indicator information and the third indicator information. For example, the terminal can calculate the difference or statistical error between the first indicator information and the third indicator information to obtain the first test performance information of the target AI unit. The first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the effectiveness or failure of the target AI unit.

[0264] In some other implementations, the terminal can obtain the first test performance information of the target AI unit based on the first indicator information and the first target indicator information. For example, the terminal can calculate the difference or statistical error between the first indicator information and the first target indicator information to obtain the first test performance information of the target AI unit. The first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the effectiveness or failure of the target AI unit.

[0265] In some other implementations, the terminal can obtain the first test performance information of the target AI unit based on the first threshold information. For example, the terminal can compare the similarity index, difference, or statistical error between the first CSI feedback information and the third CSI feedback information with the first threshold information to obtain the first test performance information of the target AI unit; or, it can compare the similarity index, difference, or statistical error between the first CSI feedback information and the second CSI feedback information with the first threshold information to obtain the first test performance information of the target AI unit; or, it can compare the difference or statistical error between the first index information and the third index information with the first threshold information to obtain the first test performance information of the target AI unit; or, it can compare the first index information with the first threshold information to obtain the first test performance information of the target AI unit, etc. This first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the effectiveness or failure of the target AI unit.

[0266] In some other implementations, the terminal can obtain the first test performance information of the target AI unit based on the first indicator information, the first threshold information, and the first target indicator information. For example, the terminal can calculate the difference or statistical error between the first indicator information and the first target indicator information, and compare the difference or statistical error between the first indicator information and the first target indicator information with the first threshold information to obtain the first test performance information of the target AI unit. This first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the effectiveness or failure of the target AI unit.

[0267] In some other implementations, the terminal can obtain the first test performance information of the target AI unit based on the first indicator information and the first threshold information. For example, the terminal can compare the first indicator information and the first threshold information to obtain the first test performance information of the target AI unit.

[0268] Optionally, the first dataset includes K data units, each of which includes first target CSI information, where K is an integer greater than 1.

[0269] In some alternative embodiments, K can be 20, 50, or 100, etc. Exemplarily, each of the above data units may include data from at least one time unit, which may include a subframe, a time slot, or a sub-time slot, etc.

[0270] Optionally, the first indicator information is determined based on the performance indicator information corresponding to L data units, where L is an integer less than or equal to K;

[0271] The performance metric information corresponding to each data unit is determined according to at least one of the following:

[0272] The third CSI feedback information and the corresponding first CSI feedback information for each data unit;

[0273] The first target CSI information and the corresponding second reconstructed CSI information for each data unit;

[0274] Estimated performance metrics information corresponding to each data unit.

[0275] In this embodiment, the aforementioned L data units are data units among the aforementioned K data units. The second reconstructed CSI information corresponding to each data unit can be understood as the reconstructed CSI information obtained based on the first CSI feedback information obtained from the first target CSI information of that data unit, or the reconstructed CSI information obtained by the terminal based on the target AI unit. The estimated performance index information corresponding to each data unit can be understood as the performance index information of that data unit estimated by the terminal.

[0276] The first indicator information mentioned above is determined based on the indicator information corresponding to L data units. For example, the first indicator information can be the maximum value, minimum value, median value, average value, or statistical value obtained based on a preset statistical algorithm for the indicator information corresponding to L data units. This embodiment does not limit this.

[0277] In this embodiment, the first indicator information is determined based on the performance indicator information corresponding to L data units, which helps to ensure the reliability of the calculated first indicator information.

[0278] Optionally, the first indicator information is determined based on the performance indicator information corresponding to the M data unit groups, where M is an integer less than or equal to K;

[0279] The performance index information corresponding to each data unit group is the performance index information determined based on at least two or all data units in each data unit group.

[0280] In this embodiment, the data units in the aforementioned M data unit groups are the data units in the aforementioned K data units. Each data unit group may include at least two data units from the aforementioned K data units, and the performance indicator information corresponding to each data unit group can be determined based on some or all of the data units in that data unit group. For example, one performance indicator can be estimated for every 4 data units, and K / 4 performance indicator information can be estimated from a dataset, and then the first indicator information can be determined by combining the K / 4 performance indicator information; or, for another example, one performance indicator can be estimated for every 10 data units, and multiple performance indicator information can be estimated from a dataset.

[0281] In this embodiment, the first indicator information is determined based on the performance indicator information corresponding to the M data unit groups, which helps to ensure the reliability of the calculated first indicator information.

[0282] Optionally, the second dataset is associated with fourth information, which includes at least one of the following:

[0283] Second dataset identification information;

[0284] Indication information or type information of channel measurement information;

[0285] The fourth indicator information is a performance indicator information calculated based on the second target CSI information and the third reconstructed CSI information. The third reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, and the second target CSI information is the target CSI information obtained by the network-side device based on the first channel measurement information.

[0286] The second indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit;

[0287] The second target indicator information is the expected performance indicator information of the target AI unit;

[0288] Second threshold information;

[0289] Format indication information, wherein the format indication information is used to indicate the format of the target CSI information;

[0290] Channel characteristic information;

[0291] Scene information.

[0292] The aforementioned second dataset identifier (Dataset id) information can be used to identify the second dataset. In some optional embodiments, the aforementioned second dataset identifier information can also be used to indicate other information about the device acquiring the second dataset, such as scene information, area information, cell information, etc.

[0293] The indication information of the aforementioned channel measurement information can be used to indicate the first channel measurement information. The type information of the aforementioned channel measurement information can be used to indicate the type of the first channel measurement information.

[0294] The aforementioned fourth indicator information can be understood as expected performance indicator information. In some optional embodiments, the aforementioned fourth indicator information can be performance indicator information calculated by the terminal's peer device (e.g., network-side device or server) based on the second target CSI information and the third reconstructed CSI information. The aforementioned second target CSI information can be target CSI information obtained by the terminal's peer device based on the first channel measurement information. The aforementioned second target CSI information can also be referred to as expected target CSI information. The aforementioned third reconstructed CSI information is reconstructed CSI information calculated by the network-side device, which can be understood as the third reconstructed CSI information being reconstructed CSI information obtained by the network-side AI unit based on the network-side device.

[0295] The performance metric indication information of the aforementioned target AI unit is used to indicate the performance metric of the target AI unit. The type information of the aforementioned performance metric of the target AI unit is used to indicate the type of performance metric of the target AI unit. It can be understood that the aforementioned second metric indication information can be used to indicate the performance metric of the target AI unit or the type of performance metric of the target AI unit in the second dataset.

[0296] The aforementioned second threshold information is used to indicate the lower limit of the expected indicator obtained based on the second dataset, or, the second threshold information is used to indicate the lower limit of the indicator that the target AI unit needs to meet, or, the second threshold information is used to indicate the lower limit of the indicator, or, the second threshold information is used to indicate the lower limit of the performance obtained based on the first dataset, or, the second threshold information is used to indicate the lower limit of the performance that the target AI unit needs to meet, or, the second threshold information is used to indicate the lower limit of the performance, etc.

[0297] The above format indication information is used to indicate the format of the target CSI information. For example, the above format indication information can be used to indicate the format of the target CSI information obtained based on the first channel measurement information.

[0298] The aforementioned channel feature information can be used to indicate the channel features associated with the second dataset, such as line-of-sight (LOS) or non-line-of-sight (NLOS), channel quality information, etc. The aforementioned scene information can be used to indicate the scene associated with the second dataset, such as network-side condition information, indoor or outdoor information, scene classification indicators, cell information, area information, etc.

[0299] In this embodiment, the second dataset is also associated with at least one of the following: the second dataset identification information, the channel measurement information indication information or the channel measurement information type information, the fourth indicator information, the second indicator indication information, the second target indicator information, the second threshold information, the format indication information, the channel feature information, and the scene information. This second dataset facilitates more accurate and convenient performance testing of the target AI unit.

[0300] In some optional embodiments, the method further includes:

[0301] The terminal receives fourth information associated with the second dataset, the fourth information including at least one of the following: the second dataset identification information, the channel measurement information indication information or the channel measurement information type information, the fourth indicator information, the second indicator indication information, the second target indicator information, the second threshold information, the format indication information, the channel feature information, and the scene information.

[0302] In some optional embodiments, the second dataset further includes fourth information, which includes at least one of the following: second dataset identification information, indication information of the channel measurement information or type information of the channel measurement information, the fourth indicator information, the second indicator indication information, the second target indicator information, the second threshold information, the format indication information, the channel feature information, and the scene information.

[0303] Optionally, the second operation further includes at least one of the following:

[0304] Report the second CSI feedback information;

[0305] The second indicator information is reported. The second indicator information is the performance indicator information of the target AI unit estimated by the terminal on the second dataset. Alternatively, the second indicator information is the performance indicator information calculated based on the first real CSI information and the fourth reconstructed CSI information. The fourth reconstructed CSI information is the reconstructed CSI information calculated by the terminal or the reconstructed CSI information calculated by the network-side device. The first real CSI information is the CSI information obtained by the terminal based on the first channel measurement information.

[0306] Obtain the second test performance information of the target AI unit;

[0307] Report the second test performance information of the target AI unit.

[0308] The second indicator information mentioned above is the performance indicator information of the target AI unit on the second dataset estimated by the terminal. For example, the terminal can estimate the performance indicator of the target AI unit based on a performance estimation algorithm or AI unit. Alternatively, the second indicator information is the performance indicator information calculated based on the first true CSI information and the fourth reconstructed CSI information, wherein the fourth reconstructed CSI information is the reconstructed CSI information calculated by the terminal, for example, the reconstructed CSI information obtained by the terminal-side AI unit and the second CSI feedback information, or the second reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, for example, the network-side device obtains the reconstructed CSI information based on the network-side AI unit and the second CSI feedback information and sends it to the terminal. The first true CSI information mentioned above is the CSI information obtained by the terminal according to the first channel measurement information, for example, the first true CSI information is the target CSI information obtained from the first channel measurement information, or the first true CSI information is the CSI information determined based on the target CSI information obtained from the first channel measurement information.

[0309] It is understood that, if the second operation further includes obtaining second test performance information of the target AI unit, the aforementioned second information does not include the second test performance information of the target AI unit. In some optional embodiments, the terminal may obtain the second test performance information of the target AI unit based on at least one of the first information, the second information, and the second dataset.

[0310] In this embodiment, the terminal acquires and / or reports at least one of the second indicator information and the second test performance information. This is beneficial for determining the overall performance of the target CSI processing and the target AI unit on the UE side, or for further determining whether there is an offset in the condition information or data distribution on the UE side, so that the peer device (such as the network-side device, server) can further determine whether the UE side is adapted to the current target AI unit.

[0311] Optionally, obtaining the second test performance information of the target AI unit includes:

[0312] The second test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or...

[0313] The second test performance information of the target AI unit is obtained based on the second CSI feedback information and the fourth CSI feedback information; or...

[0314] The second test performance information of the target AI unit is obtained based on the second indicator information; or...

[0315] The second test performance information of the target AI unit is obtained based on the second and fourth indicator information; or...

[0316] The second test performance information of the target AI unit is obtained based on the second indicator information and the second target indicator information; or...

[0317] The second test performance information of the target AI unit is obtained based on the second threshold information; or...

[0318] The second test performance information of the target AI unit is obtained based on the second indicator information and the second threshold information; or...

[0319] The second test performance information of the target AI unit is obtained based on the second indicator information, the second threshold information, and the second target indicator information.

[0320] In some implementations, the terminal can obtain the second test performance information of the target AI unit based on the first CSI feedback information and the second CSI feedback information. For example, when the first target CSI information is target CSI information obtained based on the first channel measurement information, the terminal can obtain the similarity index information, difference, or statistical error between the first CSI feedback information and the second CSI feedback information, and obtain the second test performance information of the target AI unit based on the similarity index information, difference, or statistical error. The second test performance information can reflect whether there is a performance degradation problem caused by the data processing method not matching the target AI unit.

[0321] In other implementations, the terminal can obtain the second test performance information of the target AI unit based on the second CSI feedback information and the fourth CSI feedback information. For example, the terminal can obtain the similarity index information, difference, or statistical error between the second CSI feedback information and the fourth CSI feedback information, and obtain the second test performance information of the target AI unit based on the similarity index information, difference, or statistical error. The second test performance information can reflect whether there is a performance degradation problem caused by the data processing method not matching the target AI unit.

[0322] The aforementioned fourth CSI feedback information can be understood as expected CSI feedback information, such as codebook information compressed by the AI ​​unit. In some optional embodiments, the peer device of the terminal (e.g., a network-side device or a server) can send the fourth feedback information to the terminal. In some optional embodiments, the aforementioned fourth CSI feedback information can be the target CSI information obtained by the peer device of the terminal based on the first channel measurement information and the CSI feedback information obtained by the AI ​​unit on the peer device side. For example, the target CSI information obtained by the peer device based on the first channel measurement information can be the target CSI information obtained by the peer device by processing the first channel measurement information using an estimation method. In some embodiments, the aforementioned estimation method can be understood as the processing method or processing mode by which the peer device estimates the target CSI information obtained by the terminal based on the channel measurement information.

[0323] In other implementations, the terminal can obtain the second test performance information of the target AI unit based on the second indicator information. For example, the terminal can compare the second indicator information with a threshold predefined by the protocol or a threshold configured by the network-side device or server to obtain the second test performance information of the target AI unit. Alternatively, the terminal can compare the second indicator information with the fourth indicator information to obtain the second test performance information of the target AI unit. Through this second test performance information, the overall performance of the UE side in obtaining the target CSI processing and the target AI unit can be determined. Or, it can be further determined whether there is an offset in the condition information or data distribution of the UE side, so that the peer device (such as the network-side device or server) can further determine whether there is a problem with the UE side adapting to the current target AI unit.

[0324] In some other implementations, the terminal can obtain the second test performance information of the target AI unit based on the second indicator information and the fourth indication information. For example, the terminal can calculate the difference or statistical error between the second indicator information and the fourth indicator information to obtain the second test performance information of the target AI unit. The second test performance information can be used to determine whether there are any problems with the UE side's acquisition of the target CSI processing and the target AI unit.

[0325] In some other implementations, the terminal can obtain the second test performance information of the target AI unit based on the second indicator information and the second target indicator information. For example, the terminal can calculate the difference or statistical error between the second indicator information and the second target indicator information to obtain the second test performance information of the target AI unit. The second test performance information can reflect the performance of the AI ​​unit on the second dataset, thereby determining the effectiveness or failure of the target AI unit.

[0326] In some other implementations, the terminal can obtain the second test performance information of the target AI unit based on the second threshold information. For example, the terminal can compare the similarity index, difference, or statistical error between the first CSI feedback information and the second CSI feedback information with the second threshold information to obtain the second test performance information of the target AI unit; or, it can compare the difference or statistical error between the second index information and the fourth index information with the second threshold information to obtain the second test performance information of the target AI unit; or, it can compare the second index information with the second threshold information to obtain the second test performance information of the target AI unit, etc. This second test performance information can reflect the performance of the AI ​​unit on the second dataset, thereby determining the effectiveness or failure of the target AI unit.

[0327] In some other implementations, the terminal can obtain the second test performance information of the target AI unit based on the second indicator information, the second threshold information, and the second target indicator information. For example, the terminal can calculate the difference or statistical error between the second indicator information and the second target indicator information, and compare the difference or statistical error with the second threshold information to obtain the second test performance information of the target AI unit. This second test performance information can reflect the performance of the AI ​​unit on the second dataset, thereby determining the effectiveness or failure of the target AI unit.

[0328] In another embodiment, the terminal can obtain second test performance information of the target AI unit based on the second indicator information and the second threshold information. For example, the terminal can compare the second indicator information and the second threshold information to obtain the second test performance information of the target AI unit. This second test performance information can reflect the performance of the AI ​​unit on a second dataset, thereby determining the effectiveness or failure of the target AI unit.

[0329] Optionally, the second dataset includes P data units, each of which includes first channel measurement information, where P is an integer greater than 1.

[0330] In some alternative embodiments, P can be 20, 50, or 100, etc. Exemplarily, each of the above data units may include data from at least one time unit, which may include a subframe, a time slot, or a sub-time slot, etc.

[0331] Optionally, the second indicator information is determined based on the performance indicator information corresponding to Q data units, where Q is an integer less than or equal to P;

[0332] The performance metric information corresponding to each data unit is determined according to at least one of the following:

[0333] The second CSI feedback information and the fourth CSI feedback information corresponding to each data unit;

[0334] The first real CSI information and the fourth reconstructed CSI information corresponding to each data unit, wherein the first real CSI information corresponding to each data unit is CSI information obtained based on the first channel measurement information of each data unit;

[0335] Estimated performance metrics information corresponding to each data unit.

[0336] The aforementioned fourth CSI feedback information can be found in the relevant descriptions of the foregoing embodiments, and will not be repeated here.

[0337] In this embodiment, the aforementioned Q data units are data units among the aforementioned P data units. The first true CSI information corresponding to each data unit can be understood as the true CSI information obtained based on the first channel measurement information of that data unit; the fourth reconstructed CSI information corresponding to each data unit can be understood as the reconstructed CSI information obtained based on the second CSI feedback information obtained from the first channel measurement information of that data unit. The estimated performance index information corresponding to each data unit can be understood as the performance index information of that data unit estimated by the terminal.

[0338] The aforementioned second indicator information is determined based on the indicator information corresponding to the Q data units. For example, the aforementioned second indicator information may be the maximum value, minimum value, median value, average value, or statistical value obtained based on a preset statistical algorithm for the indicator information corresponding to the Q data units. This embodiment does not limit this.

[0339] In this embodiment, the second indicator information is determined based on the performance indicator information corresponding to the Q data units, which helps to ensure the reliability of the calculated second indicator information.

[0340] Optionally, the first indicator information is determined based on the performance indicator information corresponding to the N data unit groups, where N is an integer less than or equal to P;

[0341] The performance index information corresponding to each data unit group is the performance index information determined based on all data units in each data unit group.

[0342] In this embodiment, the data units in the aforementioned N data unit groups are the data units in the aforementioned P data units. Each data unit group may include at least two data units from the aforementioned P data units, and the performance indicator information corresponding to each data unit group can be determined based on some or all of the data units in that data unit group. For example, one performance indicator can be estimated for every 6 data units, and P / 6 performance indicator information can be estimated from a dataset. Then, the second indicator information can be determined by combining the P / 6 performance indicator information.

[0343] In this embodiment, the second indicator information is determined based on the performance indicator information corresponding to the N data unit groups, which helps to ensure the reliability of the calculated second indicator information.

[0344] Optionally, the performance metric information includes information obtained based on at least one of the following: Squared Generalized Cosine Similarity (SGCS) and Minimum Mean Squared Error (MMSE).

[0345] For example, the performance metric information may include at least one of SGCS and MMSE, or the performance metric information may be information determined according to at least one of SGCS and MMSE.

[0346] Optionally, the second operation further includes:

[0347] The first true CSI information is reported. The first true CSI information is the CSI information obtained by the terminal based on the first channel measurement information. The first true CSI information is used to obtain the second test performance information of the target AI unit.

[0348] The acquisition of the aforementioned ground-truth CSI information can be found in the relevant descriptions of the foregoing embodiments, and will not be repeated here.

[0349] The terminal reports the first true CSI information, allowing the peer device (e.g., a network-side device or server) to conveniently obtain the second test performance information of the target AI unit based on this information. For example, the terminal can report the first true CSI information to the network-side device, enabling the network-side device to obtain the second test performance information of the target AI unit based on the first true CSI information and the reconstructed CSI information obtained from the second CSI feedback information.

[0350] Optionally, the second operation further includes:

[0351] Receive first indication information, which is used to instruct the terminal to report the first real CSI information.

[0352] For example, the network-side device or server can send a first indication message to the terminal, and upon receiving the first indication message, the terminal can report the first real CSI information to the network-side device or server.

[0353] Optionally, the first indication information includes at least one of the following:

[0354] Orthogonal beam vector number;

[0355] Indication information for Type II parameter combinations;

[0356] Indication information of the angle-delay field codebook;

[0357] The codebook type information of the first real CSI information.

[0358] In this embodiment, the terminal can report the first real CSI information based on the first indication information, which helps to ensure that the reported real CSI information is more in line with the requirements.

[0359] Optionally, the first true CSI information includes the following: a codebook matrix, a compressed codebook matrix.

[0360] The codebook matrix mentioned above, for example, W N*B , where N is the number of ports, B is the number of frequency domain units, or the codebook matrix in the angle-delay domain.

[0361] The compressed codebook matrix mentioned above includes, for example, Type II codebook information such as weighted amplitude coefficients, phase information, and orthogonal beam vectors.

[0362] Optionally, the second dataset is used to test at least one of the performance of the target AI unit and the performance of the first processing unit, wherein the first processing unit is used by the terminal to process channel measurement information to obtain target CSI information.

[0363] In this embodiment, when the terminal obtains channel measurement information, it first processes the channel measurement information based on the first processing unit to obtain target CSI information, and then processes the obtained target CSI information based on the target AI unit to obtain CSI feedback information. It should be noted that the processing method of the channel measurement information by the first processing unit is the same as the aforementioned data processing method. Since the second dataset includes the first channel measurement information, at least one of the performance of the target AI unit and the performance of the first processing unit can be tested based on the second dataset.

[0364] Optionally, the method further includes:

[0365] If the first test performance information of the target AI unit meets the first performance requirement, but the second test performance information of the target AI unit does not meet the second performance requirement, the terminal performs at least one of the following:

[0366] Collect a third dataset;

[0367] Report the third dataset;

[0368] Send a second instruction message, which is used to instruct the collection of a third dataset;

[0369] The third dataset includes data from the terminal or data from a first type of terminal, where the first type refers to the type of terminal.

[0370] For example, the first test performance information of the target AI unit meeting the first performance requirement may include the first test performance information of the target AI unit indicating that the target AI unit is effective under the first dataset. The second test performance information of the target AI unit not meeting the second performance requirement may include the second test performance information of the target AI unit indicating that the target AI unit is ineffective or its effectiveness is reduced under the second dataset (e.g., the reduction in effectiveness exceeds a preset value).

[0371] It is understood that if the first test performance information of the target AI unit meets the first performance requirement, but the second test performance information of the target AI unit does not meet the second performance requirement, it may be because the relevant data of the terminal or the terminal type (i.e., the first type) is not included in the data used to train the target AI unit, resulting in a decrease or failure of the target AI unit's performance. In this case, the relevant data of the terminal or the terminal type (i.e., the third dataset) can be collected to update the target AI unit.

[0372] It is understood that if the first test performance information of the target AI unit meets the first performance requirement, and the second test performance information of the target AI unit does not meet the second performance requirement, it indicates that the target AI unit is effective. However, the processing of the first processing unit of the terminal or the condition information of the terminal does not match the data of the model obtained by the network side.

[0373] In some embodiments, if the target AI unit is trained by the terminal, the terminal can collect a third dataset and update the target AI unit based on the third dataset, which is beneficial to improving the performance of the target AI unit.

[0374] In other embodiments, if the target AI unit is trained by a network-side device or server, the terminal can report a third dataset for the network-side device or server to update the target AI unit. Alternatively, the terminal can send a second instruction message to instruct the network-side device or server to collect the third dataset and update the target AI unit, which is beneficial to improving the performance of the target AI unit.

[0375] In other embodiments, if the target AI unit is trained by a network-side device or server, the terminal can report a third dataset and the reason for the failure of the target AI unit so that the network-side device or server can update the target AI unit.

[0376] Optionally, the third dataset includes at least one of the following:

[0377] The terminal or the terminal of the first type acquires target CSI information based on CSI measurements;

[0378] Data type identification information, or terminal type identification information, or terminal identification information;

[0379] Data feature information based on target CSI information obtained from CSI measurements;

[0380] The CSI measurement is associated with at least one of the channel characteristic information and scene information.

[0381] The identification information for the aforementioned data types can be used to indicate the data type of the third dataset. The identification information for the aforementioned terminal types can be used to indicate the terminal type associated with the third dataset. The identification information for the aforementioned terminals can be used to indicate the terminal associated with the third dataset.

[0382] The aforementioned data characteristics may include, but are not limited to, the variance and mean of the data.

[0383] The aforementioned channel characteristic information, such as LOS or NLOS, channel quality information, etc. The aforementioned scenario information, such as network-side condition information, indoor or outdoor information, scenario classification indicators, cell information, area information, etc.

[0384] Optionally, the method further includes at least one of the following:

[0385] The terminal receives at least one of an updated first dataset and an updated second dataset;

[0386] The terminal updates the target AI unit based on at least one of the updated first dataset and the updated second dataset;

[0387] The terminal updates the target AI unit based on the third dataset;

[0388] The terminal receives the updated target AI unit.

[0389] The following examples illustrate this embodiment:

[0390] In some embodiments, the terminal can receive an updated second dataset and an updated target AI unit, and then test the updated target AI unit based on the updated second dataset, which helps to ensure the performance of the target AI unit.

[0391] In other embodiments, the terminal may receive at least one of an updated first dataset and an updated second dataset, and update the target AI unit based on at least one of the updated first dataset and the updated second dataset, which helps to ensure the performance of the target AI unit.

[0392] In some other embodiments, the terminal can update the target AI unit based on a third dataset and test the updated target AI unit based on the updated second dataset, which helps to ensure the performance of the target AI unit.

[0393] Optionally, the method further includes:

[0394] The terminal performs a third operation, which includes:

[0395] Perform CSI measurements to obtain CSI information for the third target;

[0396] The fifth CSI feedback information is obtained based on the third target CSI information and the target AI unit;

[0397] The fifth CSI feedback information is reported, which is used to obtain the second monitoring result information of the target AI unit and the channel information of the terminal for the network-side device to determine.

[0398] In this embodiment, the terminal can obtain second channel measurement information through CSI measurement, and can process the second channel measurement information to obtain third target CSI information. The third target CSI information is then processed by the target AI unit to obtain and report fifth CSI feedback information. Subsequently, the network-side device can determine the terminal's channel information and obtain the second monitoring result information of the target AI unit based on the CSI feedback information. For example, the network-side device can obtain reconstructed CSI information based on the fifth CSI feedback information, and obtain the second monitoring result information of the target AI unit based on the obtained reconstructed CSI information and the second real CSI information. The second real CSI information can be the CSI information obtained based on the third target CSI information.

[0399] This embodiment can monitor the measured performance of the target AI unit through the third operation, thereby further ensuring the performance of CSI processing based on the target AI unit.

[0400] Optionally, the third operation further includes at least one of the following:

[0401] The second real CSI information is reported. The second real CSI information is the CSI information obtained by the terminal based on the third target CSI information. The second real CSI information is used to obtain the second monitoring result information of the target AI unit.

[0402] Obtain the second monitoring result information of the target AI unit.

[0403] In some implementations, the terminal can report ground truth CSI information, which facilitates network-side devices to obtain the second monitoring result information of the target AI unit based on the ground truth CSI information.

[0404] In other embodiments, the terminal can obtain the second monitoring result information of the target AI unit based on the second real CSI information and the reconstructed CSI information obtained based on the fifth CSI feedback information.

[0405] Optionally, the method further includes:

[0406] If the first test performance information of the target AI unit meets the first performance requirement, and the second test performance information of the target AI unit meets the second performance requirement, then if the second monitoring result information of the target AI unit does not meet the third performance requirement, the terminal executes at least one of the following:

[0407] Collect the fourth dataset;

[0408] Report the fourth dataset;

[0409] Send a third instruction message, which is used to instruct the collection of a fourth dataset;

[0410] The fourth dataset includes data under the first scenario or data under the first data distribution, where the first scenario is the scenario corresponding to the fifth CSI feedback information, and the first data distribution is the data distribution corresponding to the fifth CSI feedback information.

[0411] For example, the first test performance information of the target AI unit satisfying the first performance requirement may include the first test performance information of the target AI unit indicating that the target AI unit is effective under the first dataset. The second test performance information of the target AI unit satisfying the second performance requirement may include the second test performance information of the target AI unit indicating that the target AI unit is effective under the second dataset. The second monitoring result information of the target AI unit not satisfying the third performance requirement may include the second monitoring result information of the target AI unit indicating that the target AI unit has failed or its effectiveness has decreased, etc.

[0412] It is understood that if the first test performance information of the target AI unit meets the first performance requirement, the second test performance information of the target AI unit meets the second performance requirement, and the second monitoring result information of the target AI unit does not meet the third performance requirement, it indicates that the performance degradation of the target AI unit may be due to inconsistent data distribution or inconsistent channel conditions and scene distribution. In this case, data under the first scene or data under the first data distribution can be collected to update the target AI unit.

[0413] In some embodiments, if the target AI unit is trained by the terminal, the terminal can collect a fourth dataset and update the target AI unit based on the fourth dataset, which is beneficial to improving the performance of the target AI unit.

[0414] In other embodiments, if the target AI unit is trained by a network-side device or server, the terminal can report a fourth dataset for the network-side device or server to update the target AI unit. Alternatively, the terminal can send a third instruction message to instruct the network-side device or server to collect the fourth dataset and update the target AI unit, which is beneficial to improving the performance of the target AI unit.

[0415] Optionally, the fourth dataset includes at least one of the following:

[0416] The target CSI information obtained by the terminal in the first scenario or the terminal under the first data distribution based on CSI measurement;

[0417] Scene information;

[0418] Data distribution information.

[0419] The aforementioned scenario information can be used to indicate the first scenario. The aforementioned data distribution information can be used to indicate the first data distribution. For example, the aforementioned data distribution information may include at least one of data distribution identification information, data feature information, etc. The aforementioned data feature information may include, but is not limited to, the variance of the data, the mean of the data, etc.

[0420] Optionally, the terminal performs a third operation, including:

[0421] If the first test performance information of the target AI unit meets the first performance requirement, the terminal performs a third operation;

[0422] or,

[0423] If the first test performance information of the target AI unit meets the first performance requirement and the second test performance information of the target AI unit meets the second performance requirement, the terminal performs the third operation.

[0424] In some embodiments, if the first test performance information of the target AI unit meets the first performance requirement, the terminal performs a third operation. For example, the terminal first performs the first operation to test the validity of the target AI unit under the first dataset and determine the deployment performance of the target AI unit. If the target AI unit is valid under the first dataset, the third operation can be performed to monitor the actual test performance of the target AI unit. If the actual test performance of the target AI unit meets the corresponding performance requirement, the second operation is performed to test the validity of the target AI unit under the second dataset and determine whether the terminal-side data processing method matches the target AI unit. If the target AI unit fails or its performance degrades under the second dataset, it can be determined that the terminal-side data processing method does not match the target AI unit.

[0425] In other embodiments, if the first test performance information of the target AI unit meets the first performance requirement and the second test performance information of the target AI unit meets the second performance requirement, the terminal performs a third operation. That is, the terminal first performs the first and second operations to test the effectiveness of the target AI unit under the first dataset and the second dataset. If the target AI unit is effective under both the first and second datasets, the third operation is performed to monitor the actual test performance of the target AI unit. If the actual test performance of the target AI unit decreases, it can be determined that the performance degradation is caused by inconsistent data distribution or inconsistent channel conditions and scene distribution.

[0426] For example, the terminal may first perform a first operation to test the effectiveness of the target AI unit under the first dataset; if the target AI unit is effective under the first dataset, perform a second operation to test the effectiveness of the target AI unit under the second dataset; if the target AI unit is effective under the second dataset, perform a third operation to monitor the measured performance of the target AI unit.

[0427] Optionally, the terminal performs at least one of the first operation and the second operation, including:

[0428] If the second monitoring result information of the target AI unit does not meet the third performance requirement, the terminal performs at least one of the first operation and the second operation.

[0429] In this embodiment, the terminal may first perform a third operation to monitor the measured performance of the target AI unit. If the measured performance of the target AI unit does not meet the corresponding performance requirements, at least one of the first and second operations may be performed. For example, the terminal may first perform a third operation to monitor the measured performance of the target AI unit; if the measured performance of the target AI unit does not meet the corresponding performance requirements, a first operation may be performed to test the effectiveness of the target AI unit under the first dataset; if the target AI unit is effective under the first dataset, a second operation may be performed to test the effectiveness of the target AI unit under the second dataset.

[0430] Optionally, the first dataset is predefined by the protocol, or the first dataset is configured or sent by the first device;

[0431] And / or,

[0432] The second dataset is predefined by the protocol, or the second dataset is configured or sent by the first device.

[0433] Optionally, the target AI unit includes at least one of the following:

[0434] The terminal is a first AI unit used to acquire CSI feedback information;

[0435] The network-side device is used as a second AI unit to acquire reconstructed CSI information;

[0436] The terminal is a reference AI unit used to acquire CSI feedback information;

[0437] Network-side devices are used to acquire reference AI units for reconstructing CSI information;

[0438] The AI ​​unit used in the test by the terminal;

[0439] AI units used in network-side equipment testing;

[0440] The AI ​​unit used in the test equipment;

[0441] The terminal is used to match the AI ​​unit of the AI ​​unit used in the test;

[0442] Network-side devices are used to match AI units used in testing;

[0443] The testing equipment is used to match the AI ​​units used in the test.

[0444] Please refer to Figure 5, which is a flowchart of an AI unit processing method provided in an embodiment of this application. The method can be executed by a first device, and as shown in Figure 5, it includes the following steps:

[0445] Step 501: The first device performs at least one of the fourth and fifth operations;

[0446] The fourth operation includes at least one of the following:

[0447] Send a first dataset, which includes first target CSI information;

[0448] Send a second dataset, which includes the first channel measurement information;

[0449] The fifth operation includes at least one of the following: receiving fifth information, receiving sixth information;

[0450] The fifth piece of information includes at least one of the following:

[0451] First CSI feedback information;

[0452] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0453] The first test performance information of the target AI unit;

[0454] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0455] The sixth piece of information includes at least one of the following:

[0456] Second CSI feedback information;

[0457] The first monitoring result information of the target AI unit;

[0458] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0459] The second test performance information of the target AI unit;

[0460] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0461] In this embodiment, the first device may include a network-side device or a server. The fifth information may include some or all of the first information, and the sixth information may include some or all of the second information.

[0462] In some alternative embodiments, the first device performs the fourth operation and the second device performs the fifth operation, wherein one of the first device and the second device is a network-side device and the other is a server. For example, the server performs the fourth operation and the network-side device performs the fifth operation, or the network-side device performs the fourth operation and the server performs the fifth operation.

[0463] Optionally, the first dataset further includes at least one of the following:

[0464] First dataset identification information;

[0465] The indication information or type information of the target CSI information;

[0466] Third CSI feedback information.

[0467] Optionally, the first dataset is associated with third information, which includes at least one of the following:

[0468] The third indicator information is a performance indicator information calculated based on the first target CSI information and the first reconstructed CSI information, wherein the first reconstructed CSI information is the reconstructed CSI information calculated by the network-side device.

[0469] The first indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit;

[0470] The first target indicator information is the expected performance indicator information of the target AI unit;

[0471] First threshold information.

[0472] Optionally, the second dataset is associated with fourth information, which includes at least one of the following:

[0473] Second dataset identification information;

[0474] Indication information or type information of channel measurement information;

[0475] The fourth indicator information is a performance indicator information calculated based on the second target CSI information and the third reconstructed CSI information. The third reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, and the second target CSI information is the target CSI information obtained by the network-side device based on the first channel measurement information.

[0476] The second indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit;

[0477] The second target indicator information is the expected performance indicator information of the target AI unit;

[0478] Second threshold information;

[0479] Format indication information, wherein the format indication information is used to indicate the format of the target CSI information;

[0480] Channel characteristic information;

[0481] Scene information.

[0482] Optionally, the fifth operation further includes at least one of the following:

[0483] Obtain the first test performance information of the target AI unit;

[0484] Obtain the second test performance information of the target AI unit.

[0485] In some optional embodiments, the first device may obtain first test performance information of the target AI unit based on at least one of the fifth information, the sixth information, and the first dataset; and may obtain second test performance information of the target AI unit based on at least one of the fifth information, the sixth information, and the second dataset.

[0486] Optionally, obtaining the first test performance information of the target AI unit includes:

[0487] The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or...

[0488] The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the third CSI feedback information; or...

[0489] The first test performance information of the target AI unit is obtained based on the fifth indicator information; or...

[0490] The first test performance information of the target AI unit is obtained based on the fifth and third indicator information; or...

[0491] The first test performance information of the target AI unit is obtained based on the fifth indicator information and the first target indicator information; or...

[0492] The first test performance information of the target AI unit is obtained based on the first threshold information; or...

[0493] The first test performance information of the target AI unit is obtained based on the fifth indicator information and the first threshold information; or...

[0494] The first test performance information of the target AI unit is obtained based on the fifth indicator information, the first threshold information, and the first target indicator information.

[0495] The fifth indicator information includes either the first indicator information or the sixth indicator information. The sixth indicator information is the performance indicator information of the target AI unit on the first dataset estimated by the first device, or the sixth indicator information is the performance indicator information calculated based on the first target CSI information and the fifth reconstructed CSI information. The fifth reconstructed CSI information is the reconstructed CSI information calculated by the network-side device.

[0496] The sixth indicator information mentioned above is the performance indicator information of the target AI unit on the first dataset estimated by the first device. For example, the first device can estimate the performance indicator information of the target AI unit based on a performance estimation algorithm or an AI unit. The fifth reconstructed CSI information mentioned above is the reconstructed CSI information calculated by the network-side device, which can be understood as the fifth reconstructed CSI information being the reconstructed CSI information obtained by the network-side device based on the network-side AI unit.

[0497] In some implementations, the first device can obtain the first test performance information of the target AI unit based on the first CSI feedback information and the second CSI feedback information. For example, when the first target CSI information is target CSI information obtained based on the first channel measurement information, the first device can obtain the similarity index information, difference, or statistical error between the first CSI feedback information and the second CSI feedback information, and obtain the first test performance information of the target AI unit based on the similarity index information, difference, or statistical error. The first test performance information can reflect whether there is a performance degradation problem caused by the data processing method not matching the target AI unit.

[0498] In other embodiments, the first device can obtain the first test performance information of the target AI unit based on the first CSI feedback information and the third CSI feedback information. For example, the first device can obtain the similarity index information, difference or statistical error between the first CSI feedback information and the third CSI feedback information, and obtain the first test performance information of the target AI unit based on the similarity index information, difference or statistical error. The first test performance information can reflect whether there are problems in the deployment of the AI ​​unit.

[0499] In some other implementations, the first device can obtain the first test performance information of the target AI unit based on the fifth indicator information. For example, the first device can compare the fifth indicator information with a threshold predefined in the protocol or a threshold configured by the network-side device or server to obtain the first test performance information of the target AI unit. Alternatively, the first device can compare the fifth indicator information with the third indicator information to obtain the first test performance information of the target AI unit. This first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the validity or failure of the target AI unit.

[0500] In some other implementations, the first device can obtain the first test performance information of the target AI unit based on the fifth indicator information and the third indication information. For example, the first device can calculate the difference or statistical error between the fifth indicator information and the third indicator information to obtain the first test performance information of the target AI unit. The first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the effectiveness or failure of the target AI unit.

[0501] In some other implementations, the first device can obtain the first test performance information of the target AI unit based on the fifth indicator information and the first target indicator information. For example, the first device can calculate the difference or statistical error between the fifth indicator information and the first target indicator information to obtain the first test performance information of the target AI unit. The first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the effectiveness or failure of the target AI unit.

[0502] In some other embodiments, the first device can obtain the first test performance information of the target AI unit based on the first threshold information. For example, the first device can compare the similarity index, difference, or statistical error between the first CSI feedback information and the third CSI feedback information with the first threshold information to obtain the first test performance information of the target AI unit; or, it can compare the similarity index, difference, or statistical error between the first CSI feedback information and the second CSI feedback information with the first threshold information to obtain the first test performance information of the target AI unit; or, it can compare the difference or statistical error between the fifth index information and the third index information with the first threshold information to obtain the first test performance information of the target AI unit; or, it can compare the fifth index information with the first threshold information to obtain the first test performance information of the target AI unit, etc. This first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the effectiveness or failure of the target AI unit.

[0503] In some other implementations, the first device can obtain the first test performance information of the target AI unit based on the fifth indicator information, the first threshold information, and the first target indicator information. For example, the first device can calculate the difference or statistical error between the fifth indicator information and the first target indicator information, and compare the difference or statistical error between the fifth indicator information and the first target indicator information with the first threshold information to obtain the first test performance information of the target AI unit. This first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the effectiveness or failure of the target AI unit.

[0504] In some other embodiments, the first device can obtain the first test performance information of the target AI unit based on the fifth indicator information and the first threshold information. For example, the first device can compare the fifth indicator information and the first threshold information to obtain the first test performance information of the target AI unit. This first test performance information can reflect the performance of the AI ​​unit on the first dataset, thereby determining the effectiveness or failure of the target AI unit.

[0505] Optionally, obtaining the second test performance information of the target AI unit includes:

[0506] The second test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or...

[0507] The second test performance information of the target AI unit is obtained based on the second CSI feedback information and the fourth CSI feedback information; or...

[0508] The second test performance information of the target AI unit is obtained based on the seventh indicator information; or...

[0509] The second test performance information of the target AI unit is obtained based on the seventh and fourth indicator information; or...

[0510] The second test performance information of the target AI unit is obtained based on the seventh indicator information and the second target indicator information; or...

[0511] The second test performance information of the target AI unit is obtained based on the second threshold information; or...

[0512] The second test performance information of the target AI unit is obtained based on the seventh indicator information and the second threshold information; or...

[0513] The second test performance information of the target AI unit is obtained based on the seventh indicator information, the second threshold information, and the second target indicator information.

[0514] The seventh indicator information includes either the second indicator information or the eighth indicator information. The eighth indicator information is the performance indicator information of the target AI unit on the second dataset estimated by the first device, or the eighth indicator information is the performance indicator information calculated based on the first real CSI information and the sixth reconstructed CSI information, or the eighth indicator information is the performance indicator information calculated based on the fourth target CSI information and the sixth reconstructed CSI information. The sixth reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, and the fourth target CSI information is the target CSI information estimated based on the first channel measurement information.

[0515] The aforementioned fourth target CSI information can be the target CSI information obtained by the first device processing the first channel measurement information using an estimation method. In some embodiments, the aforementioned estimation method can be understood as a processing method or approach by which the first device estimates the target CSI information obtained by the terminal based on the channel measurement information.

[0516] As shown in Figure 7, the terminal measures the CSI-RS to obtain CSI-RS related side information, such as (first channel measurement information). Optionally, UE-side processing can be performed on the measured channel, such as channel autocorrelation, SVD decomposition, phase normalization, etc., to obtain the target CSI information or the true CSI information. It is understood that different UEs may adopt different UE-side processing methods; for example, the methods for channel autocorrelation, SVD decomposition, and phase normalization may differ. Therefore, the aforementioned fourth target CSI information can be one or more target CSI information obtained by the first device using various different estimation methods to process the first channel measurement information. In some embodiments, the above estimation method can be understood as one or more processing methods or approaches by which the first device estimates the terminal's acquisition of target CSI information based on channel measurement information.

[0517] In some implementations, the first device can obtain the second test performance information of the target AI unit based on the first CSI feedback information and the second CSI feedback information. For example, when the first target CSI information is target CSI information obtained based on the first channel measurement information, the first device can obtain the similarity index information, difference, or statistical error between the first CSI feedback information and the second CSI feedback information, and obtain the second test performance information of the target AI unit based on the similarity index information, difference, or statistical error. The second test performance information can reflect whether there is a performance degradation problem caused by the data processing method not matching the target AI unit.

[0518] In other embodiments, the first device can obtain the second test performance information of the target AI unit based on the second CSI feedback information and the fourth CSI feedback information. For example, the first device can obtain the similarity index information, difference, or statistical error between the second CSI feedback information and the fourth CSI feedback information, and obtain the second test performance information of the target AI unit based on the similarity index information, difference, or statistical error. The second test performance information can reflect whether there is a performance degradation problem caused by the data processing method not matching the target AI unit.

[0519] The aforementioned fourth CSI feedback information can be understood as expected CSI feedback information, such as codebook information compressed by the AI ​​unit. In some optional embodiments, the peer device of the terminal (e.g., a network-side device or a server) can send the fourth feedback information to the terminal. In some optional embodiments, the aforementioned fourth CSI feedback information can be the second target CSI information obtained by the peer device of the terminal based on the first channel measurement information and the CSI feedback information obtained by the AI ​​unit on the peer device side. For example, the aforementioned second target CSI information can be the target CSI information obtained by the peer device by processing the first channel measurement information using an estimation method. In some embodiments, the aforementioned estimation method can be understood as the processing method or processing mode by which the peer device estimates the target CSI information obtained by the terminal based on the channel measurement information.

[0520] In other implementations, the first device can obtain the second test performance information of the target AI unit based on the seventh indicator information. For example, the first device can compare the seventh indicator information with a threshold predefined by the protocol or a threshold configured by the network-side device or server to obtain the second test performance information of the target AI unit. Alternatively, the first device can compare the seventh indicator information with the fourth indicator information to obtain the second test performance information of the target AI unit. The second test performance information can be used to determine the overall performance of the target CSI processing and the target AI unit on the UE side, or to further determine whether there is an offset in the condition information or data distribution on the UE side, so that the peer device (such as the network-side device or server) can further determine whether there is a problem with the UE side adapting to the current target AI unit.

[0521] In some other implementations, the first device can obtain the second test performance information of the target AI unit based on the seventh indicator information and the fourth indication information. For example, the first device can calculate the difference or statistical error between the seventh indicator information and the fourth indicator information to obtain the second test performance information of the target AI unit. The second test performance information can be used to determine whether there are any problems with the UE side's acquisition of the target CSI processing and the target AI unit.

[0522] In some other implementations, the first device can obtain the second test performance information of the target AI unit based on the seventh indicator information and the second target indicator information. For example, the first device can calculate the difference or statistical error between the seventh indicator information and the second target indicator information to obtain the second test performance information of the target AI unit. The second test performance information can reflect the performance of the AI ​​unit on the second dataset, thereby determining the effectiveness or failure of the target AI unit.

[0523] In some other embodiments, the first device can obtain the second test performance information of the target AI unit based on the second threshold information. For example, the first device can compare the similarity index information, difference, or statistical error between the first CSI feedback information and the second CSI feedback information with the second threshold information to obtain the second test performance information of the target AI unit. Alternatively, the first device can compare the difference or statistical error between the seventh index information and the fourth index information with the second threshold information to obtain the second test performance information of the target AI unit. Or, the first device can compare the seventh index information with the second threshold information to obtain the second test performance information of the target AI unit, etc.

[0524] In some other implementations, the first device can obtain the second test performance information of the target AI unit based on the seventh indicator information, the second threshold information, and the second target indicator information. For example, the first device can calculate the difference or statistical error between the seventh indicator information and the second target indicator information, and compare the difference or statistical error with the second threshold information to obtain the second test performance information of the target AI unit. This second test performance information can reflect the performance of the AI ​​unit on the second dataset, thereby determining the effectiveness or failure of the target AI unit.

[0525] In another embodiment, the first device can obtain the second test performance information of the target AI unit based on the seventh indicator information and the second threshold information. For example, the first device can obtain the second test performance information of the target AI unit by comparing the seventh indicator information and the second threshold information.

[0526] Optionally, the fifth operation further includes:

[0527] Receive the first real CSI information.

[0528] Optionally, the fifth operation further includes:

[0529] Send a first instruction message, which is used to instruct the terminal to report the first real CSI information.

[0530] Optionally, the first indication information includes at least one of the following:

[0531] Orthogonal beam vector number;

[0532] Indication information for Type II parameter combinations;

[0533] Indication information of the angle delay domain codebook;

[0534] The codebook type information of the first real CSI information.

[0535] Optionally, the fourth target CSI information includes at least two target CSI information, wherein the at least two target CSI information are target CSI information obtained by using at least two estimation methods;

[0536] The eighth indicator information is determined based on at least two performance indicator information, and each of the at least two performance indicator information is determined based on each of the at least two target CSI information and the sixth reconstructed CSI information.

[0537] The estimation method described above is used to obtain target CSI information based on channel measurement information. In some embodiments, the estimation method described above can be understood as a processing method or approach by which the first device estimates the terminal's acquisition of target CSI information based on channel measurement information, such as the processing method used by the first processing unit described above.

[0538] For example, the first device can use multiple estimation methods to process the first channel measurement information to obtain at least two target CSI information, and can obtain at least two performance index information based on the at least two target CSI information and the sixth reconstructed CSI information, and then determine the eighth index information based on the at least two performance index information. For example, the eighth index information can be the minimum, maximum or average value of the at least two performance index information.

[0539] For example, the first device can use multiple estimation methods to process the first channel measurement information to obtain at least two target CSI information, and can obtain at least two performance index information based on the at least two target CSI information and the sixth reconstructed CSI information. Then, it can determine the eighth index information based on the at least two performance index information and the second threshold. For example, the eighth index information can be the minimum, maximum, or average value of the at least two performance index information.

[0540] Optionally, the first dataset includes K data units, each of which includes first target CSI information, where K is an integer greater than 1;

[0541] And / or,

[0542] The second dataset includes M data units, each of which includes first channel measurement information, where M is an integer greater than 1.

[0543] Optionally, the first indicator information is determined based on the performance indicator information corresponding to L data units, where L is an integer less than or equal to K;

[0544] The performance metric information corresponding to each data unit is determined according to at least one of the following:

[0545] The third CSI feedback information and the corresponding first CSI feedback information for each data unit;

[0546] The first target CSI information and the corresponding second reconstructed CSI information for each data unit;

[0547] Estimated performance metrics information corresponding to each data unit.

[0548] Optionally, the sixth indicator information is determined based on the performance indicator information corresponding to I data units, where I is an integer less than or equal to K;

[0549] The performance metric information corresponding to each data unit is determined according to at least one of the following:

[0550] The third CSI feedback information and the corresponding first CSI feedback information for each data unit;

[0551] The first target CSI information and the corresponding fifth reconstructed CSI information for each data unit;

[0552] Estimated performance metrics information corresponding to each data unit.

[0553] The aforementioned I data units are data units within the aforementioned K data units. The estimated performance index information corresponding to each data unit can be understood as the performance index information of that data unit estimated by the first device.

[0554] The sixth indicator information mentioned above is determined based on the indicator information corresponding to I data units. For example, the sixth indicator information can be the maximum value, minimum value, median value, average value, or statistical value obtained based on a preset statistical algorithm for the indicator information corresponding to I data units. This embodiment does not limit this.

[0555] In this embodiment, the sixth indicator information is determined based on the performance indicator information corresponding to I data units, which helps to ensure the reliability of the calculated sixth indicator information.

[0556] Optionally, the first indicator information is determined based on the performance indicator information corresponding to the M data unit groups, where M is an integer less than or equal to K;

[0557] The performance index information corresponding to each data unit group is the performance index information determined based on at least two or all data units in each data unit group.

[0558] Optionally, the sixth indicator information is determined based on the performance indicator information corresponding to the L data unit groups, where L is an integer less than or equal to K;

[0559] The performance index information corresponding to each data unit group is the performance index information determined based on at least two or all data units in each data unit group.

[0560] In this embodiment, the data units in the aforementioned L data unit groups are the data units in the aforementioned K data units. Each data unit group may include at least two data units from the aforementioned K data units, and the performance indicator information corresponding to each data unit group can be determined based on some or all of the data units in that data unit group. For example, one performance indicator can be estimated for every four data units, and K / 4 performance indicator information can be estimated from a dataset. Then, the sixth indicator information can be determined by combining the K / 4 performance indicator information.

[0561] In this embodiment, the sixth indicator information is determined based on the performance indicator information corresponding to the L data unit groups, which helps to ensure the reliability of the calculated sixth indicator information.

[0562] Optionally, the second indicator information is determined based on the performance indicator information corresponding to Q data units, where Q is an integer less than or equal to P;

[0563] The performance metric information corresponding to each data unit is determined according to at least one of the following:

[0564] The second CSI feedback information and the fourth CSI feedback information corresponding to each data unit;

[0565] The first real CSI information and the fourth reconstructed CSI information corresponding to each data unit, wherein the first real CSI information corresponding to each data unit is CSI information obtained based on the first channel measurement information of each data unit;

[0566] Estimated performance metrics information corresponding to each data unit.

[0567] Optionally, the eighth indicator information is determined based on the performance indicator information corresponding to J data units, where J is an integer less than or equal to P;

[0568] The performance metric information corresponding to each data unit is determined according to at least one of the following:

[0569] The second CSI feedback information and the fourth CSI feedback information corresponding to each data unit;

[0570] The first real CSI information and the sixth reconstructed CSI information corresponding to each data unit, wherein the first real CSI information corresponding to each data unit is CSI information obtained based on the first channel measurement information of each data unit;

[0571] Estimated performance metrics information corresponding to each data unit.

[0572] In this embodiment, the aforementioned J data units are data units among the aforementioned P data units. The first true CSI information corresponding to each data unit can be understood as the true CSI information obtained based on the first channel measurement information of that data unit; the sixth reconstructed CSI information corresponding to each data unit can be understood as the reconstructed CSI information obtained based on the second CSI feedback information obtained from the first channel measurement information of that data unit. The estimated performance index information corresponding to each data unit can be understood as the performance index information of that data unit estimated by the first device based on the estimation method.

[0573] The aforementioned eighth indicator information is determined based on the indicator information corresponding to the J data units. For example, the aforementioned eighth indicator information can be the maximum value, minimum value, median value, average value, or statistical value obtained based on a preset statistical algorithm for the indicator information corresponding to the J data units. This embodiment does not limit this.

[0574] In this embodiment, the eighth indicator information is determined based on the performance indicator information corresponding to the J data units, which helps to ensure the reliability of the calculated eighth indicator information.

[0575] Optionally, the eighth indicator information is determined based on the performance indicator information corresponding to the S data unit groups, where S is an integer less than or equal to P;

[0576] The performance index information corresponding to each data unit group is the performance index information determined based on all data units in each data unit group.

[0577] In this embodiment, the data units in the aforementioned S data unit groups are the data units in the aforementioned P data units. Each data unit group may include at least two data units from the aforementioned P data units, and the performance indicator information corresponding to each data unit group can be determined based on some or all of the data units in that data unit group. For example, one performance indicator can be estimated for every 6 data units, and P / 6 performance indicator information can be estimated from one dataset. Then, the P / 6 performance indicator information can be combined to determine the eighth indicator information.

[0578] In this embodiment, the eighth indicator information is determined based on the performance indicator information corresponding to the S data unit groups, which helps to ensure the reliability of the calculated eighth indicator information.

[0579] Optionally, the performance metrics information includes information obtained based on at least one of the following: squared generalized cosine similarity (SGCS) and minimum mean square error (MMSE).

[0580] Optionally, the method further includes:

[0581] If the first test performance information of the target AI unit meets the first performance requirement, but the second test performance information of the target AI unit does not meet the second performance requirement, the first device performs at least one of the following:

[0582] Collect a third dataset;

[0583] Send a second instruction message, which is used to instruct the collection of a third dataset;

[0584] The third dataset includes data from terminals that provide CSI feedback based on the target AI unit or data from terminals of a first type, wherein the first type is the type of terminal that provides CSI feedback based on the target AI unit.

[0585] Optionally, the method further includes at least one of the following:

[0586] The first device sends an updated second dataset;

[0587] The first device sends an updated target AI unit.

[0588] Optionally, the method further includes:

[0589] The first device receives the fifth CSI feedback information;

[0590] The first device obtains the second monitoring result information of the target AI unit based on the fifth CSI feedback information.

[0591] Optionally, the method further includes:

[0592] Receive second real CSI information;

[0593] The first device obtains the second monitoring result information of the target AI unit based on the fifth CSI feedback information, including:

[0594] The first device obtains the second monitoring result information of the target AI unit based on the fifth CSI feedback information and the second real CSI information.

[0595] Optionally, the method further includes:

[0596] If the first test performance information of the target AI unit meets the first performance requirement, and the second test performance information of the target AI unit meets the second performance requirement, then if the second monitoring result information of the target AI unit does not meet the third performance requirement, the first device shall perform at least one of the following:

[0597] Collect the fourth dataset;

[0598] Send a third instruction message, which is used to instruct the collection of a fourth dataset;

[0599] The fourth dataset includes data under the first scenario or data under the first data distribution, where the first scenario is the scenario corresponding to the fifth CSI feedback information, and the first data distribution is the data distribution corresponding to the fifth CSI feedback information.

[0600] It should be noted that the implementation method of this method can be found in the relevant description of the embodiment shown in Figure 4, and will not be repeated here.

[0601] The embodiments of this application are illustrated below with examples:

[0602] Example 1: Referring to Figure 6, the solid black line is used to verify the effectiveness of the target AI unit in the first dataset, and the dashed black line is used to monitor the real-time performance of the target AI unit. Exemplarily, the AI ​​unit processing method provided in this application embodiment includes the following steps:

[0603] Step a1: The terminal receives performance testing dataset 1 (i.e., the first dataset mentioned above) sent by the network-side device. This performance testing dataset 1 includes at least target CSI information (i.e., the first target CSI information mentioned above). For details regarding the content of the performance testing dataset 1, please refer to the relevant description of the first dataset mentioned above; it will not be repeated here.

[0604] Step a2: The UE processes the target CSI information of the performance test dataset 1 based on the encoder (i.e., the target AI unit), obtains CSI feedback information (i.e., the first CSI feedback information), and reports the first CSI feedback information.

[0605] Step a3: The network-side device obtains the reconstructed target CSI (i.e., reconstructed CSI information) based on the decoder and the first CSI feedback information, and obtains the first SGCS (i.e., SGCS1) based on the reconstructed CSI information and the first target CSI information.

[0606] Step a4: The UE performs CSI measurement to obtain raw channel information (i.e., channel measurement information), processes the raw channel information to obtain target CSI information (i.e., third target CSI information), processes the third target CSI information based on the encoder to obtain fifth CSI feedback information, and reports the fifth CSI feedback information and the real CSI information (i.e., second real CSI information).

[0607] Step a5: The network-side device obtains the reconstructed target CSI (i.e., reconstructed CSI information) based on the decoder and the fifth CSI feedback information, and obtains the third SGCS (i.e., SGCS3) based on the reconstructed CSI information and the second real CSI information.

[0608] It should be noted that the execution order between step a1 and step a4 is not limited in the embodiments of this application.

[0609] Example 2: Referring to Figure 7, the black solid line is used to verify the effectiveness of the target AI unit under the first dataset, the gray solid line is used to verify the effectiveness of the target AI unit under the second dataset, and the black dashed line is used to monitor the real-time performance of the target AI unit. Exemplarily, the AI ​​unit processing method provided in this application embodiment includes the following steps:

[0610] Step b1: The terminal receives the performance test dataset 1 (Performance testing dbtbset 1) (i.e., the first dataset mentioned above) sent by the network-side device. The performance test dataset 1 includes at least the target CSI (Tbrget CSI) information (i.e., the first target CSI information mentioned above). The relevant content of the performance test dataset 1 can be found in the description of the first dataset mentioned above, and will not be repeated here.

[0611] Step b2: The UE processes the target CSI information of the performance test dataset 1 based on the encoder (i.e., the target BI unit), obtains the CSI feedback information (i.e., the first CSI feedback information), and reports the first CSI feedback information.

[0612] Step b3: The network-side device obtains the reconstructed target CSI (i.e., reconstructed CSI information) based on the decoder and the first CSI feedback information, and obtains the first SGCS (i.e., SGCS1) based on the reconstructed CSI information and the first target CSI information.

[0613] Step b4: The UE receives the performance test dataset 2 (Performance testing dbtbset2) (i.e., the second dataset mentioned above) sent by the network-side device. This performance test dataset 2 includes at least the original channel information (i.e., the first channel measurement information mentioned above). For details regarding the content of the performance test dataset 2, please refer to the relevant description of the second dataset mentioned above; it will not be repeated here.

[0614] Step b5: The UE processes the raw channel information of the performance test dataset 2 to obtain the target CSI information, processes the target CSI information based on the encoder to obtain the second CSI feedback information, and reports the second CSI feedback information and the real CSI information (i.e., the first real CSI information).

[0615] Step b6: The network-side device obtains the reconstructed target CSI (i.e., reconstructed CSI information) based on the decoder and the second CSI feedback information, and obtains the second SGCS (i.e., SGCS2) based on the reconstructed CSI information and the first real CSI information.

[0616] Step b7: The UE performs CSI measurement to obtain raw channel information (i.e., channel measurement information). This raw channel information can also be called real raw channel information. The raw channel information is processed to obtain target CSI information (i.e., third target CSI information). The third target CSI information is processed based on the encoder to obtain fifth CSI feedback information. The fifth CSI feedback information and real CSI information (i.e., second real CSI information) are reported.

[0617] Step b8: The network-side device obtains the reconstructed target CSI (i.e., reconstructed CSI information) based on the decoder and the fifth CSI feedback information, and obtains the third SGCS (i.e., SGCS3) based on the reconstructed CSI information and the second real CSI information.

[0618] It should be noted that the execution order of steps b1, b4 and b7 is not limited in the embodiments of this application.

[0619] Example 3: Referring to Figure 8, an AI-based communication method provided in this application embodiment may include the following steps:

[0620] Step 1. The Network (NW) sends a UE Capability Query to the UE. This UE Capability Query includes supported functionalities, storage, and data collection.

[0621] Step 2. The UE sends UE Capability Information to the network side. This UE Capability Information may include supported functions, storage, and data collection.

[0622] Step 3. The network sends a configuration for data collection to the UE. The UE then performs data collection based on this configuration.

[0623] Step 4. The UE reports the logged data and immediate data to the network side.

[0624] Step 5. The network side sends the model (i.e., AI unit) to the UE.

[0625] Step 6. The network sends a configuration to the UE, which allows the UE to report UE Assistance Information (UAI) through other configurations (otherConfig). The network can provide additional conditions on the NW side, FFS (Inference Configuration). Upon receiving this configuration, the UE can determine the applicable functionality.

[0626] Step 7. The UE reports the applicable functionality.

[0627] Step 8. The network side sends configuration information to the UE. This step may include:

[0628] 8-1. Perform inference configuration for the UE after reporting the applicable functions;

[0629] 8-2. The network side decides whether to provide updated configurations.

[0630] In some optional embodiments, the AI ​​unit processing method provided in this application embodiment can be executed after step 8, for example, step 401 above. If it is determined based on the AI ​​unit processing method provided in this application embodiment that the AI ​​unit is faulty or its performance is degraded, the data collection step can be returned; if it is determined based on the AI ​​unit processing method provided in this application embodiment that the AI ​​unit is valid, the model inference step can be executed.

[0631] Step 9. The UE performs inference reporting.

[0632] Step 10. The network side sends monitoring configuration to the UE.

[0633] When the UE receives the monitoring configuration, it can monitor the AI ​​unit.

[0634] Step 11. The UE performs monitoring and reporting.

[0635] In some optional embodiments, if it is determined that the AI ​​unit has failed or its performance has degraded based on the monitoring results, the AI ​​unit processing method provided in this application embodiment can be executed, such as step 401 above.

[0636] In summary, the AI ​​unit processing method provided in this application embodiment can identify the matching between the AI ​​unit and the terminal, as well as the matching problem between the AI ​​unit and the current scene and channel conditions, and can update the AI ​​unit, thereby ensuring the performance of the AI ​​unit.

[0637] It should be noted that the AI ​​unit processing method provided in this application embodiment can be executed by an AI unit processing device. This application embodiment uses an AI unit processing device executing the AI ​​unit processing method as an example to illustrate the AI ​​unit processing device provided in this application embodiment.

[0638] This application provides an AI unit processing device. As an example, the AI ​​unit processing device may be a communication device or a component within a communication device, such as a chip. The communication device may be a terminal, a network-side device, or a server, etc. Exemplarily, the terminal may include, but is not limited to, the type of terminal 11 listed above, and the network-side device may include, but is not limited to, the type of network-side device 12 listed above. This application does not impose specific limitations.

[0639] The AI ​​unit processing device includes a receiving module, a transmitting module, and a processing module. These modules can be implemented in software or hardware. When implemented in hardware, the processing module can be implemented by a processor. For example, the processor can include general-purpose processors, special-purpose processors, such as a Central Processing Unit (CPU), microprocessor, Digital Signal Processor (DSP), Artificial Intelligence (AI) processor, Graphics Processing Unit (GPU), Application Specific Integrated Circuit (ASIC), Network Processor (NP), Field Programmable Gate Array (FPGA), or other programmable logic devices, gate circuits, transistors, discrete hardware components, etc. The receiving and transmitting modules can be implemented by a communication interface, which can include one or more of the following: transceiver, pins, circuits, bus, radio frequency unit, etc.

[0640] Specifically, referring to Figure 9, when the AI ​​unit processing device is a terminal or a component in a terminal, the AI ​​unit processing device 900 includes a first processing module 901 for performing at least one of a first operation and a second operation;

[0641] The first operation includes:

[0642] Obtain a first dataset, which includes first target channel state information (CSI) information;

[0643] First information is obtained based on the first target CSI information and the target artificial intelligence (AI) unit;

[0644] The first information includes at least one of the following:

[0645] First CSI feedback information;

[0646] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0647] The first test performance information of the target AI unit;

[0648] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0649] The second operation includes:

[0650] Obtain a second dataset, which includes the first channel measurement information;

[0651] The second information is obtained based on the first channel measurement information and the target AI unit;

[0652] The second information includes at least one of the following:

[0653] Second CSI feedback information;

[0654] The first monitoring result information of the target AI unit;

[0655] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0656] The second test performance information of the target AI unit;

[0657] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0658] Optionally, the first dataset further includes at least one of the following:

[0659] First dataset identification information;

[0660] The indication information or type information of the target CSI information;

[0661] Third CSI feedback information.

[0662] Optionally, the first dataset is associated with third information, which includes at least one of the following:

[0663] The third indicator information is a performance indicator information calculated based on the first target CSI information and the first reconstructed CSI information, wherein the first reconstructed CSI information is the reconstructed CSI information calculated by the network-side device.

[0664] The first indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit;

[0665] The first target indicator information is the expected performance indicator information of the target AI unit;

[0666] First threshold information.

[0667] Optionally, the first operation further includes at least one of the following:

[0668] Report the first CSI feedback information;

[0669] The first indicator information is reported, which is the performance indicator information of the target AI unit estimated by the terminal on the first dataset, or the first indicator information is the performance indicator information calculated based on the first target CSI information and the second reconstructed CSI information, where the second reconstructed CSI information is the reconstructed CSI information calculated by the terminal or the reconstructed CSI information calculated by the network side device.

[0670] Obtain the first test performance information of the target AI unit;

[0671] Report the first test performance information of the target AI unit.

[0672] Optionally, obtaining the first test performance information of the target AI unit includes:

[0673] The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or...

[0674] The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the third CSI feedback information; or...

[0675] Obtain the first test performance information of the target AI unit based on the first indicator information; or...

[0676] The first test performance information of the target AI unit is obtained based on the first indicator information and the third indicator information; or...

[0677] The first test performance information of the target AI unit is obtained based on the first indicator information and the first target indicator information; or...

[0678] The first test performance information of the target AI unit is obtained based on the first threshold information; or...

[0679] The first test performance information of the target AI unit is obtained based on the first indicator information and the first threshold information; or...

[0680] The first test performance information of the target AI unit is obtained based on the first indicator information, the first threshold information, and the first target indicator information.

[0681] Optionally, the first dataset includes K data units, each of which includes first target CSI information, where K is an integer greater than 1.

[0682] Optionally, the first indicator information is determined based on the performance indicator information corresponding to L data units, where L is an integer less than or equal to K;

[0683] The performance metric information corresponding to each data unit is determined according to at least one of the following:

[0684] The third CSI feedback information and the corresponding first CSI feedback information for each data unit;

[0685] The first target CSI information and the corresponding second reconstructed CSI information for each data unit;

[0686] Estimated performance metrics information corresponding to each data unit.

[0687] Optionally, the first indicator information is determined based on the performance indicator information corresponding to the M data unit groups, where M is an integer less than or equal to K;

[0688] The performance index information corresponding to each data unit group is the performance index information determined based on at least two or all data units in each data unit group.

[0689] Optionally, the second dataset is associated with fourth information, which includes at least one of the following:

[0690] Second dataset identification information;

[0691] Indication information or type information of channel measurement information;

[0692] The fourth indicator information is a performance indicator information calculated based on the second target CSI information and the third reconstructed CSI information. The third reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, and the second target CSI information is the target CSI information obtained by the network-side device based on the first channel measurement information.

[0693] The second indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit;

[0694] The second target indicator information is the expected performance indicator information of the target AI unit;

[0695] Second threshold information;

[0696] Format indication information, wherein the format indication information is used to indicate the format of the target CSI information;

[0697] Channel characteristic information;

[0698] Scene information.

[0699] Optionally, the second operation further includes at least one of the following:

[0700] Report the second CSI feedback information;

[0701] The second indicator information is reported. The second indicator information is the performance indicator information of the target AI unit estimated by the terminal on the second dataset. Alternatively, the second indicator information is the performance indicator information calculated based on the first real CSI information and the fourth reconstructed CSI information. The fourth reconstructed CSI information is the reconstructed CSI information calculated by the terminal or the reconstructed CSI information calculated by the network-side device. The first real CSI information is the CSI information obtained by the terminal based on the first channel measurement information.

[0702] Obtain the second test performance information of the target AI unit;

[0703] Report the second test performance information of the target AI unit.

[0704] Optionally, obtaining the second test performance information of the target AI unit includes:

[0705] The second test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or...

[0706] The second test performance information of the target AI unit is obtained based on the second CSI feedback information and the fourth CSI feedback information; or...

[0707] The second test performance information of the target AI unit is obtained based on the second indicator information; or...

[0708] The second test performance information of the target AI unit is obtained based on the second and fourth indicator information; or...

[0709] The second test performance information of the target AI unit is obtained based on the second indicator information and the second target indicator information; or...

[0710] The second test performance information of the target AI unit is obtained based on the second threshold information; or...

[0711] The second test performance information of the target AI unit is obtained based on the second indicator information and the second threshold information; or...

[0712] The second test performance information of the target AI unit is obtained based on the second indicator information, the second threshold information, and the second target indicator information.

[0713] Optionally, the second dataset includes P data units, each of which includes first channel measurement information, where P is an integer greater than 1.

[0714] Optionally, the second indicator information is determined based on the performance indicator information corresponding to Q data units, where Q is an integer less than or equal to P;

[0715] The performance metric information corresponding to each data unit is determined according to at least one of the following:

[0716] The second CSI feedback information and the fourth CSI feedback information corresponding to each data unit;

[0717] The first real CSI information and the fourth reconstructed CSI information corresponding to each data unit, wherein the first real CSI information corresponding to each data unit is CSI information obtained based on the first channel measurement information of each data unit;

[0718] Estimated performance metrics information corresponding to each data unit.

[0719] Optionally, the first indicator information is determined based on the performance indicator information corresponding to the N data unit groups, where N is an integer less than or equal to P;

[0720] The performance index information corresponding to each data unit group is the performance index information determined based on all data units in each data unit group.

[0721] Optionally, the performance metrics information includes information obtained based on at least one of the following: squared generalized cosine similarity (SGCS) and minimum mean square error (MMSE).

[0722] Optionally, the second operation further includes:

[0723] The first true CSI information is reported. The first true CSI information is the CSI information obtained by the terminal based on the first channel measurement information. The first true CSI information is used to obtain the second test performance information of the target AI unit.

[0724] Optionally, the second operation further includes:

[0725] Receive first indication information, which is used to instruct the terminal to report the first real CSI information.

[0726] Optionally, the first indication information includes at least one of the following:

[0727] Orthogonal beam vector number;

[0728] Indication information for Type II parameter combinations;

[0729] Indication information of the angle delay domain codebook;

[0730] The codebook type information of the first real CSI information.

[0731] Optionally, the first true CSI information includes the following: a codebook matrix, a compressed codebook matrix.

[0732] Optionally, the second dataset is used to test at least one of the performance of the target AI unit and the performance of the first processing unit, wherein the first processing unit is used by the terminal to process channel measurement information to obtain target CSI information.

[0733] Optionally, the first processing module is further configured to:

[0734] If the first test performance information of the target AI unit meets the first performance requirement, but the second test performance information of the target AI unit does not meet the second performance requirement, at least one of the following shall be performed:

[0735] Collect a third dataset;

[0736] Report the third dataset;

[0737] Send a second instruction message, which is used to instruct the collection of a third dataset;

[0738] The third dataset includes data from the terminal or data from a first type of terminal, where the first type refers to the type of terminal.

[0739] Optionally, the third dataset includes at least one of the following:

[0740] The terminal or the terminal of the first type acquires target CSI information based on CSI measurements;

[0741] Data type identification information, or terminal type identification information, or terminal identification information;

[0742] Data feature information based on target CSI information obtained from CSI measurements;

[0743] The CSI measurement is associated with at least one of the channel characteristic information and scene information.

[0744] Optionally, the apparatus further includes a receiving module for receiving at least one of an updated first dataset, an updated second dataset, and an updated target AI unit;

[0745] The first processing module is further configured to perform at least one of the following:

[0746] The target AI unit is updated based on at least one of the updated first dataset and the updated second dataset;

[0747] The target AI unit is updated based on the third dataset.

[0748] Optionally, the first processing module is further configured to:

[0749] Perform a third operation, the third operation including:

[0750] Perform CSI measurements to obtain CSI information for the third target;

[0751] The fifth CSI feedback information is obtained based on the third target CSI information and the target AI unit;

[0752] The fifth CSI feedback information is reported, which is used to obtain the second monitoring result information of the target AI unit and the channel information of the terminal for the network-side device to determine.

[0753] Optionally, the third operation further includes at least one of the following:

[0754] The second real CSI information is reported. The second real CSI information is the CSI information obtained by the terminal based on the third target CSI information. The second real CSI information is used to obtain the second monitoring result information of the target AI unit.

[0755] Obtain the second monitoring result information of the target AI unit.

[0756] Optionally, the first processing module is further configured to:

[0757] If the first test performance information of the target AI unit meets the first performance requirement, and the second test performance information of the target AI unit meets the second performance requirement, then if the second monitoring result information of the target AI unit does not meet the third performance requirement, then at least one of the following shall be executed:

[0758] Collect the fourth dataset;

[0759] Report the fourth dataset;

[0760] Send a third instruction message, which is used to instruct the collection of a fourth dataset;

[0761] The fourth dataset includes data under the first scenario or data under the first data distribution, where the first scenario is the scenario corresponding to the fifth CSI feedback information, and the first data distribution is the data distribution corresponding to the fifth CSI feedback information.

[0762] Optionally, the fourth dataset includes at least one of the following:

[0763] The target CSI information obtained by the terminal in the first scenario or the terminal under the first data distribution based on CSI measurement;

[0764] Scene information;

[0765] Data distribution information.

[0766] Optionally, the first processing module is specifically used for:

[0767] If the first test performance information of the target AI unit meets the first performance requirement, the third operation is performed.

[0768] or,

[0769] If the first test performance information of the target AI unit meets the first performance requirement and the second test performance information of the target AI unit meets the second performance requirement, then the third operation is performed.

[0770] Optionally, the first processing module is specifically used for:

[0771] If the second monitoring result information of the target AI unit does not meet the third performance requirement, at least one of the first operation and the second operation shall be performed.

[0772] Optionally, the first dataset is predefined by the protocol, or the first dataset is configured or sent by the first device;

[0773] And / or,

[0774] The second dataset is predefined by the protocol, or the second dataset is configured or sent by the first device.

[0775] Optionally, the target AI unit includes at least one of the following:

[0776] The terminal is a first AI unit used to acquire CSI feedback information;

[0777] The network-side device is used as a second AI unit to acquire reconstructed CSI information;

[0778] The terminal is a reference AI unit used to acquire CSI feedback information;

[0779] Network-side devices are used to acquire reference AI units for reconstructing CSI information;

[0780] The AI ​​unit used in the test by the terminal;

[0781] AI units used in network-side equipment testing;

[0782] The AI ​​unit used in the test equipment;

[0783] The terminal is used to match the AI ​​unit of the AI ​​unit used in the test;

[0784] Network-side devices are used to match AI units used in testing;

[0785] The testing equipment is used to match the AI ​​units used in the test.

[0786] The AI ​​unit processing device provided in this application embodiment can implement the various processes implemented in the method embodiment of FIG4 and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0787] Referring to Figure 10, when the AI ​​unit processing device is a network-side device or a component in a network-side device, the AI ​​unit processing device 1000 includes a second processing module 1001 for performing at least one of the fourth and fifth operations;

[0788] The fourth operation includes at least one of the following:

[0789] Send a first dataset, which includes first target CSI information;

[0790] Send a second dataset, which includes the first channel measurement information;

[0791] The fifth operation includes at least one of the following: receiving fifth information, receiving sixth information;

[0792] The fifth piece of information includes at least one of the following:

[0793] First CSI feedback information;

[0794] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0795] The first test performance information of the target AI unit;

[0796] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0797] The sixth piece of information includes at least one of the following:

[0798] Second CSI feedback information;

[0799] The first monitoring result information of the target AI unit;

[0800] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0801] The second test performance information of the target AI unit;

[0802] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0803] Optionally, the first dataset further includes at least one of the following:

[0804] First dataset identification information;

[0805] The indication information or type information of the target CSI information;

[0806] Third CSI feedback information.

[0807] Optionally, the first dataset is associated with third information, which includes at least one of the following:

[0808] The third indicator information is a performance indicator information calculated based on the first target CSI information and the first reconstructed CSI information, wherein the first reconstructed CSI information is the reconstructed CSI information calculated by the network-side device.

[0809] The first indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit;

[0810] The first target indicator information is the expected performance indicator information of the target AI unit;

[0811] First threshold information.

[0812] Optionally, the second dataset is associated with fourth information, which includes at least one of the following:

[0813] Second dataset identification information;

[0814] Indication information or type information of channel measurement information;

[0815] The fourth indicator information is a performance indicator information calculated based on the second target CSI information and the third reconstructed CSI information. The third reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, and the second target CSI information is the target CSI information obtained by the network-side device based on the first channel measurement information.

[0816] The second indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit;

[0817] The second target indicator information is the expected performance indicator information of the target AI unit;

[0818] Second threshold information;

[0819] Format indication information, wherein the format indication information is used to indicate the format of the target CSI information;

[0820] Channel characteristic information;

[0821] Scene information.

[0822] Optionally, the fifth operation further includes at least one of the following:

[0823] Obtain the first test performance information of the target AI unit;

[0824] Obtain the second test performance information of the target AI unit.

[0825] Optionally, obtaining the first test performance information of the target AI unit includes:

[0826] The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or...

[0827] The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the third CSI feedback information; or...

[0828] The first test performance information of the target AI unit is obtained based on the fifth indicator information; or...

[0829] The first test performance information of the target AI unit is obtained based on the fifth and third indicator information; or...

[0830] The first test performance information of the target AI unit is obtained based on the fifth indicator information and the first target indicator information; or...

[0831] The first test performance information of the target AI unit is obtained based on the first threshold information; or...

[0832] The first test performance information of the target AI unit is obtained based on the fifth indicator information and the first threshold information; or...

[0833] The first test performance information of the target AI unit is obtained based on the fifth indicator information, the first threshold information, and the first target indicator information.

[0834] The fifth indicator information includes either the first indicator information or the sixth indicator information. The sixth indicator information is the performance indicator information of the target AI unit on the first dataset estimated by the first device, or the sixth indicator information is the performance indicator information calculated based on the first target CSI information and the fifth reconstructed CSI information. The fifth reconstructed CSI information is the reconstructed CSI information calculated by the network-side device.

[0835] Optionally, obtaining the second test performance information of the target AI unit includes:

[0836] The second test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or...

[0837] The second test performance information of the target AI unit is obtained based on the second CSI feedback information and the fourth CSI feedback information; or...

[0838] The second test performance information of the target AI unit is obtained based on the seventh indicator information; or...

[0839] The second test performance information of the target AI unit is obtained based on the seventh and fourth indicator information; or...

[0840] The second test performance information of the target AI unit is obtained based on the seventh indicator information and the second target indicator information; or...

[0841] The second test performance information of the target AI unit is obtained based on the second threshold information; or...

[0842] The second test performance information of the target AI unit is obtained based on the seventh indicator information and the second threshold information; or...

[0843] The second test performance information of the target AI unit is obtained based on the seventh indicator information, the second threshold information, and the second target indicator information.

[0844] The seventh indicator information includes either the second indicator information or the eighth indicator information. The eighth indicator information is the performance indicator information of the target AI unit on the second dataset estimated by the first device, or the eighth indicator information is the performance indicator information calculated based on the first real CSI information and the sixth reconstructed CSI information, or the eighth indicator information is the performance indicator information calculated based on the fourth target CSI information and the sixth reconstructed CSI information. The sixth reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, and the fourth target CSI information is the target CSI information estimated based on the first channel measurement information.

[0845] Optionally, the fifth operation further includes:

[0846] Receive the first real CSI information.

[0847] Optionally, the fifth operation further includes:

[0848] Send a first instruction message, which is used to instruct the terminal to report the first real CSI information.

[0849] Optionally, the first indication information includes at least one of the following:

[0850] Orthogonal beam vector number;

[0851] Indication information for Type II parameter combinations;

[0852] Indication information of the angle delay domain codebook;

[0853] The codebook type information of the first real CSI information.

[0854] Optionally, the fourth target CSI information includes at least two target CSI information, wherein the at least two target CSI information are target CSI information obtained by using at least two estimation methods;

[0855] The eighth indicator information is determined based on at least two performance indicator information, and each of the at least two performance indicator information is determined based on each of the at least two target CSI information and the sixth reconstructed CSI information.

[0856] Optionally, the first dataset includes K data units, each of which includes first target CSI information, where K is an integer greater than 1;

[0857] And / or,

[0858] The second dataset includes M data units, each of which includes first channel measurement information, where M is an integer greater than 1.

[0859] Optionally, the performance metrics information includes information obtained based on at least one of the following: squared generalized cosine similarity (SGCS) and minimum mean square error (MMSE).

[0860] Optionally, the processing module is further configured to:

[0861] If the first test performance information of the target AI unit meets the first performance requirement, but the second test performance information of the target AI unit does not meet the second performance requirement, at least one of the following shall be performed:

[0862] Collect a third dataset;

[0863] Send a second instruction message, which is used to instruct the collection of a third dataset;

[0864] The third dataset includes data from terminals that provide CSI feedback based on the target AI unit or data from terminals of a first type, wherein the first type is the type of terminal that provides CSI feedback based on the target AI unit.

[0865] Optionally, the device further includes a transmitting module for at least one of the following:

[0866] Send the updated second dataset;

[0867] Send the updated target AI unit.

[0868] Optionally, the device further includes a receiving module for receiving fifth CSI feedback information;

[0869] The second processing module is also used to obtain the second monitoring result information of the target AI unit based on the fifth CSI feedback information.

[0870] Optionally, the device further includes a receiving module for receiving second real CSI information;

[0871] The second processing module is specifically used for:

[0872] The second monitoring result information of the target AI unit is obtained based on the fifth CSI feedback information and the second real CSI information.

[0873] Optionally, the second processing module is further configured to:

[0874] If the first test performance information of the target AI unit meets the first performance requirement, and the second test performance information of the target AI unit meets the second performance requirement, then if the second monitoring result information of the target AI unit does not meet the third performance requirement, then at least one of the following shall be executed:

[0875] Collect the fourth dataset;

[0876] Send a third instruction message, which is used to instruct the collection of a fourth dataset;

[0877] The fourth dataset includes data under the first scenario or data under the first data distribution, where the first scenario is the scenario corresponding to the fifth CSI feedback information, and the first data distribution is the data distribution corresponding to the fifth CSI feedback information.

[0878] The AI ​​unit processing device provided in this application embodiment can implement the various processes implemented in the method embodiment of FIG5 and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0879] As shown in Figure 11, this application embodiment also provides a communication device 1100, including a processor 1101 and a memory 1102. The memory 1102 stores programs or instructions that can run on the processor 1101. For example, when the communication device 1100 is a terminal, the program or instructions executed by the processor 1101 implement the various steps of the above-described terminal-side AI unit processing method embodiment and achieve the same technical effect. When the communication device 1100 is a first device, the program or instructions executed by the processor 1101 implement the various steps of the above-described first device-side AI unit processing method embodiment and achieve the same technical effect. To avoid repetition, this will not be described again here.

[0880] This 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 programs or instructions to implement the steps in the method embodiment shown in FIG4. This terminal embodiment corresponds to the above-described terminal-side method embodiment, and all implementation processes and methods of the above-described method embodiments can be applied to this terminal embodiment and can achieve the same technical effect. The terminal may be the AI ​​unit processing device shown in FIG9. Specifically, FIG12 is a schematic diagram of the hardware structure of a terminal implementing an embodiment of this application.

[0881] The terminal 1200 includes, but is not limited to, at least some of the following components: radio frequency unit 1201, network module 1202, audio output unit 1203, input unit 1204, sensor 1205, display unit 1206, user input unit 1207, interface unit 1208, memory 1209, and processor 1210.

[0882] Those skilled in the art will understand that terminal 1200 may also include a power supply (such as a battery) for powering various components. The power supply can be logically connected to processor 1210 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The terminal structure shown in Figure 12 does not constitute a limitation on the terminal. The terminal may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here.

[0883] It should be understood that, in this embodiment, the input unit 1204 may include a graphics processor 12041 and a microphone 12042. The graphics processor 12041 processes image data of still images or videos obtained by an image capture device (such as a camera) in video capture mode or image capture mode. The display unit 1206 may include a display panel 12061, which may be configured in the form of a liquid crystal display, an organic light-emitting diode, or the like. The user input unit 1207 includes a touch panel 12071 and at least one of other input devices 12072. The touch panel 12071 is also called a touch screen. The touch panel 12071 may include a touch detection device and a touch controller. Other input devices 12072 may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, and joysticks, which will not be described in detail here.

[0884] In this embodiment, after receiving downlink data from the network-side device, the radio frequency unit 1201 can transmit it to the processor 1210 for processing; in addition, the radio frequency unit 1201 can send uplink data to the network-side device. Typically, the radio frequency unit 1201 includes, but is not limited to, antennas, amplifiers, transceivers, couplers, low-noise amplifiers, duplexers, etc.

[0885] The memory 1209 can be used to store software programs or instructions, as well as various data. The memory 1209 may primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area may store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, the memory 1209 may include volatile memory or non-volatile memory. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DRRAM). The memory 1209 in this embodiment includes, but is not limited to, these and any other suitable types of memory.

[0886] Processor 1210 may include one or more processing units; optionally, processor 1210 integrates an application processor and a modem processor, wherein the application processor mainly handles operations involving the operating system, user interface, and applications, and the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into processor 1210.

[0887] The processor 1210 is configured to perform at least one of the first operation and the second operation;

[0888] The first operation includes:

[0889] Obtain a first dataset, which includes first target channel state information (CSI) information;

[0890] First information is obtained based on the first target CSI information and the target artificial intelligence (AI) unit;

[0891] The first information includes at least one of the following:

[0892] First CSI feedback information;

[0893] The first indicator information is the performance indicator information of the target AI unit on the first dataset.

[0894] The first test performance information of the target AI unit;

[0895] First failure reason information, which is used to indicate the reason for the failure of the target AI unit;

[0896] The second operation includes:

[0897] Obtain a second dataset, which includes the first channel measurement information;

[0898] The second information is obtained based on the first channel measurement information and the target AI unit;

[0899] The second information includes at least one of the following:

[0900] Second CSI feedback information;

[0901] The first monitoring result information of the target AI unit;

[0902] The second indicator information is the performance indicator information of the target AI unit on the second dataset.

[0903] The second test performance information of the target AI unit;

[0904] The second failure reason information is used to indicate the reason for the failure of the target AI unit.

[0905] It is understood that the implementation process of each implementation method mentioned in this embodiment can refer to the relevant description of the aforementioned terminal-side AI unit processing method embodiment, and achieve the same or corresponding technical effects. To avoid repetition, it will not be described again here.

[0906] This application also provides a first device, including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the steps of the method embodiment shown in FIG5. This first device embodiment corresponds to the above-described first device method embodiment, and all implementation processes and methods of the above-described method embodiments can be applied to this first device embodiment and can achieve the same technical effect.

[0907] Specifically, this application embodiment also provides a network-side device, which may be the AI ​​unit processing device shown in FIG10. As shown in FIG13, the network-side device 1300 includes: an antenna 1301, a radio frequency device 1302, a baseband device 1303, a processor 1304, and a memory 1305. The antenna 1301 is connected to the radio frequency device 1302. In the uplink direction, the radio frequency device 1302 receives information through the antenna 1301 and sends the received information to the baseband device 1303 for processing. In the downlink direction, the baseband device 1303 processes the information to be transmitted and sends it to the radio frequency device 1302, which processes the received information and then transmits it through the antenna 1301.

[0908] The method executed by the network-side device in the above embodiments can be implemented in the baseband device 1303, which includes a baseband processor.

[0909] The baseband device 1303 may include at least one baseband board, on which multiple chips are disposed, as shown in FIG13. One of the chips is, for example, a baseband processor, which is connected to the memory 1305 via a bus interface to call the program in the memory 1305 to execute the network device operation shown in the above method embodiment.

[0910] The network-side device may also include a network interface 1306, such as a Common Public Radio Interface (CPRI).

[0911] Specifically, the network-side device 1300 in this application embodiment further includes: instructions or programs stored in memory 1305 and executable on processor 1304. Processor 1304 calls the instructions or programs in memory 1305 to execute the methods executed by each module shown in FIG10 and achieve the same technical effect. To avoid repetition, it will not be described in detail here.

[0912] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the AI ​​unit processing method embodiments described above and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0913] The processor mentioned above is the processor in the terminal described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk. In some examples, the readable storage medium may be a non-transient readable storage medium.

[0914] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described AI unit processing method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0915] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0916] This application also provides a computer program / program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the AI ​​unit processing method embodiments described above, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0917] This application also provides a wireless communication system, including: a terminal and a first device, wherein the terminal can be used to execute the steps of the AI ​​unit processing method described above, and the first device can be used to execute the steps of the AI ​​unit processing method described above.

[0918] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0919] From the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of computer software products plus necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, magnetic disk, optical disk, etc.) and includes several instructions to cause the terminal or network-side device to execute the methods described in the various embodiments of this application.

[0920] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other implementations under the guidance of this application without departing from the spirit and scope of the claims. All of these implementations are within the protection scope of this application.

Claims

An AI unit processing method, comprising: The terminal performs at least one of the first and second operations; The first operation includes: Obtain a first dataset, which includes first target channel state information (CSI) information; First information is obtained based on the first target CSI information and the target artificial intelligence (AI) unit; The first information includes at least one of the following: First CSI feedback information; The first indicator information is the performance indicator information of the target AI unit on the first dataset. The first test performance information of the target AI unit; First failure reason information, which is used to indicate the reason for the failure of the target AI unit; The second operation includes: Obtain a second dataset, which includes the first channel measurement information; The second information is obtained based on the first channel measurement information and the target AI unit; The second information includes at least one of the following: Second CSI feedback information; The first monitoring result information of the target AI unit; The second indicator information is the performance indicator information of the target AI unit on the second dataset. The second test performance information of the target AI unit; The second failure reason information is used to indicate the reason for the failure of the target AI unit. According to the method of claim 1, wherein, The first dataset also includes at least one of the following: First dataset identification information; The indication information or type information of the target CSI information; Third CSI feedback information. According to the method of claim 1, wherein, The first dataset is associated with third information, which includes at least one of the following: The third indicator information is a performance indicator information calculated based on the first target CSI information and the first reconstructed CSI information, wherein the first reconstructed CSI information is the reconstructed CSI information calculated by the network-side device. The first indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit; The first target indicator information is the expected performance indicator information of the target AI unit; First threshold information. The method according to any one of claims 1 to 3, wherein, The first operation further includes at least one of the following: Report the first CSI feedback information; The first indicator information is reported, which is the performance indicator information of the target AI unit estimated by the terminal on the first dataset, or the first indicator information is the performance indicator information calculated based on the first target CSI information and the second reconstructed CSI information, where the second reconstructed CSI information is the reconstructed CSI information calculated by the terminal or the reconstructed CSI information calculated by the network side device. Obtain the first test performance information of the target AI unit; Report the first test performance information of the target AI unit. The method according to claim 4, wherein, The acquisition of the first test performance information of the target AI unit includes: The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or... The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the third CSI feedback information; or... Obtain the first test performance information of the target AI unit based on the first indicator information; or... The first test performance information of the target AI unit is obtained based on the first indicator information and the third indicator information; or... The first test performance information of the target AI unit is obtained based on the first indicator information and the first target indicator information; or... The first test performance information of the target AI unit is obtained based on the first threshold information; or... The first test performance information of the target AI unit is obtained based on the first indicator information and the first threshold information; or... The first test performance information of the target AI unit is obtained based on the first indicator information, the first threshold information, and the first target indicator information. The method according to any one of claims 1 to 5, wherein, The first dataset includes K data units, each of which includes first target CSI information, where K is an integer greater than 1. The method according to claim 6, wherein, The first indicator information is determined based on the performance indicator information corresponding to L data units, where L is an integer less than or equal to K; The performance metric information corresponding to each data unit is determined according to at least one of the following: The third CSI feedback information and the corresponding first CSI feedback information for each data unit; The first target CSI information and the corresponding second reconstructed CSI information for each data unit; Estimated performance metrics information corresponding to each data unit. The method according to claim 7, wherein, The first indicator information is determined based on the performance indicator information corresponding to M data unit groups, where M is an integer less than or equal to K; The performance index information corresponding to each data unit group is the performance index information determined based on at least two or all data units in each data unit group. The method according to any one of claims 1 to 8, wherein, The second dataset is associated with fourth information, which includes at least one of the following: Second dataset identification information; Indication information or type information of channel measurement information; The fourth indicator information is a performance indicator information calculated based on the second target CSI information and the third reconstructed CSI information. The third reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, and the second target CSI information is the target CSI information obtained by the network-side device based on the first channel measurement information. The second indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit; The second target indicator information is the expected performance indicator information of the target AI unit; Second threshold information; Format indication information, wherein the format indication information is used to indicate the format of the target CSI information; Channel characteristic information; Scene information. The method according to any one of claims 1 to 9, wherein, The second operation also includes at least one of the following: Report the second CSI feedback information; The second indicator information is reported. The second indicator information is the performance indicator information of the target AI unit estimated by the terminal on the second dataset. Alternatively, the second indicator information is the performance indicator information calculated based on the first real CSI information and the fourth reconstructed CSI information. The fourth reconstructed CSI information is the reconstructed CSI information calculated by the terminal or the reconstructed CSI information calculated by the network-side device. The first real CSI information is the CSI information obtained by the terminal based on the first channel measurement information. Obtain the second test performance information of the target AI unit; Report the second test performance information of the target AI unit. The method according to claim 10, wherein, The acquisition of the second test performance information of the target AI unit includes: The second test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or... The second test performance information of the target AI unit is obtained based on the second CSI feedback information and the fourth CSI feedback information; or... The second test performance information of the target AI unit is obtained based on the second indicator information; or... The second test performance information of the target AI unit is obtained based on the second and fourth indicator information; or... The second test performance information of the target AI unit is obtained based on the second indicator information and the second target indicator information; or... The second test performance information of the target AI unit is obtained based on the second threshold information; or... The second test performance information of the target AI unit is obtained based on the second indicator information and the second threshold information; or... The second test performance information of the target AI unit is obtained based on the second indicator information, the second threshold information, and the second target indicator information. The method according to any one of claims 1 to 11, wherein, The second dataset includes P data units, each of which includes first channel measurement information, where P is an integer greater than 1. The method according to claim 12, wherein, The second indicator information is determined based on the performance indicator information corresponding to Q data units, where Q is an integer less than or equal to P; The performance metric information corresponding to each data unit is determined according to at least one of the following: The second CSI feedback information and the fourth CSI feedback information corresponding to each data unit; The first real CSI information and the fourth reconstructed CSI information corresponding to each data unit, wherein the first real CSI information corresponding to each data unit is CSI information obtained based on the first channel measurement information of each data unit; Estimated performance metrics information corresponding to each data unit. The method according to claim 12, wherein, The first indicator information is determined based on the performance indicator information corresponding to N data unit groups, where N is an integer less than or equal to P; The performance index information corresponding to each data unit group is the performance index information determined based on at least two or all data units in each data unit group. The method according to any one of claims 1 to 14, wherein, The performance metrics information includes information obtained based on at least one of the following: squared generalized cosine similarity (SGCS) and minimum mean square error (MMSE). The method according to any one of claims 1 to 15, wherein, The second operation also includes: The first true CSI information is reported. The first true CSI information is the CSI information obtained by the terminal based on the first channel measurement information. The first true CSI information is used to obtain the second test performance information of the target AI unit. The method according to claim 16, wherein, The second operation also includes: Receive first indication information, which is used to instruct the terminal to report the first real CSI information. The method according to claim 17, wherein, The first indication information includes at least one of the following: Orthogonal beam vector number; Indication information for Type II parameter combinations; Indication information of the angle delay domain codebook; The codebook type information of the first real CSI information. The method according to any one of claims 10 to 18, wherein, The first true CSI information includes the following: codebook matrix, compressed codebook matrix. The method according to any one of claims 1 to 19, wherein, The second dataset is used to test at least one of the performance of the target AI unit and the performance of the first processing unit, wherein the first processing unit is used by the terminal to process channel measurement information to obtain target CSI information. The method according to any one of claims 1 to 20, the method further comprising: If the first test performance information of the target AI unit meets the first performance requirement, but the second test performance information of the target AI unit does not meet the second performance requirement, the terminal performs at least one of the following: Collect a third dataset; Report the third dataset; Send a second instruction message, which is used to instruct the collection of a third dataset; The third dataset includes data from the terminal or data from a first type of terminal, where the first type refers to the type of terminal. The method according to claim 21, wherein, The third dataset includes at least one of the following: The terminal or the terminal of the first type acquires target CSI information based on CSI measurements; Data type identification information, or terminal type identification information, or terminal identification information; Data feature information based on target CSI information obtained from CSI measurements; The CSI measurement is associated with at least one of the channel characteristic information and scene information. The method according to claim 21 or 22, further comprising at least one of the following: The terminal receives at least one of an updated first dataset and an updated second dataset; The terminal updates the target AI unit based on at least one of the updated first dataset and the updated second dataset; The terminal updates the target AI unit based on the third dataset; The terminal receives the updated target AI unit. The method according to any one of claims 1 to 23, the method further comprising: The terminal performs a third operation, which includes: Perform CSI measurements to obtain CSI information for the third target; The fifth CSI feedback information is obtained based on the third target CSI information and the target AI unit; The fifth CSI feedback information is reported, which is used to obtain the second monitoring result information of the target AI unit and the channel information of the terminal for the network-side device to determine. The method according to claim 24, wherein, The third operation also includes at least one of the following: The second real CSI information is reported. The second real CSI information is the CSI information obtained by the terminal based on the third target CSI information. The second real CSI information is used to obtain the second monitoring result information of the target AI unit. Obtain the second monitoring result information of the target AI unit. The method according to claim 24 or 25, further comprising: If the first test performance information of the target AI unit meets the first performance requirement, and the second test performance information of the target AI unit meets the second performance requirement, then if the second monitoring result information of the target AI unit does not meet the third performance requirement, the terminal executes at least one of the following: Collect the fourth dataset; Report the fourth dataset; Send a third instruction message, which is used to instruct the collection of a fourth dataset; The fourth dataset includes data under the first scenario or data under the first data distribution, where the first scenario is the scenario corresponding to the fifth CSI feedback information, and the first data distribution is the data distribution corresponding to the fifth CSI feedback information. The method according to claim 26, wherein, The fourth dataset includes at least one of the following: The target CSI information obtained by the terminal in the first scenario or the terminal under the first data distribution based on CSI measurement; Scene information; Data distribution information. The method according to any one of claims 24 to 27, wherein, The terminal performs a third operation, including: If the first test performance information of the target AI unit meets the first performance requirement, the terminal performs a third operation; or, If the first test performance information of the target AI unit meets the first performance requirement and the second test performance information of the target AI unit meets the second performance requirement, the terminal performs a third operation. The method according to any one of claims 24 to 26, wherein, The terminal performs at least one of the first operation and the second operation, including: If the second monitoring result information of the target AI unit does not meet the third performance requirement, the terminal performs at least one of the first operation and the second operation. The method according to any one of claims 1 to 29, wherein, The first dataset is predefined by the protocol, or the first dataset is configured or sent by the first device; And / or, The second dataset is predefined by the protocol, or the second dataset is configured or sent by the first device. The method according to any one of claims 1 to 30, wherein, The target AI unit includes at least one of the following: The terminal is a first AI unit used to acquire CSI feedback information; The network-side device is used as a second AI unit to acquire reconstructed CSI information; The terminal is a reference AI unit used to acquire CSI feedback information; Network-side devices are used to acquire reference AI units for reconstructing CSI information; The AI ​​unit used in the test by the terminal; AI units used in network-side equipment during testing; The AI ​​unit used in the test equipment; The terminal is used to match the AI ​​unit of the AI ​​unit used in the test; Network-side devices are used to match AI units used in testing; The testing equipment is used to match the AI ​​units used in the test. An AI unit processing method, comprising: The first device performs at least one of the fourth and fifth operations; The fourth operation includes at least one of the following: Send a first dataset, which includes first target CSI information; Send a second dataset, which includes the first channel measurement information; The fifth operation includes at least one of the following: receiving fifth information, receiving sixth information; The fifth piece of information includes at least one of the following: First CSI feedback information; The first indicator information is the performance indicator information of the target AI unit on the first dataset. The first test performance information of the target AI unit; First failure reason information, which is used to indicate the reason for the failure of the target AI unit; The sixth piece of information includes at least one of the following: Second CSI feedback information; The first monitoring result information of the target AI unit; The second indicator information is the performance indicator information of the target AI unit on the second dataset. The second test performance information of the target AI unit; The second failure reason information is used to indicate the reason for the failure of the target AI unit. The method according to claim 32, wherein, The first dataset also includes at least one of the following: First dataset identification information; The indication information or type information of the target CSI information; Third CSI feedback information. The method according to claim 32 or 33, wherein, The first dataset is associated with third information, which includes at least one of the following: The third indicator information is a performance indicator information calculated based on the first target CSI information and the first reconstructed CSI information, wherein the first reconstructed CSI information is the reconstructed CSI information calculated by the network-side device. The first indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit; The first target indicator information is the expected performance indicator information of the target AI unit; First threshold information. The method according to any one of claims 32 to 34, wherein, The second dataset is associated with fourth information, which includes at least one of the following: Second dataset identification information; Indication information or type information of channel measurement information; The fourth indicator information is a performance indicator information calculated based on the second target CSI information and the third reconstructed CSI information. The third reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, and the second target CSI information is the target CSI information obtained by the network-side device based on the first channel measurement information. The second indicator indication information includes at least one of the following: indication information of the performance indicator of the target AI unit and type information of the performance indicator of the target AI unit; The second target indicator information is the expected performance indicator information of the target AI unit; Second threshold information; Format indication information, wherein the format indication information is used to indicate the format of the target CSI information; Channel characteristic information; Scene information. The method according to any one of claims 32 to 35, wherein, The fifth operation also includes at least one of the following: Obtain the first test performance information of the target AI unit; Obtain the second test performance information of the target AI unit. The method according to claim 36, wherein, The acquisition of the first test performance information of the target AI unit includes: The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or... The first test performance information of the target AI unit is obtained based on the first CSI feedback information and the third CSI feedback information; or... The first test performance information of the target AI unit is obtained based on the fifth indicator information; or... The first test performance information of the target AI unit is obtained based on the fifth and third indicator information; or... The first test performance information of the target AI unit is obtained based on the fifth indicator information and the first target indicator information; or... The first test performance information of the target AI unit is obtained based on the first threshold information; or... The first test performance information of the target AI unit is obtained based on the fifth indicator information and the first threshold information; The first test performance information of the target AI unit is obtained based on the fifth indicator information, the first threshold information, and the first target indicator information. The fifth indicator information includes either the first indicator information or the sixth indicator information. The sixth indicator information is the performance indicator information of the target AI unit on the first dataset estimated by the first device, or the sixth indicator information is the performance indicator information calculated based on the first target CSI information and the fifth reconstructed CSI information. The fifth reconstructed CSI information is the reconstructed CSI information calculated by the network-side device. The method according to claim 36 or 37, wherein, The acquisition of the second test performance information of the target AI unit includes: The second test performance information of the target AI unit is obtained based on the first CSI feedback information and the second CSI feedback information; or... The second test performance information of the target AI unit is obtained based on the second CSI feedback information and the fourth CSI feedback information; or... The second test performance information of the target AI unit is obtained based on the seventh indicator information; or... The second test performance information of the target AI unit is obtained based on the seventh and fourth indicator information; or... The second test performance information of the target AI unit is obtained based on the seventh indicator information and the second target indicator information; or... The second test performance information of the target AI unit is obtained based on the second threshold information; or... The second test performance information of the target AI unit is obtained based on the seventh indicator information and the second threshold information; or... The second test performance information of the target AI unit is obtained based on the seventh indicator information, the second threshold information, and the second target indicator information. The seventh indicator information includes either the second indicator information or the eighth indicator information. The eighth indicator information is the performance indicator information of the target AI unit on the second dataset estimated by the first device, or the eighth indicator information is the performance indicator information calculated based on the first real CSI information and the sixth reconstructed CSI information, or the eighth indicator information is the performance indicator information calculated based on the fourth target CSI information and the sixth reconstructed CSI information. The sixth reconstructed CSI information is the reconstructed CSI information calculated by the network-side device, and the fourth target CSI information is the target CSI information estimated based on the first channel measurement information. The method according to claim 38, wherein, The fifth operation also includes: Receive the first real CSI information. The method according to claim 39, wherein, The fifth operation also includes: Send a first instruction message, which is used to instruct the terminal to report the first real CSI information. The method according to claim 40, wherein, The first indication information includes at least one of the following: Orthogonal beam vector number; Indication information for Type II parameter combinations; Indication information of the angle delay domain codebook; The codebook type information of the first real CSI information. The method according to claim 38, wherein, The fourth target CSI information includes at least two target CSI information, which are target CSI information obtained by at least two estimation methods. The eighth indicator information is determined based on at least two performance indicator information, and each of the at least two performance indicator information is determined based on each of the at least two target CSI information and the sixth reconstructed CSI information. The method according to any one of claims 32 to 42, wherein, The first dataset includes K data units, each of which includes first target CSI information, where K is an integer greater than 1; And / or, The second dataset includes M data units, each of which includes first channel measurement information, where M is an integer greater than 1. The method according to any one of claims 32 to 43, wherein, The performance metrics information includes information obtained based on at least one of the following: squared generalized cosine similarity (SGCS) and minimum mean square error (MMSE). The method according to any one of claims 32 to 44, the method further comprising: If the first test performance information of the target AI unit meets the first performance requirement, but the second test performance information of the target AI unit does not meet the second performance requirement, the first device performs at least one of the following: Collect a third dataset; Send a second instruction message, which is used to instruct the collection of a third dataset; The third dataset includes data from terminals that provide CSI feedback based on the target AI unit or data from terminals of a first type, wherein the first type is the type of terminal that provides CSI feedback based on the target AI unit. The method according to claim 45, further comprising at least one of the following: The first device sends an updated second dataset; The first device sends an updated target AI unit. The method according to any one of claims 32 to 46, the method further comprising: The first device receives the fifth CSI feedback information; The first device obtains the second monitoring result information of the target AI unit based on the fifth CSI feedback information. The method according to claim 47, further comprising: Receive second real CSI information; The first device obtains the second monitoring result information of the target AI unit based on the fifth CSI feedback information, including: The first device obtains the second monitoring result information of the target AI unit based on the fifth CSI feedback information and the second real CSI information. The method according to claim 47 or 48, further comprising: If the first test performance information of the target AI unit meets the first performance requirement, and the second test performance information of the target AI unit meets the second performance requirement, then if the second monitoring result information of the target AI unit does not meet the third performance requirement, the first device shall perform at least one of the following: Collect the fourth dataset; Send a third instruction message, which is used to instruct the collection of a fourth dataset; The fourth dataset includes data under the first scenario or data under the first data distribution, where the first scenario is the scenario corresponding to the fifth CSI feedback information, and the first data distribution is the data distribution corresponding to the fifth CSI feedback information. An AI unit processing device includes: A first processing module is configured to perform at least one of a first operation and a second operation; The first operation includes: Obtain a first dataset, which includes first target channel state information (CSI) information; First information is obtained based on the first target CSI information and the target artificial intelligence (AI) unit; The first information includes at least one of the following: First CSI feedback information; The first indicator information is the performance indicator information of the target AI unit on the first dataset. The first test performance information of the target AI unit; First failure reason information, which is used to indicate the reason for the failure of the target AI unit; The second operation includes: Obtain a second dataset, which includes the first channel measurement information; The second information is obtained based on the first channel measurement information and the target AI unit; The second information includes at least one of the following: Second CSI feedback information; The first monitoring result information of the target AI unit; The second indicator information is the performance indicator information of the target AI unit on the second dataset. The second test performance information of the target AI unit; The second failure reason information is used to indicate the reason for the failure of the target AI unit. An AI unit processing device includes: The second processing module is used to perform at least one of the fourth and fifth operations; The fourth operation includes at least one of the following: Send a first dataset, which includes first target CSI information; Send a second dataset, which includes the first channel measurement information; The fifth operation includes at least one of the following: receiving fifth information, receiving sixth information; The fifth piece of information includes at least one of the following: First CSI feedback information; The first indicator information is the performance indicator information of the target AI unit on the first dataset. The first test performance information of the target AI unit; First failure reason information, which is used to indicate the reason for the failure of the target AI unit; The sixth piece of information includes at least one of the following: Second CSI feedback information; The first monitoring result information of the target AI unit; The second indicator information is the performance indicator information of the target AI unit on the second dataset. The second test performance information of the target AI unit; The second failure reason information is used to indicate the reason for the failure of the target AI unit. A terminal includes a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the AI ​​unit processing method as described in any one of claims 1 to 31. A first device includes a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the AI ​​unit processing method as claimed in any one of claims 32 to 49. A readable storage medium storing a program or instructions that, when executed by a processor, implement the steps of the AI ​​unit processing method as described in any one of claims 1 to 31, or implement the steps of the AI ​​unit processing method as described in any one of claims 32 to 49.

Images