WO2025232707A1 - Csi data processing method and apparatus, terminal, network side device, medium, and product - Google Patents

Abstract

The present application relates to the technical field of communications, and discloses a channel state information (CSI) data processing method and apparatus, a terminal, a network side device, a medium, and a product. The method in embodiments of the present application comprises: a target node acquires first information, wherein the first information comprises at least one of the following information associated with a target artificial intelligence (AI) unit: information used for grouping CSI data; information for performing domain transformation on the CSI data; and information for combining the CSI data, and the target AI unit comprises at least one of the following: a first AI unit for acquiring target CSI by a terminal; a second AI unit for acquiring the reconstructed CSI by a network side device; a reference AI unit for acquiring the target CSI or the reconstructed CSI by the terminal or the network side device; a third AI unit used by the terminal, the network side device or a test device during testing; and a fifth AI unit for matching a fourth AI unit by the terminal, the network side device or the test device during testing.

Description

CSI data processing methods, devices, terminals, network-side equipment, media, and products

Cross-reference to related applications

This application claims priority to Chinese Patent Application No. 202410578172.0, filed on May 10, 2024, the entire contents of which are incorporated herein by reference.

Technical Field

This application belongs to the field of communication technology, specifically relating to a CSI data processing method, apparatus, terminal, network-side equipment, medium, and product.

Background Technology

Information theory dictates that accurate channel state information (CSI) is crucial for channel capacity. This is especially true for multi-antenna systems, where the transmitter can optimize signal transmission based on CSI to better match the channel conditions. For example, the channel quality indicator (CQI) can be used to select a suitable modulation and coding scheme (MCS) for link adaptation; the precoding matrix indicator (PMI) can be used to achieve eigenbeamforming, maximizing the strength of the received signal, or to suppress interference (such as inter-cell interference or multi-user interference). Therefore, since the introduction of multi-input multi-output (MIMO) technology, CSI acquisition has been a hot research topic.

Currently, artificial intelligence (AI) has been widely used in various fields, and the integration of AI with the communications field is deepening.

In some related technologies, the transmission and processing of CSI data, especially the transmission and processing of CSI data performed by AI units, usually involves directly performing transmission and processing on the CSI data, which results in poor transmission and processing performance of CSI data.

Summary of the Invention

This application provides a CSI data processing method, apparatus, terminal, network-side equipment, medium, and product that can solve the problem of poor CSI data transmission and processing performance.

In a first aspect, a method for processing Channel State Information (CSI) data is provided, executed by a target node. The method includes: the target node acquiring first information, the first information including at least one of the following information associated with a target artificial intelligence (AI) unit:

Information used to group CSI data;

Information used for domain transformation of CSI data;

Information used to combine CSI data;

The target AI unit includes at least one of the following:

The terminal is used to acquire the first AI unit of the target CSI;

Network-side devices are used to acquire the second AI unit for reconstructing CSI;

A reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI;

The third AI unit used in testing by the terminal, network-side equipment, or testing equipment;

The terminal, network-side device, or test equipment is used to match the fifth AI unit of the fourth AI unit used in the test;

The CSI data refers to CSI data from one or more time slots.

Secondly, a CSI data processing apparatus is provided, comprising:

The acquisition module is configured to acquire first information, which includes at least one of the following information associated with the target artificial intelligence (AI) unit:

Information used to group CSI data;

Information used for domain transformation of CSI data;

Information used to combine CSI data;

The target AI unit includes at least one of the following:

The terminal is used to acquire the first AI unit of the target CSI;

Network-side devices are used to acquire the second AI unit for reconstructing CSI;

A reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI;

The third AI unit used in testing by the terminal, network-side equipment, or testing equipment;

The terminal, network-side device, or test equipment is used to match the fifth AI unit of the fourth AI unit used in the test;

The CSI data refers to CSI data from one or more time slots.

Thirdly, a CSI data processing apparatus is provided, the apparatus being configured to perform the steps of the method described in the first aspect.

Fourthly, 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.

Fifthly, a terminal is provided, including a processor and a communication interface, wherein the processor or communication interface is used to acquire first information, the first information including at least one of the following information associated with a target artificial intelligence (AI) unit:

Information used to group CSI data;

Information used for domain transformation of CSI data;

Information used to combine CSI data;

The target AI unit includes at least one of the following:

The terminal is used to acquire the first AI unit of the target CSI;

Network-side devices are used to acquire the second AI unit for reconstructing CSI;

A reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI;

The third AI unit used in testing by the terminal, network-side equipment, or testing equipment;

The terminal, network-side device, or test equipment is used to match the fifth AI unit of the fourth AI unit used in the test;

The CSI data refers to CSI data from one or more time slots.

In a sixth aspect, a network-side device is provided, the network-side 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 first aspect.

In a seventh aspect, a network-side device is provided, including a processor and a communication interface, wherein the processor or communication interface is used to acquire first information, the first information including at least one of the following information associated with a target artificial intelligence (AI) unit:

Information used to group CSI data;

Information used for domain transformation of CSI data;

Information used to combine CSI data;

The target AI unit includes at least one of the following:

The terminal is used to acquire the first AI unit of the target CSI;

Network-side devices are used to acquire the second AI unit for reconstructing CSI;

A reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI;

The third AI unit used in testing by the terminal, network-side equipment, or testing equipment;

The terminal, network-side device, or test equipment is used to match the fifth AI unit of the fourth AI unit used in the test;

The CSI data refers to CSI data from one or more time slots.

In an eighth 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.

A ninth aspect provides a wireless communication system, comprising: a terminal and a network-side device, wherein the terminal is configured to perform the steps of the method described in the first aspect, and/or the network-side device is configured to perform the steps of the method described in the first aspect.

In a tenth aspect, a chip is provided, the chip including a processor and a communication interface coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect.

Eleventhly, a computer program/program product is provided, the computer program/program product being stored in a storage medium, the computer program/program product being executed by at least one processor to perform the steps of the method as described in the first aspect.

In this embodiment, the target node acquires first information, which includes at least one of the following information associated with the target artificial intelligence (AI) unit: information for grouping CSI data; information for performing domain transformation on CSI data; and information for combining CSI data. The target AI unit includes at least one of the following: a first AI unit used by a terminal to acquire the target CSI; a second AI unit used by a network-side device to acquire and reconstruct the CSI; a reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI; a third AI unit used by a terminal, network-side device, or test device in testing; and a fifth AI unit used by a terminal, network-side device, or test device to match a fourth AI unit used in testing. The CSI data is CSI data from one or more time slots. This embodiment optimizes the CSI data to be transmitted by acquiring at least one of the grouping, domain transformation, and combination information related to the CSI data, thereby improving the transmission performance of CSI.

Attached Figure Description

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

Figure 2 is a schematic diagram of a CSI compression scheme applicable to the embodiments of this application;

Figure 3 is a schematic diagram of a CSI reporting scheme applicable to an embodiment of this application;

Figure 4 is a schematic diagram of another CSI reporting scheme applicable to the embodiments of this application;

Figure 5 is a flowchart of a CSI data processing method provided in an embodiment of this application;

Figure 6 is a flowchart of another CSI data processing method provided in an embodiment of this application;

Figure 7 is a schematic diagram of the CSI data processing effect provided in the embodiments of this application;

Figure 8 is a flowchart of another CSI data processing method provided in an embodiment of this application;

Figure 9a is a flowchart of another CSI data processing method provided in an embodiment of this application;

Figure 9b is a flowchart of another CSI data processing method provided in an embodiment of this application;

Figure 9c is a flowchart of another CSI data processing method provided in an embodiment of this application;

Figure 10 is a flowchart of another CSI data processing method provided in an embodiment of this application;

Figure 11 is a schematic diagram of a CSI data partitioning method provided in an embodiment of this application;

Figure 12 is a schematic diagram of a CSI data processing device provided in an embodiment of this application;

Figure 13 is a schematic diagram of a communication device provided in an embodiment of this application;

Figure 14 is a schematic diagram of a terminal provided in an embodiment of this application;

Figure 15 is a schematic diagram of a network-side device provided in an embodiment of this application.

Detailed Implementation

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.

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.

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.

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 in the systems and radio technologies mentioned above, as well as in 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) communication systems.

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. 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. In this context, a base station may be referred to as a 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 base station is not limited to specific technical terms. It should be noted that in this application embodiment, only a base station in an NR system is used as an example for introduction, and the specific type of base station is not limited.

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), and Unified Data Warehouse (UDM). The core network equipment includes: Data Repository (UDR), Home Subscriber Server (HSS), Centralized Network Configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (or L-NEF), Binding Support Function (BSF), 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 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 changes in subsequent protocol versions (e.g., 6G), it will still be within the scope of protection of this application.

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).

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

Artificial intelligence (AI) has been widely applied in various fields. AI models 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 in some embodiments, but it does not limit the specific type of AI model.

A neural network consists of neurons, typically with _a1 , _a2, ..., _aK as inputs, w as weights (multiplicative coefficients), b as biases (additive coefficients), and σ(.) as the activation function. Common activation functions include sigmoid, tanh, rectified linear function, or rectified linear unit (ReLU).

The parameters of a neural network are optimized using optimization algorithms. An optimization algorithm is a class of algorithms that minimizes or maximizes an objective function (sometimes called a loss function). The objective function is often a mathematical combination of model parameters and data. For example, given data X and its corresponding label Y, a neural network model f(.) can be built. With the model, the predicted output f(x) can be obtained from the input x, and the difference between the predicted value and the true value (label) (f(x) - Y) can be calculated; this is the loss function. The goal of neural network training is to find a suitable W (a vector of weights w) b that minimizes the value of the aforementioned loss function. The smaller the loss value, the closer the model is to the reality.

Most common optimization algorithms are based on error back propagation (BP). The basic idea of BP 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 error back propagation. Error back propagation involves propagating the output error back to the input layer through the hidden layers in a certain form, 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 cyclical. 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. Common optimization algorithms include Gradient Descent, Stochastic Gradient Descent (SGD), Mini-batch Gradient Descent, Momentum, Nesterov (named after its inventor, specifically referring to stochastic gradient descent with momentum), Adaptive Gradient Descent (Adagrad), Adadelta, Root Mean Square Proportional (RMSprop), and Adaptive Moment Estimation (Adam). During error backpropagation, these algorithms calculate the gradient by taking the derivative/partial derivative of the error/loss obtained from the loss function with respect to the current neuron, adding the learning rate and the effects of previous gradients/derivatives/partial derivatives, and then propagating this gradient to the previous layer.

In this application embodiment, the AI unit/AI model may also be referred to as an AI unit, AI model, 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 AI unit/AI model may refer to a processing unit capable of implementing specific algorithms, formulas, processing flows, capabilities, etc., related to AI. Or, the AI unit/AI model may be a processing method, algorithm, function, module, or unit for a specific dataset. Alternatively, the AI unit/AI model may be a processing method, algorithm, function, module, or unit running on AI/ML related hardware such as a graphics processing unit (GPU), neural network processing unit (NPU), tensor processing unit (TPU), or application-specific integrated circuit (ASIC). This application does not specifically limit this. Optionally, the specific dataset includes the input and/or output of the AI unit/AI model.

Optionally, the identifier (identification information) 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, characteristic, capability, or module related to the AI/ML. This application embodiment does not specifically limit this.

This application relates to the application of Channel State Information (CSI) compression in this embodiment. For ease of understanding, some relevant aspects of CSI are introduced below.

Typically, access network equipment, taking a base station as an example, transmits Channel State Information Reference Signals (CSI-RS) on certain time-frequency resources within a specific time slot. The terminal performs channel estimation based on the CSI-RS, calculates the channel information for that slot, and feeds back the codebook information to the base station via PMI. The base station then combines the codebook information fed back by the terminal to assemble the channel information. Before the next CSI report, the base station uses this information for data precoding and multi-user scheduling. Here, PMI is a part of the CSI data.

To further reduce CSI feedback overhead, the terminal can change the PMI reporting for each sub-band to reporting PMI according to the delay. Since the channels in the delay domain are more concentrated, the PMI of all sub-bands can be approximated with PMIs of less delay, that is, the delay domain information is compressed before reporting. Similarly, to reduce overhead, the base station can pre-encode the CSI-RS and send the encoded CSI-RS to the terminal. The terminal sees the channel corresponding to the encoded CSI-RS. The terminal only needs to select a few ports with high strength from the ports indicated by the network side (for example, a channel with port 32) and report the coefficients corresponding to these ports.

Furthermore, to better compress channel information, neural networks or machine learning methods can be used.

Specifically, the terminal compresses and encodes the channel information, and the base station decodes the compressed content to recover the channel information. At this stage, the base station's decoding network and the terminal's encoding network need to be jointly trained to achieve a reasonable matching degree. The neural network is a joint neural network composed of the terminal's encoder and the base station's decoder, and is jointly trained by the network side. After training, the base station sends the encoder network to the terminal. During inference (the application phase of the model), the terminal estimates CSI-RS, calculates the channel information, and uses the calculated channel information or the original estimated channel information to obtain the encoding result through the encoding network. The encoded result is then sent to the base station, which receives the encoded result and inputs it into the decoding network to recover the channel information.

CSI compression is a typical two-end model use case, meaning the complete CSI compression model needs to be deployed on different communication nodes. Currently, most considerations involve deploying the encoder on the user terminal (UE) and the decoder on the network (NW). The (sub)models deployed on multiple nodes need to be paired with each other to function correctly. Considering the characteristics of two-end models, the protocol identifies several basic training collaboration types for AI/ML CSI compression models:

1) Joint training at a single entity (or type 1)

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

2) Joint training at multiple entities (or type 2)

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.

3) Separate training on multiple nodes (or type 3)

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 trains its own model based on this information, ensuring that each node (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 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 input and output data of the NW-side model) 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 input and output data of the UE-side model) to the UE side.

An example of the AI-based CSI/PMI compression process is shown in Figure 2. The UE's expected CSI, target CSI, or codebook W _A*B (where A is the number of CSI ports and B is the number of subbands) is compressed using AI, such as into an AI-based PMI value, and then reported to the network-side device. The network-side device performs decompression to obtain W′ _A*B .

Optionally, this embodiment of the application introduces the utilization of time-domain CSI correlation based on spatial frequency domain CSI compression. That is, CSI from multiple slots can be compressed together, thereby further reducing the overhead of CSI reporting or improving the accuracy of CSI reporting. For example, as shown in Figure 3, CSI from four slots are jointly compressed and reported, and the CSI from each slot can be regarded as a spatial frequency domain CSI report, where the internal information stream corresponds to the output information of the encoder intermediate node.

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 Figure 4), while progressive reporting is to report CSI on each slot sequentially in an autoregressive manner (as shown in Figure 3).

The following description, in conjunction with the accompanying drawings, details the CSI data processing method, apparatus, terminal, network-side equipment, medium, and product provided in this application through some embodiments and application scenarios.

Referring to Figure 5, which is a flowchart of a CSI data processing method provided in an embodiment of this application, for a target node, the method includes the following steps:

Step 501: The target node obtains first information, which includes at least one of the following information associated with the target artificial intelligence (AI) unit:

Information used to group CSI data;

Information used for domain transformation of CSI data;

Information used to combine CSI data;

The target AI unit includes at least one of the following:

The terminal is used to acquire the first AI unit of the target CSI;

Network-side devices are used to acquire the second AI unit for reconstructing CSI;

A reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI;

The third AI unit used in testing by the terminal, network-side equipment, or testing equipment;

The terminal, network-side device, or test equipment is used to match the fifth AI unit of the fourth AI unit used in the test;

The CSI data refers to CSI data from one or more time slots.

In this embodiment, the target node can be a terminal, a network-side device, a core network device, or a test device. In some embodiments, the target node can also be any one of a data compression device, a data decompression device, a data encoding device, a data decoding device, a data compression-decompression device, or a data encoding-decoding device. The aforementioned data compression device, data decompression device, data encoding device, data decoding device, data compression-decompression device, and data encoding-decoding device can be a terminal, a network-side device, a core network device, or a test device.

It is understood that if the data compression device obtains the second AI unit of the reconstructed CSI or the reference AI unit of the reconstructed CSI, it determines its own AI unit for compression based on the AI unit of the reconstructed CSI.

It is understood that the target node can determine the AI unit for the final application based on the reference AI unit. In some embodiments, the target node directly uses the reference AI unit; in other embodiments, the target node can perform operations such as pruning and quantization on the reference AI unit to obtain the AI unit for the final application.

It is understood that CSI data can be a channel matrix obtained from measuring signals, a codebook matrix, or data obtained by quantizing the channel matrix or codebook matrix. The quantization can be based on a larger frequency domain granularity or can be quantization using the eType II reporting method.

In this embodiment of the application, reconstruction can also be described as rebuilding.

The target node can obtain the first information based on protocol agreements or preset rules; or it can obtain it by receiving the first information sent by the peer device. For example, when the target node is a terminal, the terminal can receive the first information sent by the network-side device, and when the target node is a network-side device, the network-side device can receive the first information sent by the terminal.

For example, the target AI unit is specified in the protocol, and the target node obtains some or all of the first information according to the target AI unit specified in the protocol.

For example, the target AI unit is specified in the protocol, but it is still necessary to obtain relevant information through signaling. The target node obtains some first information based on the target AI unit specified in the protocol and obtains some first information based on signaling.

Regarding obtaining the first information, optionally, in the case where at least one AI unit is defined in the standard, the target node in this application embodiment may be to obtain at least one AI unit defined in the standard, and obtaining the target AI unit may include obtaining relevant information of the target AI unit through signaling.

Regarding obtaining the first information, optionally, when the standard defines the model structure and all parameters related to the model structure are fixed, obtaining the target AI unit can be understood as pre-generating the target AI unit according to the standard.

Regarding obtaining the first information, optionally, where the standard defines a model structure and the parameters of the model structure are variable (e.g., training is required), obtaining the first information can be understood as including the reference model structure and the obtained parameters to determine the target AI unit.

The target AI unit associated with the first piece of information mentioned above can be on the terminal side, the network side device side, or the test device side.

In this embodiment, the CSI data can also be described as channel data or codebook data. It is understood that CSI data can be a channel matrix obtained from measuring signals, a codebook matrix, or data obtained by quantizing a channel matrix or codebook matrix. The quantization can be based on a larger frequency domain granularity or can be quantization using the eType II reporting method.

Optionally, the CSI data for the one or more time slots includes at least one of the following:

First CSI data;

CSI data from multiple time slots to be grouped;

CSI data from one or more time slots to be domain transformed;

Type 1 CSI data;

Second type of CSI data;

Wherein, the first CSI data is CSI data acquired in the first time slot, and the first time slot is the time slot of the most recent measured CSI-RS, the time slot used to generate CSI reporting correlation, or the time slot of predicted CSI correlation;

The first type of CSI data includes at least one of first CSI data and second CSI data, wherein the second CSI data is one or more CSI data acquired prior to the first CSI data;

The second type of CSI data is related data of the first type of CSI data, wherein the related data of the first type of CSI data is data obtained by processing the first CSI data and/or the second CSI data through a first preset process, and/or the second type of CSI data includes data obtained by processing the related data of the first CSI data and/or the second CSI data through a second preset process.

In this embodiment of the application, the acquisition of the above-mentioned CSI data can be understood as the measured CSI data (ground truth CSI). For example, it can refer to the actual measured CSI true value, or the value of the actual measured CSI true value after data processing (e.g., quantization processing). The quantization processing can be based on a larger frequency domain granularity, or it can be quantization using the eType II reporting method.

The first type of CSI data mentioned above can be understood as the acquired CSI data, and the second type of CSI data can be understood as the data after the acquired CSI data has undergone a first preset processing. The first preset processing may include processing performed by the target AI unit or processing by other algorithms.

Optionally, the first CSI data includes at least one of the following: CSI data at the predicted time point, and CSI data at the measured time point;

The second CSI data includes at least one of the following: CSI data at the predicted time point, and CSI data at the measured time point.

In this embodiment, the CSI data can be either measured CSI data or predicted CSI data. In an optional embodiment, if the first CSI data is a predicted CSI, it can be understood that the CSI associated with the first time slot is a predicted CSI.

In this embodiment, the first type of CSI data includes at least one of first CSI data and second CSI data, and the second type of CSI data is related data of the first type of CSI data. This can be understood as the second type of CSI data including at least one of related data of the first CSI data and related data of the second CSI data. Specifically, the related data of the first CSI data is data obtained by processing the first CSI data and/or at least one CSI data previously acquired before the first CSI data through a first preset processing step; the related data of the second CSI data is data obtained by processing the second CSI data and/or at least one CSI data previously acquired before the second CSI data through a second preset processing step.

Optionally, the second type of CSI data includes at least one of the following:

The cached data after the first CSI data and/or the second CSI data have undergone the first preset processing;

The first CSI data and/or the second CSI data are cached data after passing through a preset unit of the target AI unit;

The data output by the sub-unit of the target AI unit obtained by combining the first CSI data and/or the second CSI data;

The final layer output data of the target AI unit obtained by combining the first CSI data and/or the second CSI data;

The cached data after the first CSI data and/or the related data of the second CSI data have undergone the first preset processing;

The relevant data of the first CSI data and/or the second CSI data are cached data after passing through the preset unit of the target AI unit;

The data output by the subunit of the target AI unit is obtained by combining the relevant data of the first CSI data and/or the second CSI data;

The final layer output data of the target AI unit is obtained by combining the first CSI data and/or the relevant data of the second CSI data;

The related data of the second CSI data is data obtained by processing at least one CSI data previously acquired before the second CSI data through a second preset process.

In this embodiment, the second type of CSI data can be understood as CSI data after data processing. This data processing can involve processing the CSI data itself, or combining different CSI data. The processing can be a preset processing algorithm, or it can be processed by a target AI unit. Processing by the target AI unit can be done by the entire target AI unit, or by some of its sub-units. These sub-units can also be described as intermediate units or constituent units of the target AI unit. Processing by some sub-units can be the result of cascading processing of multiple sub-units, or it can be the sum of processing results from multiple sub-units, a weighted sum, or a mapping result, etc.

Optionally, the second type of CSI data includes data output by a sub-unit of the target AI unit obtained by combining the first CSI data and/or the second CSI data, and/or, the second type of CSI data includes data output by a sub-unit of the target AI unit obtained by combining relevant data from the first CSI data and/or the second CSI data, and the second type of CSI data includes at least one of the following:

The data output by the transform module, a sub-module of the target AI unit;

The data output by the forward feedback module of the sub-module of the target AI unit;

The data output by the attention layer module, a sub-module of the target AI unit;

The data output by the convolutional feature extraction layer module of the target AI unit;

The output data of multiple cascaded modules of the target AI unit;

The sum of the output data from multiple modules of the target AI unit;

The target AI unit's multiple modules output data mapping data.

Optionally, the second type of CSI data is data obtained by combining the first CSI data and/or the second CSI data and performing a third preset processing on the data output by a subunit of the target AI unit, and/or, the second type of CSI data includes data obtained by combining the first CSI data and/or the second CSI data and performing a third preset processing on the data output by a subunit of the target AI unit, wherein the third preset processing includes at least one of the following:

Cascaded processing; mapping processing; dimensional scaling processing; quantization processing.

Optionally, the first CSI data mentioned above can be the current CSI data.

In this embodiment, the target node acquires first information for at least one of grouping, domain transformation, and combination of CSI data. The information for grouping CSI data may include grouping algorithms, grouping rules, grouping execution modules, and grouping execution parameters; the information for domain transformation of CSI data may include domain transformation algorithms, domain transformation rules, domain transformation execution modules, and domain transformation execution parameters; and the information for combining CSI data may include CSI data combination algorithms, CSI data combination rules, CSI data combination execution modules, and CSI data combination execution parameters.

In this embodiment of the application, the execution of at least one of the above-mentioned grouping, domain transformation and combination of CSI data can be performed by the target AI unit or by other units or modules other than the target AI unit.

The target AI unit can be composed of multiple sub-units, such as transform units, CNN units, RNN units, fully connected units, etc.

The parameters of the target AI unit include one or more of the following:

The type of AI unit, such as transform or fully connected;

The depth of the AI unit, such as the number of layers;

Configuration parameters of the AI unit subunit;

Quantization methods, such as scalar quantization (SQ) or vector quantization (VQ);

The parameters of the target AI unit may also include hyperparameters, which include one or more of the following:

Learning rate;

Loss function;

Batch size;

Regularization techniques and strength;

Optimization algorithm.

Optionally, the first information includes at least one of the following: composition information, model structure, or model parameters of the target AI unit, thereby determining at least one of the following information through the above information:

Information used to group CSI data;

Information used for domain transformation of CSI data;

Information used to combine CSI data;

In other words, at least one of the following is achieved through some or all of the sub-units of the target AI unit:

Group the CSI data;

Domain transformation of CSI data

Combine CSI data;

In some embodiments, combining CSI data can be understood as combining the first CSI data with multiple CSI data preceding it and processing them to obtain the target data. Figures 9b and 9c illustrate one combination method within the target AI model. For example, a mask with a lower triangular matrix structure is introduced on the attention score of the attention subunit of the target AI unit to ensure that CSI data following the first CSI data is not used when processing the first CSI data, and that the second type of CSI data or the target data can be obtained based on the CSI data preceding the first CSI data. It can be understood that the CSI data may include the first CSI data, the CSI data preceding the first CSI data, and the data following the first CSI data. Introducing a mask with a lower triangular matrix structure on the attention score of the attention subunit of the target AI unit ensures that data following the first CSI data is not used when processing the first CSI data.

In some embodiments, the target AI unit includes a subunit for performing domain transformation, thereby enabling domain transformation of the CSI data.

In some embodiments, referring to FIG8, the CSI data can be combined through the model structure of the target AI unit.

In an optional implementation, step 501 can be replaced by the following steps, or the following steps can be performed after step 501:

The target node performs the first operation on the CSI data to obtain the target data;

The first operation includes at least one of the following:

The CSI data from the multiple time slots are grouped;

Perform domain transformation on the CSI data of the one or more time slots;

Combine CSI data from multiple time slots.

In this embodiment of the application, combining CSI data from multiple time slots can be done by combining first data and second data. The first data includes at least one of first CSI data and related data of the first CSI data. The second data includes at least one of second CSI data and related data of the second CSI data. The second CSI data is at least one CSI data acquired before the first CSI data.

In this embodiment of the application, the CSI data of the multiple time slots (e.g., N, where N is a positive integer greater than 1) are grouped to obtain CSI data groups. The number of CSI data included in each CSI data group can be the same or different. For example, each CSI data group includes M CSI data, or the first CSI data group includes _M1 CSI data and the second CSI data group includes _M2 CSI data, where _M1 ≠ _M2 .

In this embodiment of the application, by grouping the data from N slots, the grouped data can be processed together, thereby improving the data processing efficiency.

Optionally, after grouping, at least one data group in the resulting CSI data group includes CSI data from multiple time slots.

Optionally, M is a positive integer greater than 1.

Optionally, the information used for grouping CSI data includes at least one of the following:

The CSI data from N time slots are grouped to obtain K CSI data groups; where N and K are positive integers greater than 1, and adjacent CSI data groups include CSI data from at least one of the same time slots.

The CSI data from the first time slot is combined with the M-1 CSI data preceding the first time slot to form a CSI data group.

In other words, the grouping of the CSI data from the multiple time slots includes at least one of the following:

The CSI data from N time slots are grouped to obtain K CSI data groups; where N and K are positive integers greater than 1, and adjacent CSI data groups include CSI data from at least one time slot.

The CSI data from the first time slot is combined with the M-1 CSI data preceding the first time slot to form a CSI data group.

In this embodiment of the application, adjacent CSI data groups include CSI data in at least one identical time slot. For example, as shown in FIG6, each CSI data group includes CSI data in four time slots. The first CSI data group and the second CSI data group include CSI data in three identical time slots (slot2 channel data, slot3 channel data, and slot4 channel data), the second CSI data group and the third CSI data group include CSI data in three identical time slots (slot3 channel data, slot4 channel data, and slot5 channel data), and so on.

It is understandable that the number of CSI data points with the same time slots included in different adjacent CSI data groups can be different. For example, the first CSI data group and the second CSI data group may include M ₃ CSI data points with the same time slots, and the second CSI data group and the third CSI data group may include M ₄ CSI data points with the same time slots, where M ₃ ≠ M ₄ .

Optionally, each CSI data group includes CSI data from M time slots, where 1 < M < N and K ≤ N - M + 1.

In the embodiments of this application, the domain transformation can be performed after grouping, or it can be performed on CSI data of one or more time slots.

Optionally, in this embodiment, information for performing domain transformation on CSI data is obtained to perform domain transformation (e.g., Fourier transform) on the CSI data related to AI processing. The transformed channel has good sparsity, thereby enabling the AI unit to have better processing performance. For example, as shown in Figure 7, part (a) of Figure 7 shows the case without domain transformation, where the channel is relatively oscillating. Part (b) of Figure 7 shows the case after domain transformation, where the transformed channel has good sparsity, allowing the AI unit to identify it better.

Optionally, the information used for domain transformation of CSI data includes at least one of the following:

Perform domain transformation on a per-CSI data set basis;

Based on the second indication information or the number of time slots agreed upon in the protocol, perform domain transformation on the CSI data of one or more time slots.

In this embodiment of the application, the above-mentioned domain transformation can be performed on a grouping basis, or it can be performed based on the number of time slots agreed upon by the indication information or protocol.

For ease of understanding, the following example illustrates the embodiments of this application by performing grouping and/or domain transformation on N slots (e.g., channel or codebook data).

a) Divide the channel or codebook data of the N slots into K groups, each group including the channel or codebook data of M slots:

Input 4*13*32*2: 4 (M, number of slots), 13 (number of sub-bands), 32 (antennas or ports), 2 (channels); it can be understood that 4, 13, 32, and 2 are examples of the number of slots, number of sub-bands, number of ports, and number of channels, and can also be other optional values.

Optionally, the data of M slots can be transformed into data of Z ₁ channels, such as reshaping to: 13*32*(8 channels), turning 4 slots into 8 channels;

Optionally, the N slots (e.g., channel or codebook data) are divided into K groups, and the i-th group and the (i+1)-th group of data include at least one set of data from the same slot;

In an optional embodiment, the i-th group and the (i+1)-th group of data include data from M-1 identical slots, as shown in Figure 6. Each group includes a channel with 4 time slots, and adjacent groups include data from 3 identical time slots.

In an optional embodiment, K <= N-M+1.

b) Perform a Fourier transform on the data from M slots (e.g., channel or codebook data) or a set of data (e.g., channel or codebook data) to the second domain space (e.g., antenna domain or Doppler domain, etc.).

If the input data from M slots is subjected to a Fourier transform and transformed to the Doppler domain, the data obtained is 4*13*32*2(4(M, number of slots), 13(number of sub-bands), 32(antennas), 2(channels)). It can be understood that 4, 13, 32, and 2 are examples of the number of slots, number of sub-bands, number of ports, and number of channels, and can also be other optional values.

Optionally, the data transformed to the Doppler domain can be reshaped into _Z2 channel data, such as reshaping it to: 13*32*(8 channels), turning 4 slots into 8 channels;

c) Following a) and/or b), the method further includes performing a first AI process on the data after the first operation, such as grouped data, or grouped data after Doppler transformation, as shown in Figure 2. The input is: WA _*B is the WA _*B data of M slots, the PMI value is the PMI value after data compression of the WA _*B data of M slots, and the output is the decoded W′A _*B corresponding to the last time slot after M slots.

Optionally, the information used to combine the CSI data indicates at least one of the following:

The CSI data from the multiple time slots are weighted and summed.

The CSI data from the multiple time slots are input into the target AI unit;

The first CSI data and the second type of CSI data are input into different sub-units of the target AI unit;

The second type of CSI data is input into the target AI unit after the sub-unit of the target AI unit into which the first CSI data is input;

The second type of CSI data is input into the sub-unit of the target AI unit.

In this embodiment of the application, the weighted summation of the CSI data of the plurality of time slots can be performed by at least one of the following: weighted summation of the first CSI data and the second CSI data; weighted summation of the related data of the first CSI data and the second CSI data; weighted summation of the related data of the first CSI data and the second CSI data; and weighted summation of the related data of the first CSI data and the related data of the second CSI data.

Similarly, inputting CSI data from multiple time slots into the target AI unit can be achieved by inputting the first CSI data, the second CSI data, related data of the first CSI data, and related data of the second CSI data into the target AI unit.

In this embodiment of the application, CSI data can be input into different sub-units of the AI unit.

Optionally, inputting the first CSI data and the second type of CSI data into different sub-units of the target AI unit includes:

The first CSI data is input into the first sub-unit of the target AI unit.

The second type of CSI data is input into the second sub-unit of the target AI unit.

The second subunit is the subunit following the first subunit in the target AI unit.

Optionally, the weighted summation of the CSI data from the multiple time slots includes:

The first type of CSI data and/or the second type of CSI data are weighted and summed before the third sub-unit of the target AI unit;

The third sub-unit includes at least one of the following: a transform module, a forward feedback module, an attention layer module, and a convolutional feature extraction layer module.

Optionally, the sub-unit includes at least one of the following: a transform module, an attention layer module, a feedforward module, and a convolutional feature extraction layer module.

In this embodiment of the application, the relevant data of the first CSI data can be input before the transform module of the sub-module of the target AI unit used to obtain the target second CSI data or the relevant data of the second CSI data;

The relevant data of the first CSI data is input before the attention layer module of the submodule of the target AI unit used to acquire the target second CSI data or related data of the second CSI data;

The relevant data of the first CSI data is input before the forward feedback module of the submodule of the target AI unit used to acquire the target second CSI data or related data of the second CSI data;

The relevant data of the first CSI data is input before the convolutional feature extraction layer of the submodule of the target AI unit used to obtain the target second CSI data or the relevant data of the second CSI data.

For example, as shown in Figure 8, CSI data (e.g., CSI for slot 1) and related data from its preceding slot CSI data (CSI for slot 0) can be input into different sub-modules of the target AI unit and then combined to obtain related data from CSI data (CSI for slot 1) for data accumulation. It is understood that slot 0 or slot 1 describes the order of the CSI data slots and does not necessarily specifically refer to slots 0 or 1.

Optionally, the accumulated information of the N-1 slots can be used as an intermediate input for calculating the CSI feedback of the Nth slot.

For a more specific example, referring to part (a) of Figure 8, the transform output of the (N-1)th slot can be accumulated to the input of the attention layer of the Nth slot (optionally, the transform output of the (N-1)th slot is accumulated by a coefficient and then accumulated to the Nth slot; alternatively, the input of the current slot also needs to be multiplied by a coefficient).

Referring to part (b) of Figure 8, the output of the attention layer of the (N-1)th slot can be accumulated to the input of the attention layer of the Nth slot (optionally, the output of the attention layer of the (N-1)th slot is accumulated and multiplied by a coefficient to the Nth slot; alternatively, the input of the current slot also needs to be multiplied by a coefficient).

Referring to part (c) of Figure 8, the output of the feedforward layer of the N-1th slot can be accumulated to the input of the attention layer of the Nth slot (optionally, the output of the attention layer of the N-1th slot is accumulated and multiplied by a coefficient to the Nth slot; optionally, the input of the current slot also needs to be multiplied by a coefficient).

It is understood that, referring to Figure 8, in this embodiment of the application, the cumulative intermediate information of the (N-1)th slot is also obtained based on the cumulative intermediate information of the (N-2)th slot and the CSI data of the (N-1)th slot. Therefore, the cumulative intermediate information of the (N-1)th slot is a cumulative sum of the intermediate information of the aforementioned multiple slots.

It is worth noting that the description is based on the encoding side (compression side) model structure, and the acquisition of reconstructed CSI follows a similar approach. That is, optionally, the accumulated intermediate information of the N-1 slots can be used as an intermediate input for calculating the CSI reconstruction information of the Nth slot.

In this embodiment of the application, the above combination method can be understood as using the codebooks of multiple slots to obtain the codebook compression CSI feedback information of the current slot. The AI model (wherein the AI model can be a reference model implemented by the UE, or a model used for testing or alignment between the terminal side and the network side, etc.) is shown in Figure 8 above. The AI model includes, but is not limited to:

• The Transform module (including the attention module and the forward feedback module);

• Fully connected model (optional: include a fully connected model before and/or after the transform);

• Quantization module;

Intermediate information across slots.

It is understood that the embodiments of this application are not limited to the Transform model shown in Figure 8, and other AI models such as the CNN model (as shown in Figure 9a) can also be used. As shown in Figure 9a, CSI compression of the spatiotemporal frequency domain TSF can be achieved using a fully connected CNN, and the intermediate information of slot k is sent to slot k+1: Here, it can be understood that slot k+1 is a temporal description of a CSI data after slot k, and the temporal CSI data can be several slots after slot k, or it can be understood as the next CSI data obtained after slot k.

The intermediate information accumulated info of slot k is the information obtained through the mapping layer after the input and output of the convolutional feature extraction layer are concatenated.

The input to the slot k+1 convolutional layer is the sum of the information of slot k and the accumulated info (or multiplied by a certain proportion and added).

In this embodiment, by utilizing the intermediate information of the preceding slot, the information of the current slot can be extracted more effectively, allowing the AI model to process it more efficiently.

Another example, a CSI data processing example, is shown in Figure 10:

First, the input is fed into a convolutional long short-term memory network (cov LSTM) with 8 channels, and then the output has 8 channels.

Then, it is compressed to 2 channels through convolutional cov, and output as a fully connected FC 16-dimensional output with 4-bit quantization per dimension.

In this embodiment of the application, when M CSI data are processed into one processing unit, the output of the decoder is the codebook of the last slot of the input side. This can be understood as the dimension of the input side being M times the dimension of the parsing side, or as the input side needing to perform a Doppler transformation, while the output side not needing to transform back from the Doppler domain.

In some embodiments, the AI unit at the compression end includes a Doppler transformation module, while the decompression end does not have a Doppler transformation module.

Optionally, the information used to combine the CSI data includes:

Adjust the CSI data to be combined to be data of the same dimension.

In this embodiment of the application, optionally, the data to be combined, such as the two sums, have the same dimension, such as the transform output of N-1 slots and the input of the attention layer are both Z*L dimensional.

Optionally, in the embodiments of this application, information for combining CSI data is obtained so that CSI data from multiple time slots can be used together for CSI processing. This can better utilize historical information or information from other data to help CSI in a specific time slot, thereby optimizing the performance of CSI processing.

In this embodiment of the application, for CSI compression, the correlation method that obtains the grouping and domain transformation information of CSI data is more suitable for the packing compression method, while the correlation method that obtains the combination information of CSI data is more suitable for the progressive compression method.

Optionally, the first information may further include at least one of the following:

The target AI unit shall have at least one of the following: the model structure of the subunit used to group the CSI data or the parameter information of the model subunit;

The target AI unit has at least one of the model structure or model parameter information of the subunit used to perform domain transformation on the CSI data;

The target AI unit includes at least one of the model structure or model parameter information of the subunit used to combine the CSI data.

In this embodiment, the target AI unit performs at least one of the following operations: grouping, domain transformation, and combination of CSI data. The first information obtained by the target node can indicate the model structure and/or parameter information of the target AI unit sub-unit performing the above operations. It can be understood that, based on the aforementioned first information, the target node can determine the architecture of the target AI unit performing the grouping, domain transformation, and/or combination operations of the CSI data.

It is understandable that when the first information obtained by the target node is sent by the peer device, or when the first information obtained is the model structure or parameter information of the AI unit of the peer device, the target node can construct or determine the AI unit structure or parameters required by the target node itself based on the first information provided by the peer device or the first information related to the peer device.

Optionally, the model structure or model parameter information of the sub-unit in the target AI unit used for combining CSI data includes at least one of the following:

The sub-unit is located in the target AI unit;

The type of the subunit;

The parameters of the subunit;

The connection method of the subunit;

The manner in which the CSI data is input into the sub-unit.

Optionally, the first information further includes first indication information, which indicates at least one of the following:

Whether to apply the sub-unit in the target AI unit used for grouping CSI data;

Whether to apply the sub-unit in the target AI unit used for domain transformation of CSI data;

Whether to apply the sub-unit in the target AI unit used for combining CSI data.

In this embodiment of the application, the target node determines whether to perform grouping, domain transformation and/or combination operations using the corresponding sub-unit in the target AI unit based on the first indication information.

Optionally, the first information may further include at least one of the following:

The grouping information of the CSI data;

The time slot information associated with the CSI data;

The type information of the CSI data.

In this embodiment of the application, the grouping information of the CSI data may include CSI data grouping rules, grouping results, and other information. The time slot information associated with the CSI data may include the time slot data associated with the CSI data, window parameters, and other information. The type information of the CSI data may include first type CSI data information and/or second type CSI data information.

Optionally, the method further includes:

The target node determines target data based on the first information and the target AI unit; wherein the target data includes at least one of the following:

Input data for the target artificial intelligence (AI) unit;

Output data of the target AI unit;

Target CSI data;

Relevant data for the target CSI data.

In this embodiment, based on the first information and the target AI unit, the output data or target CSI data of the target AI unit can be directly obtained, or the input data or related data of the target AI unit or target CSI data can be obtained. Furthermore, the aforementioned input data or related data of the target AI unit can be further input into the target AI unit for processing to obtain the output data or target CSI data of the target AI unit.

In this context, the target CSI data on the compression end or terminal side can be understood as CSI feedback or data reported by CSI.

In this context, the target CSI data at the decompression end or network side can be understood as reconstructed CSI data.

Optionally, the first information is also used to indicate at least one of the following:

Reshape at least a portion of the grouped CSI data groups into data with Z ₁ channels, where Z ₁ is a positive integer;

The domain-transformed CSI data set is reshaped into _Z2 channels of data, where _Z2 is a positive integer.

In other words, the first operation further includes at least one of the following:

Reshape at least a portion of the grouped CSI data groups into data with Z ₁ channels, where Z ₁ is a positive integer;

The domain-transformed CSI data set is reshaped into _Z2 channels of data, where _Z2 is a positive integer.

In this embodiment of the application, after performing data grouping, the CSI data group can optionally be reshaped to form data with Z ₁ channels. The above reshaping process can facilitate the performance of domain transformation (e.g., Fourier transform) or dimension alignment.

In this embodiment of the application, after performing data domain transformation, the CSI data after domain transformation can be optionally reshaped to obtain data with _Z2 channels (for example, as shown in Figure 6, the data after domain transformation is reshaped into 8-channel data; the number of channels after grouping is not shown in Figure 6). The above reshaping process can facilitate the alignment of data dimensions, thereby facilitating subsequent data processing.

Optionally, if the target data includes at least one of the input data of the target artificial intelligence (AI) unit and the related data of the target CSI data, the method further includes: the target node inputs the target data into the target AI unit to obtain third CSI data, wherein the third CSI data includes at least one of the reported CSI data and the reconstructed CSI data;

Alternatively, if the target data includes at least one of the output data of the target AI unit and the target CSI data, the target data includes at least one of the reported CSI data and the reconstructed CSI data.

In this embodiment of the application, after determining the input data of the target artificial intelligence (AI) unit, the input data can be input into the target AI unit, and the target AI unit can output at least one of the reported CSI data and the reconstructed CSI data. Alternatively, at least one of the reported CSI data and the reconstructed CSI data can be determined directly based on the first information and the target AI unit.

The reported CSI data and the reconstructed CSI data mentioned above can be respectively CSI compressed data and CSI decompressed data.

The embodiments of this application can also be described as follows: when the target data includes at least one of the input data of the target artificial intelligence (AI) unit and the related data of the target CSI data, the method further includes: the target node inputs the target data into the target AI unit to obtain third CSI data, wherein the third CSI data includes at least one of the reported CSI data and the reconstructed CSI data;

And/or, if the target data includes at least one of the output data of the target AI unit and the target CSI data, the target node performs a first operation on the CSI data to obtain the target data, including: inputting the data obtained by performing the first operation into the target AI unit to obtain the target data.

Optionally, when the first information includes information for grouping CSI data, the reconstructed CSI data is the reconstructed CSI data corresponding to the CSI data of the last time slot in each CSI data group;

Alternatively, if the information for grouping CSI data includes forming a CSI data group by combining the CSI data of the first time slot with the M-1 CSI data preceding the first time slot, then the reconstructed CSI data is the reconstructed CSI data corresponding to the first time slot.

Alternatively, if the information used to group the CSI data includes forming a CSI data group by combining the CSI data of the first time slot with the M-1 CSI data preceding the first time slot, the size of the reconstructed CSI data is 1/M times the size of the reconstructed CSI data of the M time slots.

Alternatively, if the first information includes information for performing domain transformation on the CSI data, the reconstructed CSI data is the reconstructed CSI data corresponding to the CSI data of the first time slot;

Alternatively, if the information for performing domain transformation on the CSI data includes comparing the CSI data of the first time slot with the CSI data of the M-1 preceding times slot, the reconstructed CSI data is the reconstructed CSI data corresponding to the first time slot.

Alternatively, if the information for performing domain transformation on the CSI data includes comparing the CSI data of the first time slot with the CSI data of M-1 times preceding the first time slot, the size of the reconstructed CSI data is 1/M times the size of the reconstructed CSI data of the M time slots.

In this embodiment, in the case of grouping, multiple CSIs may correspond to one reconstructed CSI data. Similarly, when multiple CSIs in multiple time slots jointly perform domain transformation, multiple CSIs may also correspond to one reconstructed CSI data. For the encoder-decoder scenario, the size of the input data on the encoder side may be M times the size of the output data on the decoder side, and the output data on the decoder side may be the decompressed data corresponding to the last time slot of the input data on the encoder side. For example, as shown in Figure 6, the input on the encoder side is the slot 1-4 channel data as the first group of CSI data, and the output on the decoder side is the slot 4 channel data.

It is understood that the example of the two-end memory model in Figure 6 is not intended to limit the type of AI model in the embodiments of this application.

Optionally, the CSI data in the first CSI data group are CSI data corresponding to consecutive CSI measurement resources, wherein the first CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data in the second CSI data group are CSI data corresponding to consecutive CSI reporting resources, wherein the second CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data in the third CSI data group is at least partly predicted CSI data, wherein the third CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data in the fourth CSI data group includes the CSI data of the current time slot and at least one predicted CSI data, wherein the fourth CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data in the fifth CSI data group includes CSI data acquired within the target time window, wherein the fifth CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data of the multiple time slots may include CSI data corresponding to continuous CSI measurement resources;

Alternatively, the CSI data of the multiple time slots may include CSI data corresponding to continuously reported CSI resources;

Alternatively, the CSI data in the plurality of time slots may be at least partially predicted CSI data;

Alternatively, the CSI data in the multiple time slots may include at least one predicted CSI data.

Alternatively, the CSI data for the multiple time slots may include the CSI data for the current time slot and at least one predicted CSI data.

Alternatively, the CSI data of the plurality of time slots may include CSI data of the first time slot and at least one type of CSI data;

Alternatively, the CSI data of the plurality of time slots may include CSI data of a first time slot and at least one second type of CSI data;

Alternatively, the CSI data for the multiple time slots may include predicted CSI data and at least one second type of CSI data;

Alternatively, the CSI data for the multiple time slots may include predicted CSI data and at least one type of first-type CSI data;

Alternatively, the CSI data for the multiple time slots may include CSI data acquired within the target time window.

In this embodiment of the application, CSI data is grouped based on the first information. The composition of the grouped data can be different depending on the data composition or grouping method. The data in a CSI data group can be CSI data corresponding to continuous CSI measurement resources, CSI data corresponding to continuous CSI reporting resources, or at least part of predicted CSI data, or a combination of CSI data in the current time slot and predicted CSI data. CSI grouping can also be determined based on time windows, as exemplarily shown in Figure 11.

In this embodiment of the application, the CSI data of the multiple time slots can be multiple predicted CSI data, which are compressed together and reported to the network-side device.

In this embodiment, the CSI data for the multiple time slots can be the CSI data of the current time slot and one or more predicted CSI data, which are compressed together and reported to the network-side device. The CSI data of the current time slot can also be used to monitor the performance of the target AI unit.

For CSI data across multiple time slots, similar grouped CSI data can be CSI data corresponding to continuous CSI measurement resources, CSI data corresponding to continuous CSI reporting resources, or at least partially predicted CSI data. It can also be a combination of current time slot CSI data and predicted CSI data, or CSI grouping can be determined based on time windows. Furthermore, it can be a combination of first time slot data and different types of CSI data, or a combination of predicted CSI data and different types of CSI data.

In this embodiment of the application, the predicted CSI data can also be described as a third type of CSI data.

Understandably, for AI units, the grouping method used in the testing or inference phase may be the same as that used in the training phase, or the data composition method used in the testing or inference phase may be the same as that used in the training phase, in order to obtain better processing results.

For example, when the grouped CSI data is channel or codebook data for M slots, or when the CSI data for multiple time slots is channel or codebook data for M slots, in one optional embodiment, the channel or codebook data for the M slots is channel or codebook data for M slots measured continuously; in another optional embodiment, the channel or codebook data for the M slots is predicted channel or codebook data for the M slots; in yet another optional embodiment, the channel or codebook data for the M slots includes the channel or codebook data for the current slot and the channel or codebook data for at least one predicted slot.

Optionally, the target time window is determined based on at least one of the following:

Time window length;

The number of cycles of the Channel State Information Reference Signal (CSI-RS);

The period of the Channel State Information Reference Signal (CSI-RS);

The offset value relative to the first time.

In this embodiment, the CSI or CSI of one group or multiple time slots can be channel or codebook data of M slots acquired within a target time window. The target time window can be determined based on the time window length parameter, the number of CSI-RS cycles, or the offset value of a first time, which can be a time determined based on a preset rule.

When the time window length of the group is 1, the CSI within that group undergoes spatial frequency domain compression.

Optionally, the first time is determined based on at least one of the following:

Preset reference time;

The time slot for the fourth CSI data that needs to be obtained now, wherein the fourth CSI data is the output data of the target AI unit;

Preset signal activation time;

Preset signal activation response time;

The position of the first CSI data within a grouped CSI data set;

The position of the last CSI data within a grouped CSI data group;

First time slot.

In this embodiment, the aforementioned first time can be a preset reference time, a time determined based on the location corresponding to specific CSI data, the time corresponding to the first time slot in the aforementioned embodiments, or a time determined based on the activation time and/or activation response time of a preset signal (e.g., downlink control signaling DCI). The aforementioned fourth CSI data can be understood as the target CSI data to be obtained, such as the third CSI data in the aforementioned embodiments.

Optionally, the target time window includes any of the following:

C is _{the first} cycle prior to the first time point;

The second time is _C2 cycles after the second time, where the second time is the first time or the time offset from the first time by the first offset amount;

A time window of a preset duration prior to the first time;

Wherein, the period is the CSI-RS period, or the period reported by CSI, and _C1 and _C2 are positive integers.

For example, the first time is at least one of the following:

Preset reference time;

The current slot n that needs to obtain CSI information;

The activation time or activation response time C relative to the preset signal;

The position number of the first CSI within the group;

The position number of the last CSI in the group;

The first slot relative to the predicted channel or codebook data.

Optionally, the target time window is a window counting C ₁ periods backward relative to the first time, such as a window of [slot nC ₁ * T, slot n].

Optionally, the time window is a window that is one time earlier than the first time and has a preset window length.

Optionally, the time window is a window that counts C ₂ periods after the first time, such as a window of [slot c + delta, slot c + delta + C ₂ * T].

Optionally, T above is the CSI-RS period, or the CSI reporting period, and the first offset Delta includes the signal processing time and the relative offset time.

Optionally, the first offset is determined based on the delay in acquiring the CSI data.

In this embodiment of the application, the delta is related to the first feedback delay or processing delay, which includes the delay of collecting the channel or codebook data of the M slots, such as the period of M1*CSI-RS, where M1<＝M.

Optionally, the target time window is a sliding time window.

In the embodiments of this application, the determination of CSI for groups or multiple time slots can be performed using a sliding window method.

In this embodiment of the application, during CSI data processing, CSI data from multiple time slots are processed (first operation). Specific operations include, but are not limited to, at least one of grouping, domain transformation, and data combination. By performing the above operations on multiple CSI data, i.e., using multi-time slot CSI data for CSI processing, historical information can be better utilized, or when multiple data need to be reported, information from other data can be used to assist the current slot in CSI processing (e.g., performing CSI compression and/or decompression), thereby optimizing the performance of CSI processing.

In this embodiment of the application, at least one of the grouping, domain transformation and combination related information of CSI data is obtained to optimize the CSI data to be transmitted, thereby improving the transmission performance of CSI.

The CSI data processing method provided in this application can be executed by a CSI data processing device. This application uses the execution of the CSI data processing method by a CSI data processing device as an example to illustrate the apparatus of the CSI data processing device provided in this application.

This application provides a CSI data processing apparatus. As an example, the CSI data processing apparatus 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.

The CSI data 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.

Specifically, referring to Figure 12, the CSI data processing device 1200 includes an acquisition module 1201 for acquiring first information, which includes at least one of the following information associated with the target artificial intelligence (AI) unit:

Information used to group CSI data;

Information used for domain transformation of CSI data;

Information used to combine CSI data;

The target AI unit includes at least one of the following:

The terminal is used to acquire the first AI unit of the target CSI;

Network-side devices are used to acquire the second AI unit for reconstructing CSI;

A reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI;

The third AI unit used in testing by the terminal, network-side equipment, or testing equipment;

The terminal, network-side device, or test equipment is used to match the fifth AI unit of the fourth AI unit used in the test;

The CSI data refers to CSI data from one or more time slots.

Optionally, the CSI data for the one or more time slots includes at least one of the following:

First CSI data;

CSI data from multiple time slots to be grouped;

CSI data from one or more time slots to be domain transformed;

Type 1 CSI data;

Second type of CSI data;

Wherein, the first CSI data is CSI data acquired in the first time slot, and the first time slot is the time slot of the most recent measured CSI-RS, the time slot used to generate CSI reporting correlation, or the time slot of predicted CSI correlation;

The first type of CSI data includes at least one of first CSI data and second CSI data, wherein the second CSI data is one or more CSI data acquired prior to the first CSI data;

The second type of CSI data is related data of the first type of CSI data, wherein the related data of the first type of CSI data is data obtained by processing the first CSI data and/or the second CSI data through a first preset process.

Optionally, the first information may further include at least one of the following: at least one of the model structure or parameter information of the sub-unit used to group the CSI data in the target AI unit;

The target AI unit has at least one of the model structure or model parameter information of the subunit used to perform domain transformation on the CSI data;

The target AI unit includes at least one of the model structure or model parameter information of the subunit used to combine the CSI data.

Optionally, the first information further includes first indication information, which indicates at least one of the following:

Whether to apply the sub-unit in the target AI unit used for grouping CSI data;

Whether to apply the sub-unit in the target AI unit used for domain transformation of CSI data;

Whether to apply the sub-unit in the target AI unit used for combining CSI data.

Optionally, the first information may further include at least one of the following:

The grouping information of the CSI data;

The time slot information associated with the CSI data;

The type information of the CSI data.

Optionally, the device further includes:

A determining module is configured to determine target data based on first information and a target AI unit; wherein the target data includes at least one of the following:

Input data for the target artificial intelligence (AI) unit;

Output data of the target AI unit;

Target CSI data;

Relevant data for the target CSI data.

Optionally, the first information is also used to indicate at least one of the following:

Reshape at least a portion of the grouped CSI data groups into data with Z ₁ channels, where Z ₁ is a positive integer;

The domain-transformed CSI data set is reshaped into _Z2 channels of data, where _Z2 is a positive integer.

Optionally, the information for grouping CSI data is used to indicate at least one of the following:

The CSI data from N time slots are grouped to obtain K CSI data groups; where N and K are positive integers greater than 1, and adjacent CSI data groups include CSI data from at least one of the same time slots.

The CSI data from the first time slot is combined with the M-1 CSI data preceding the first time slot to form a CSI data group.

Optionally, each CSI data group includes CSI data from M time slots, where 1 < M < N and K ≤ N - M + 1.

Optionally, the information used for domain transformation of CSI data includes at least one of the following:

Perform domain transformation on a per-CSI data set basis;

Based on the second indication information or the number of time slots agreed upon in the protocol, perform domain transformation on the CSI data of one or more time slots.

Optionally, if the target data includes at least one of the input data of the target artificial intelligence (AI) unit and the related data of the target CSI data, the method further includes: the target node inputs the target data into the target AI unit to obtain third CSI data, wherein the third CSI data includes at least one of the reported CSI data and the reconstructed CSI data;

Alternatively, if the target data includes at least one of the output data of the target AI unit and the target CSI data, the target data includes at least one of the reported CSI data and the reconstructed CSI data.

Optionally, when the first information includes information for grouping CSI data, the reconstructed CSI data is the reconstructed CSI data corresponding to the CSI data of the last time slot in each CSI data group;

Alternatively, if the information for grouping CSI data includes forming a CSI data group by combining the CSI data of the first time slot with the M-1 CSI data preceding the first time slot, then the reconstructed CSI data is the reconstructed CSI data corresponding to the first time slot.

Alternatively, if the information used to group the CSI data includes forming a CSI data group by combining the CSI data of the first time slot with the M-1 CSI data preceding the first time slot, the size of the reconstructed CSI data is 1/M times the size of the reconstructed CSI data of the M time slots.

Alternatively, if the first information includes information for performing domain transformation on the CSI data, the reconstructed CSI data is the reconstructed CSI data corresponding to the CSI data of the first time slot;

Alternatively, if the information for performing domain transformation on the CSI data includes comparing the CSI data of the first time slot with the CSI data of the M-1 preceding times slot, the reconstructed CSI data is the reconstructed CSI data corresponding to the first time slot.

Alternatively, if the information for performing domain transformation on the CSI data includes comparing the CSI data of the first time slot with the CSI data of M-1 times preceding the first time slot, the size of the reconstructed CSI data is 1/M times the size of the reconstructed CSI data of the M time slots.

Optionally, the CSI data in the first CSI data group are CSI data corresponding to consecutive CSI measurement resources, wherein the first CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data in the second CSI data group are CSI data corresponding to consecutive CSI reporting resources, wherein the second CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data in the third CSI data group is at least partly predicted CSI data, wherein the third CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data in the fourth CSI data group includes the CSI data of the current time slot and at least one predicted CSI data, wherein the fourth CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data in the fifth CSI data group includes CSI data acquired within the target time window, wherein the fifth CSI data group is at least one of the grouped CSI data groups;

Alternatively, the CSI data of the multiple time slots may include CSI data corresponding to continuous CSI measurement resources;

Alternatively, the CSI data of the multiple time slots may include CSI data corresponding to continuously reported CSI resources;

Alternatively, the CSI data in the plurality of time slots may be at least partially predicted CSI data;

Alternatively, the CSI data in the plurality of time slots may include at least one predicted CSI data.

Alternatively, the CSI data for the multiple time slots may include the CSI data for the current time slot and at least one predicted CSI data.

Alternatively, the CSI data of the plurality of time slots may include CSI data of the first time slot and at least one type of CSI data;

Alternatively, the CSI data of the plurality of time slots may include CSI data of a first time slot and at least one second type of CSI data;

Alternatively, the CSI data for the multiple time slots may include predicted CSI data and at least one second type of CSI data;

Alternatively, the CSI data of the multiple time slots may include predicted CSI data and at least one type of first-type CSI data;

Alternatively, the CSI data for the multiple time slots may include CSI data acquired within the target time window.

Optionally, the target time window is determined based on at least one of the following:

Time window length;

The number of cycles of the Channel State Information Reference Signal (CSI-RS);

The period of the Channel State Information Reference Signal (CSI-RS);

The offset value relative to the first time.

Optionally, the first time is determined based on at least one of the following:

Preset reference time;

The time slot for the fourth CSI data that needs to be obtained now, wherein the fourth CSI data is the output data of the target AI unit;

Preset signal activation time;

Preset signal activation response time;

The position of the first CSI data within a grouped CSI data set;

The position of the last CSI data within a grouped CSI data group;

First time slot.

Optionally, the target time window includes any of the following:

C is _{the first} cycle prior to the first time point;

The second time is _C2 cycles after the second time, where the second time is the first time or the time offset from the first time by the first offset amount;

A time window of a preset duration prior to the first time;

Wherein, the period is the CSI-RS period, or the period reported by CSI, and _C1 and _C2 are positive integers.

Optionally, the first offset is determined based on the delay in acquiring the CSI data.

Optionally, the target time window is a sliding time window.

Optionally, the second type of CSI data includes at least one of the following:

The cached data after the first CSI data and/or the second CSI data have undergone the first preset processing;

The first CSI data and/or the second CSI data are cached data after passing through a preset unit of the target AI unit;

The data output by the sub-unit of the target AI unit obtained by combining the first CSI data and/or the second CSI data;

The final layer output data of the target AI unit obtained by combining the first CSI data and/or the second CSI data;

The cached data after the first CSI data and/or the related data of the second CSI data have undergone the first preset processing;

The relevant data of the first CSI data and/or the second CSI data are cached data after passing through the preset unit of the target AI unit;

The data output by the subunit of the target AI unit is obtained by combining the relevant data of the first CSI data and/or the second CSI data;

The final layer output data of the target AI unit is obtained by combining the first CSI data and/or the relevant data of the second CSI data;

The related data of the second CSI data is data obtained by processing at least one CSI data previously acquired before the second CSI data through a second preset process.

Optionally, the second type of CSI data includes data output by a sub-unit of the target AI unit obtained by combining the first CSI data and/or the second CSI data, and/or, the second type of CSI data includes data output by a sub-unit of the target AI unit obtained by combining relevant data from the first CSI data and/or the second CSI data, and the second type of CSI data includes at least one of the following:

The data output by the transform module, a sub-module of the target AI unit;

The data output by the forward feedback module of the sub-module of the target AI unit;

The data output by the attention layer module, a sub-module of the target AI unit;

The data output by the convolutional feature extraction layer module of the target AI unit;

The output data of multiple cascaded modules of the target AI unit;

The sum of the output data from multiple modules of the target AI unit;

The target AI unit's multiple modules output data mapping data.

Optionally, the second type of CSI data is data obtained by combining the first CSI data and/or the second CSI data and performing a third preset processing on the data output by a subunit of the target AI unit, and/or, the second type of CSI data includes data obtained by combining the first CSI data and/or the second CSI data and performing a third preset processing on the data output by a subunit of the target AI unit, wherein the third preset processing includes at least one of the following:

Cascaded processing; mapping processing; dimensional scaling processing; quantization processing.

Optionally, the information used to combine the CSI data indicates at least one of the following:

The CSI data from the multiple time slots are weighted and summed.

The CSI data from the multiple time slots are input into the target AI unit;

The first CSI data and the second type of CSI data are input into different sub-units of the target AI unit;

The second type of CSI data is input into the target AI unit after the sub-unit of the target AI unit into which the first CSI data is input;

The second type of CSI data is input into the sub-unit of the target AI unit.

Optionally, inputting the first CSI data and the second type of CSI data into different sub-units of the target AI unit includes:

The first CSI data is input into the first sub-unit of the target AI unit.

The second type of CSI data is input into the second sub-unit of the target AI unit.

The second subunit is the subunit following the first subunit in the target AI unit.

Optionally, the weighted summation of the CSI data from the multiple time slots includes:

The first type of CSI data and/or the second type of CSI data are weighted and summed before the third sub-unit of the target AI unit;

The third sub-unit includes at least one of the following: a transform module, a forward feedback module, an attention layer module, and a convolutional feature extraction layer module.

Optionally, the sub-unit includes at least one of the following: a transform module, an attention layer module, a feedforward module, and a convolutional feature extraction layer module.

Optionally, the information used to combine the CSI data includes:

Adjust the CSI data to be combined to be data of the same dimension.

Optionally, the first CSI data includes at least one of the following: CSI data at the predicted time point, and CSI data at the measured time point;

The second CSI data includes at least one of the following: CSI data at the predicted time point, and CSI data at the measured time point.

Optionally, the model structure or model parameter information of the sub-unit in the target AI unit used for combining CSI data includes at least one of the following:

The sub-unit is located in the target AI unit;

The type of the subunit;

The parameters of the subunit;

The connection method of the subunit;

The manner in which the CSI data is input into the sub-unit.

It should be noted that the CSI data processing apparatus provided in this application embodiment is an apparatus capable of executing the above-described CSI data processing method. Therefore, all implementation methods in the above-described CSI data processing method embodiments are applicable to this electronic device and can achieve the same or similar beneficial effects. To avoid repetition, this embodiment will not elaborate further.

The apparatus provided in this application embodiment can implement the various processes implemented in the method embodiments of Figures 2 to 11 and achieve the same technical effect. To avoid repetition, it will not be described again here.

As shown in Figure 13, this application embodiment also provides a communication device 1300, including a processor 1301 and a memory 1302. The memory 1302 stores a program or instructions that can run on the processor 1301. For example, when the communication device 1300 is a terminal, the program or instructions executed by the processor 1301 implement the various steps of the above-described CSI data processing method embodiment and achieve the same technical effect. When the communication device 1300 is a network-side device, the program or instructions executed by the processor 1301 implement the various steps of the above-described CSI data processing method embodiment and achieve the same technical effect. To avoid repetition, this will not be described again here.

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 embodiments shown in Figures 2-11. This terminal embodiment corresponds to the above-described terminal-side method embodiments, and all implementation processes and methods of the above-described method embodiments can be applied to this terminal embodiment and achieve the same technical effect. The terminal may be the CSI data processing device shown in Figure 12. Specifically, Figure 14 is a schematic diagram of the hardware structure of a terminal implementing an embodiment of this application.

The terminal 1400 includes, but is not limited to, at least some of the following components: radio frequency unit 1401, network module 1402, audio output unit 1403, input unit 1404, sensor 1405, display unit 1406, user input unit 1407, interface unit 1408, memory 1409, and processor 1410.

Those skilled in the art will understand that the terminal 1400 may also include a power supply (such as a battery) for powering various components. The power supply can be logically connected to the processor 1410 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 14 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.

It should be understood that, in this embodiment, the input unit 1404 may include a graphics processor 14041 and a microphone 14042. The graphics processor 14041 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 1406 may include a display panel 14061, which may be configured in the form of a liquid crystal display, an organic light-emitting diode, or the like. The user input unit 1407 includes at least one of a touch panel 14071 and other input devices 14072. The touch panel 14071 is also called a touch screen. The touch panel 14071 may include a touch detection device and a touch controller. Other input devices 14072 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.

In this embodiment, after receiving downlink data from the network-side device, the radio frequency unit 1401 can transmit it to the processor 1410 for processing; in addition, the radio frequency unit 1401 can send uplink data to the network-side device. Typically, the radio frequency unit 1401 includes, but is not limited to, antennas, amplifiers, transceivers, couplers, low-noise amplifiers, duplexers, etc.

The memory 1409 can be used to store software programs or instructions, as well as various data. The memory 1409 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 1409 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 1409 in this embodiment includes, but is not limited to, these and any other suitable types of memory.

Processor 1410 may include one or more processing units; optionally, processor 1410 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 1410.

The radio frequency unit 1401 or the processor 1410 is configured to acquire first information, the first information including at least one of the following information associated with the target artificial intelligence (AI) unit:

Information used to group CSI data;

Information used for domain transformation of CSI data;

Information used to combine CSI data;

The target AI unit includes at least one of the following:

The terminal is used to acquire the first AI unit of the target CSI;

Network-side devices are used to acquire the second AI unit for reconstructing CSI;

A reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI;

The third AI unit used in testing by the terminal, network-side equipment, or testing equipment;

The terminal, network-side device, or test equipment is used to match the fifth AI unit of the fourth AI unit used in the test;

The CSI data refers to CSI data from one or more time slots.

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

This application also provides a network-side device, including a processor and a communication interface. The communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the steps of the method embodiment shown in Figures 2-11. This network-side device embodiment corresponds to the above-described network-side device method embodiment. All implementation processes and methods of the above-described method embodiments can be applied to this network-side device embodiment and achieve the same technical effects.

Specifically, this application embodiment also provides a network-side device, which may be the CSI data processing device shown in FIG12. As shown in FIG15, the network-side device 1500 includes: an antenna 151, a radio frequency device 152, a baseband device 153, a processor 154, and a memory 155. The antenna 151 is connected to the radio frequency device 152. In the uplink direction, the radio frequency device 152 receives information through the antenna 151 and sends the received information to the baseband device 153 for processing. In the downlink direction, the baseband device 153 processes the information to be transmitted and sends it to the radio frequency device 152. The radio frequency device 152 processes the received information and transmits it through the antenna 151.

The method executed by the network-side device in the above embodiments can be implemented in the baseband device 153, which includes a baseband processor.

The baseband device 153 may include at least one baseband board, on which multiple chips are disposed, as shown in FIG15. One of the chips is, for example, a baseband processor, which is connected to the memory 155 via a bus interface to call the program in the memory 155 and execute the network device operation shown in the above method embodiment.

The network-side device may also include a network interface 156, such as a Common Public Radio Interface (CPRI).

Specifically, the network-side device 1500 in this application embodiment further includes: instructions or programs stored in memory 155 and executable on processor 154. Processor 154 calls the instructions or programs in memory 155 to execute the methods executed by each module shown in FIG12 and achieve the same technical effect. To avoid repetition, it will not be described in detail here.

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 above-described CSI data processing method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

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.

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 CSI data processing method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

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.

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 above-described CSI data processing method embodiments, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

This application also provides a communication system, including: a terminal and a network-side device, wherein the terminal can be used to execute the steps of the CSI data processing method described above, and/or the network-side device can be used to execute the steps of the CSI data processing method described above.

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.

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.

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 (34)

A method for processing Channel State Information (CSI) data includes: The target node acquires first information, which includes at least one of the following information associated with the target artificial intelligence (AI) unit: Information used to group CSI data; Information used for domain transformation of CSI data; Information used to combine CSI data; The target AI unit includes at least one of the following: The terminal is used to acquire the first AI unit of the target CSI; Network-side devices are used to acquire the second AI unit for reconstructing CSI; A reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI; The third AI unit used in testing by the terminal, network-side equipment, or testing equipment; The terminal, network-side device, or test equipment is used to match the fifth AI unit of the fourth AI unit used in the test; The CSI data refers to CSI data from one or more time slots. According to the method of claim 1, the CSI data of the one or more time slots includes at least one of the following: First CSI data; CSI data from multiple time slots to be grouped; CSI data from one or more time slots to be domain transformed; Type 1 CSI data; Second type of CSI data; Wherein, the first CSI data is CSI data acquired in the first time slot, and the first time slot is the time slot of the most recent measured CSI-RS, the time slot used to generate CSI reporting correlation, or the time slot of predicted CSI correlation; The first type of CSI data includes at least one of first CSI data and second CSI data, wherein the second CSI data is one or more CSI data acquired prior to the first CSI data; The second type of CSI data is related data of the first type of CSI data, wherein the related data of the first type of CSI data is data obtained by processing the first CSI data and/or the second CSI data through a first preset process. According to the method of claim 1 or 2, the first information further includes at least one of the following: at least one of the model structure of the sub-unit used to group the CSI data in the target AI unit or the parameter information of the model sub-unit; The target AI unit has at least one of the model structure or model parameter information of the subunit used to perform domain transformation on the CSI data; The target AI unit includes at least one of the model structure or model parameter information of the subunit used to combine the CSI data. According to the method of claim 3, the first information further includes first indication information, the first indication information indicating at least one of the following: Whether to apply the sub-unit in the target AI unit used for grouping CSI data; Whether to apply the sub-unit in the target AI unit used for domain transformation of CSI data; Whether to apply the sub-unit in the target AI unit used for combining CSI data. The method according to any one of claims 1-4, wherein the first information further includes at least one of the following: The grouping information of the CSI data; The time slot information associated with the CSI data; The type information of the CSI data. The method according to any one of claims 1-5, wherein the method further comprises: The target node determines target data based on the first information and the target AI unit; wherein the target data includes at least one of the following: Input data for the target artificial intelligence (AI) unit; Output data of the target AI unit; Target CSI data; Relevant data for the target CSI data. The method according to any one of claims 1-6, wherein the first information is further used to indicate at least one of the following: Reshape at least a portion of the grouped CSI data groups into data with Z ₁ channels, where Z ₁ is a positive integer; The domain-transformed CSI data set is reshaped into _Z2 channels of data, where _Z2 is a positive integer. The method according to any one of claims 1-3, wherein the information for grouping CSI data is used to indicate at least one of the following: The CSI data from N time slots are grouped to obtain K CSI data groups; where N and K are positive integers greater than 1, and adjacent CSI data groups include CSI data from at least one of the same time slots. The CSI data from the first time slot is combined with the M-1 CSI data preceding the first time slot to form a CSI data group. According to the method of claim 8, each CSI data group comprises CSI data in M time slots, 1 < M < N, K ≤ N-M+1. The method according to any one of claims 1-5, wherein the information for performing domain transformation on the CSI data includes at least one of the following: Perform domain transformation on a per-CSI data set basis; Based on the second indication information or the number of time slots agreed upon in the protocol, perform domain transformation on the CSI data of one or more time slots. The method according to claim 6, wherein, If the target data includes at least one of the input data of the target artificial intelligence (AI) unit and the related data of the target CSI data, the method further includes: the target node inputs the target data into the target AI unit to obtain third CSI data, wherein the third CSI data includes at least one of the reported CSI data and the reconstructed CSI data; Alternatively, if the target data includes at least one of the output data of the target AI unit and the target CSI data, the target data includes at least one of the reported CSI data and the reconstructed CSI data. According to the method of claim 11, wherein, when the first information includes information for grouping CSI data, the reconstructed CSI data is the reconstructed CSI data corresponding to the CSI data of the last time slot in each CSI data group; Alternatively, if the information for grouping CSI data includes forming a CSI data group by combining the CSI data of the first time slot with the M-1 CSI data preceding the first time slot, then the reconstructed CSI data is the reconstructed CSI data corresponding to the first time slot. Alternatively, if the information used to group the CSI data includes forming a CSI data group by combining the CSI data of the first time slot with the M-1 CSI data preceding the first time slot, the size of the reconstructed CSI data is 1/M times the size of the reconstructed CSI data of the M time slots. Alternatively, if the first information includes information for performing domain transformation on the CSI data, the reconstructed CSI data is the reconstructed CSI data corresponding to the CSI data of the first time slot; Alternatively, if the information for performing domain transformation on the CSI data includes comparing the CSI data of the first time slot with the CSI data of the M-1 preceding times slot, the reconstructed CSI data is the reconstructed CSI data corresponding to the first time slot. Alternatively, if the information for performing domain transformation on the CSI data includes comparing the CSI data of the first time slot with the CSI data of M-1 times preceding the first time slot, the size of the reconstructed CSI data is 1/M times the size of the reconstructed CSI data of the M time slots. The method according to claim 2, wherein, The CSI data in the first CSI data group are CSI data corresponding to continuous CSI measurement resources, wherein the first CSI data group is at least one of the grouped CSI data groups; Alternatively, the CSI data in the second CSI data group are CSI data corresponding to consecutive CSI reporting resources, wherein the second CSI data group is at least one of the grouped CSI data groups; Alternatively, the CSI data in the third CSI data group is at least partly predicted CSI data, wherein the third CSI data group is at least one of the grouped CSI data groups; Alternatively, the CSI data in the fourth CSI data group includes the CSI data of the current time slot and at least one predicted CSI data, wherein the fourth CSI data group is at least one of the grouped CSI data groups; Alternatively, the CSI data in the fifth CSI data group includes CSI data acquired within the target time window, wherein the fifth CSI data group is at least one of the grouped CSI data groups; Alternatively, the CSI data of the multiple time slots may include CSI data corresponding to continuous CSI measurement resources; Alternatively, the CSI data of the multiple time slots may include CSI data corresponding to continuously reported CSI resources; Alternatively, the CSI data in the plurality of time slots may be at least partially predicted CSI data; Alternatively, the CSI data in the plurality of time slots may include at least one predicted CSI data. Alternatively, the CSI data for the multiple time slots may include the CSI data for the current time slot and at least one predicted CSI data. Alternatively, the CSI data of the plurality of time slots may include CSI data of the first time slot and at least one type of CSI data; Alternatively, the CSI data of the plurality of time slots may include CSI data of a first time slot and at least one second type of CSI data; Alternatively, the CSI data for the multiple time slots may include predicted CSI data and at least one second type of CSI data; Alternatively, the CSI data of the multiple time slots may include predicted CSI data and at least one type of first-type CSI data; Alternatively, the CSI data for the multiple time slots may include CSI data acquired within the target time window. The method of claim 13, wherein the target time window is determined according to at least one of the following: Time window length; The number of cycles of the Channel State Information Reference Signal (CSI-RS); The period of the Channel State Information Reference Signal (CSI-RS); The offset value relative to the first time. The method of claim 14, wherein the first time is determined according to at least one of the following: Preset reference time; The time slot for the fourth CSI data that needs to be obtained now, wherein the fourth CSI data is the output data of the target AI unit; Preset signal activation time; Preset signal activation response time; The position of the first CSI data within a grouped CSI data set; The position of the last CSI data within a grouped CSI data group; First time slot. The method according to claim 14 or 15, wherein the target time window comprises any one of the following: C is _{the first} cycle prior to the first time point; The second time is _C2 cycles after the second time, where the second time is the first time or the time offset from the first time by the first offset amount; A time window of a preset duration prior to the first time; Wherein, the period is the CSI-RS period, or the period reported by CSI, and _C1 and _C2 are positive integers. According to the method of claim 16, the first offset is determined based on the delay in acquiring CSI data. The method according to any one of claims 13-17, wherein the target time window is a sliding time window. The method according to claim 2, wherein the second type of CSI data includes at least one of the following: The cached data after the first CSI data and/or the second CSI data have undergone the first preset processing; The first CSI data and/or the second CSI data are cached data after passing through a preset unit of the target AI unit; The data output by the sub-unit of the target AI unit obtained by combining the first CSI data and/or the second CSI data; The final layer output data of the target AI unit obtained by combining the first CSI data and/or the second CSI data; The cached data after the first CSI data and/or the related data of the second CSI data have undergone the first preset processing; The relevant data of the first CSI data and/or the second CSI data are cached data after passing through the preset unit of the target AI unit; The data output by the subunit of the target AI unit is obtained by combining the relevant data of the first CSI data and/or the second CSI data; The final layer output data of the target AI unit is obtained by combining the first CSI data and/or the relevant data of the second CSI data; The related data of the second CSI data is data obtained by processing at least one CSI data previously acquired before the second CSI data through a second preset process. According to the method of claim 19, wherein the second type of CSI data includes data output by a subunit of the target AI unit obtained by combining the first CSI data and/or the second CSI data, and/or, the second type of CSI data includes data output by a subunit of the target AI unit obtained by combining related data of the first CSI data and/or the second CSI data, and the second type of CSI data includes at least one of the following: The data output by the transform module, a sub-module of the target AI unit; The data output by the forward feedback module of the sub-module of the target AI unit; The data output by the attention layer module, a sub-module of the target AI unit; The data output by the convolutional feature extraction layer module of the target AI unit; The output data of multiple cascaded modules of the target AI unit; The sum of the output data from multiple modules of the target AI unit; The target AI unit's multiple modules output data mapping data. According to the method of claim 19, wherein the second type of CSI data is data after performing a third preset processing on the data output by a subunit of the target AI unit obtained by combining the first CSI data and/or the second CSI data, and/or, the second type of CSI data includes data after performing a third preset processing on the data output by a subunit of the target AI unit obtained by combining the first CSI data and/or the second CSI data, wherein the third preset processing includes at least one of the following: Cascaded processing; mapping processing; dimensional scaling processing; quantization processing. The method of claim 2, wherein the information for combining CSI data is used to indicate at least one of the following: The CSI data from the multiple time slots are weighted and summed. The CSI data from the multiple time slots are input into the target AI unit; The first CSI data and the second type of CSI data are input into different sub-units of the target AI unit; The second type of CSI data is input into the target AI unit after the sub-unit of the target AI unit into which the first CSI data is input; The second type of CSI data is input into the sub-unit of the target AI unit. According to the method of claim 22, wherein inputting the first CSI data and the second type of CSI data into different sub-units of the target AI unit includes: The first CSI data is input into the first sub-unit of the target AI unit. The second type of CSI data is input into the second sub-unit of the target AI unit. The second subunit is the subunit following the first subunit in the target AI unit. According to the method of claim 22, the weighted summation of the CSI data from the plurality of time slots includes: The first type of CSI data and/or the second type of CSI data are weighted and summed before the third sub-unit of the target AI unit; The third sub-unit includes at least one of the following: a transform module, a forward feedback module, an attention layer module, and a convolutional feature extraction layer module. The method according to any one of claims 19-24, wherein the sub-unit comprises at least one of the following: a transform module, an attention layer module, a feedforward module, and a convolutional feature extraction layer module. The method according to any one of claims 1-25, wherein the information for combining the CSI data includes: Adjust the CSI data to be combined to be data of the same dimension. According to the method of claim 2, the first CSI data includes at least one of the following: predicted CSI data at the time point, and measured CSI data at the time point; The second CSI data includes at least one of the following: CSI data at the predicted time point, and CSI data at the measured time point. According to the method of claim 3, the model structure or model parameter information of the sub-unit in the target AI unit used for combining CSI data includes at least one of the following: The sub-unit is located in the target AI unit; The type of the subunit; The parameters of the subunit; The connection method of the subunit; The manner in which the CSI data is input into the sub-unit. A CSI data processing apparatus, comprising: The acquisition module is configured to acquire first information, which includes at least one of the following information associated with the target artificial intelligence (AI) unit: Information used to group CSI data; Information used for domain transformation of CSI data; Information used to combine CSI data; The target AI unit includes at least one of the following: The terminal is used to acquire the first AI unit of the target CSI; Network-side devices are used to acquire the second AI unit for reconstructing CSI; A reference AI unit used by a terminal or network-side device to acquire the target CSI or reconstruct the CSI; The third AI unit used in testing by the terminal, network-side equipment, or testing equipment; The terminal, network-side device, or test equipment is used to match the fifth AI unit of the fourth AI unit used in the test; The CSI data refers to CSI data from one or more time slots. The apparatus according to claim 29, wherein the apparatus further comprises: A determining module is configured to determine target data based on first information and a target AI unit; wherein the target data includes at least one of the following: Input data for the target artificial intelligence (AI) unit; Output data of the target AI unit; Target CSI data; Relevant data for the target CSI data. 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 CSI data processing method as claimed in any one of claims 1 to 28. A network-side 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 CSI data processing method as described in any one of claims 1 to 28. A readable storage medium storing a program or instructions that, when executed by a processor, implement the CSI data processing method as described in any one of claims 1-28. A computer program product includes computer instructions that, when executed by a processor, implement the steps of the CSI data processing method as described in any one of claims 1-28.