Communication method and device

Abstract

The invention provides a communication method and device. The training overhead of artificial intelligence in a wireless communication network can be reduced. The method comprises the following steps: acquiring first coding information; wherein the first encoding information corresponds to first channel information input into a first encoder in the first equipment, and a statistical dependency relationship between the first channel information and the first encoding information meets a first condition; determining a statistical dependency relationship between the first coding information and the second coding information; wherein the second encoding information corresponds to first channel information input into a second encoder in the second equipment, and the statistical dependency relationship between the first channel information and the second encoding information meets a first condition; and when the statistical dependency relationship between the first encoding information and the second encoding information meets a second condition, updating the second encoder according to the first encoding information.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a communication method and device. Background Art

[0002] In wireless communication networks, such as mobile communication networks, the services supported by the network are becoming more and more diverse, so the requirements that need to be met are becoming more and more diverse. For example, the network needs to be able to support ultra-high speed, ultra-low latency, and / or ultra-large connections. This feature makes network planning, network configuration, and / or resource scheduling more and more complicated. In addition, as the network becomes more and more powerful, such as supporting higher and higher spectrum, supporting high-order multiple input multiple output (MIMO) technology, supporting beamforming, and / or supporting new technologies such as beam management, network energy saving has become a hot research topic. These new requirements, new scenarios, and new features have brought unprecedented challenges to network planning, operation and maintenance, and efficient operation. In order to meet this challenge, artificial intelligence technology can be introduced into wireless communication networks to realize network intelligence. Based on this, how to effectively implement artificial intelligence in the network, for example, how to reduce the training overhead of artificial intelligence, is a problem worth studying. Summary of the invention

[0003] The present application provides a communication method and device, which can reduce the training overhead of artificial intelligence in a wireless communication network and improve communication performance.

[0004] In a first aspect, the present application provides a communication method, comprising: obtaining first coding information; wherein the first coding information corresponds to first channel information input to a first encoder applied to a first device, and the statistical dependency relationship between the first channel information and the first coding information meets a first condition; determining a statistical dependency relationship between the first coding information and second coding information; wherein the second coding information corresponds to the first channel information input to a second encoder applied to a second device, and the statistical dependency relationship between the first channel information and the second coding information meets the first condition; when the statistical dependency relationship between the first coding information and the second coding information meets the second condition, updating the second encoder according to the first coding information.

[0005] In the above design, the communicating parties first separately train autoencoders that meet the same constraints, and then update the encoder on one side with less communication overhead to achieve matching between the updated encoder and the decoder on the other side, which is beneficial to improving the performance of the autoencoders used in the communication network.

[0006] In one possible design, the difference between the first encoding information and the third encoding information is less than a set threshold, and the third encoding information corresponds to the first channel information input to the updated second encoder. Such a design can improve the matching degree between the updated second encoder and the first decoder.

[0007] In a possible design, the first encoder or the second encoder before updating is obtained through offline training. Such a design can reduce the communication overhead during the training process of the autoencoder.

[0008] In a possible design, before obtaining the first encoding information, it also includes: sending first capability information to the first device, where the first capability information is used to indicate support for updating the second encoder.

[0009] In one possible design, before updating the second encoder according to the first encoding information, it also includes: sending the statistical dependency of the first encoding information and the second encoding information to the first device; receiving first indication information from the first device, the first indication information being used to indicate that the statistical dependency of the first encoding information and the second encoding information meets the second condition, or the first indication information being used to trigger an update of the second encoder.

[0010] In one possible design, the statistical dependency of two information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or M parameters in the normalized mutual information of the two information, where M is a positive integer; wherein the two information include the first channel information and the first coding information, or the first channel information and the second coding information, and the first condition is a preconfigured condition with a mapping relationship with the M parameters. In such a design, the communicating parties use the same statistical method to calculate the statistical dependency of the input and output of the encoder, so that the statistical dependency of the input and output of the encoder in offline training can meet the same constraints.

[0011] In one possible design, the method also includes: receiving third indication information, wherein the third indication information is used to indicate a method for determining a statistical dependency relationship between the first encoding information and the second encoding information, and the second condition corresponding to the determination method.

[0012] In one possible design, the statistical dependency between the first coding information and the second coding information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or N parameters in the normalized mutual information of the first coding information and the second coding information, where N is a positive integer, and there is a mapping relationship between the second condition and the N parameters.

[0013] In one possible design, obtaining the first coded information includes: receiving the first coded information from the first device; or, receiving quantization information corresponding to the first coded information from the first device, and dequantizing the quantization information to obtain the first coded information.

[0014] In one possible design, the method is applied to the second device, and the method further includes: obtaining second channel information; encoding and quantizing the second channel information according to the updated second encoder and the first quantization method to obtain third encoded information; and sending the third encoded information. Optionally, the first quantization method is one of a plurality of preconfigured quantization methods, or the first quantization method is indicated by the first device to the second device. Through such a design, the quantization bit width (i.e., the length of the quantization information) can be adjusted to achieve different feedback overheads. During implementation, only the model under a certain bit width can be adapted online to complete model adaptation of different feedback overheads, further reducing the overhead of model adaptation.

[0015] In a second aspect, the present application provides a communication method, comprising: sending first coded information or quantization information corresponding to the first coded information; wherein the first coded information corresponds to first channel information input to a first encoder applied to a first device, and a statistical dependency relationship between the first channel information and the first coded information meets a first condition; the first coded information is used to update a second encoder applied to a second device; wherein a statistical dependency relationship between the input and output of the second encoder before the update meets the first condition.

[0016] In a possible design, the first encoder or the second encoder before updating is obtained through offline training.

[0017] In one possible design, the method is applied to a first device, and before sending first encoding information or quantization information corresponding to the first encoding information, the method also includes: receiving first capability information from the second device, the first capability information being used to indicate that the second device supports updating the second encoder.

[0018] In one possible design, the method is applied to a first device, and the method also includes: receiving a statistical dependency between the first coding information and the second coding information; wherein the second coding information corresponds to the first channel information input to the second encoder before updating, and the statistical dependency between the first channel information and the second coding information meets the first condition; sending first indication information, the first indication information is used to indicate that the statistical dependency between the first coding information and the second coding information meets the second condition, or the first indication information is used to trigger an update of the second encoder.

[0019] In one possible design, the statistical dependency relationship between two information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or M parameters in the normalized mutual information of the two information, where M is a positive integer; wherein the two information include the first channel information and the first coding information, or the first channel information and the second coding information, and the first condition is a preconfigured condition that has a mapping relationship with the M parameters.

[0020] In one possible design, the method also includes: sending third indication information, wherein the third indication information is used to indicate a method for determining a statistical dependency relationship between the first coding information and the second coding information, and the second condition corresponding to the determination method.

[0021] In one possible design, the statistical dependency between the first coding information and the second coding information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or N parameters in the normalized mutual information of the first coding information and the second coding information, where N is a positive integer, and there is a mapping relationship between the second condition and the N parameters.

[0022] In a third aspect, the present application provides a communication method, including: obtaining second coding information; wherein the second coding information corresponds to first channel information input to a second encoder applied to a second device, and the statistical dependency between the first channel information and the second coding information meets a first condition; determining the statistical dependency between the first coding information and the second coding information; wherein the first coding information corresponds to the first channel information input to a first encoder applied to a first device, and the statistical dependency between the first channel information and the first coding information meets the first condition; when the statistical dependency between the first coding information and the second coding information meets the second condition, updating the first decoder applied to the first device according to the second coding information, and the first decoder before the update matches the first encoder.

[0023] In the above design, the communicating parties first separately train the autoencoders that meet the same constraints, and then update the decoder in one party with less communication overhead to achieve matching between the updated decoder and the encoder in the other party, which is beneficial to improving the performance of the autoencoders used in the communication network.

[0024] In one possible design, the difference between the first channel information and the third channel information is less than a set threshold; wherein the third channel information is the output of the updated first decoder, and the third channel information corresponds to the second encoded information input to the updated first decoder.

[0025] In a possible design, the first encoder, the first decoder before updating, or the second encoder is obtained through offline training.

[0026] In a possible design, the statistical dependency relationship between two information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or M parameters in the normalized mutual information of the two information, where M is a positive integer; wherein the two information include the first channel information and the first coding information, or the first channel information and the second coding information; and the first condition is a preconfigured condition that has a mapping relationship with the M parameters.

[0027] In one possible design, the statistical dependency between the first coded information and the second coded information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or N parameters in the normalized mutual information of the first coded information and the second coded information, where N is a positive integer; the second condition is a preconfigured condition that has a mapping relationship with the N parameters.

[0028] In a possible design, before obtaining the second coding information, it also includes: sending first request information to the second device, where the first request information is used to request the second coding information.

[0029] In one possible design, obtaining the second coded information includes: receiving the second coded information from the second device; or, receiving quantization information corresponding to the second coded information from the second device, and dequantizing the quantization information to obtain the second coded information.

[0030] In one possible design, the method is applied to the first device, and the method also includes: receiving fourth coded information from a second device, wherein the fourth coded information is determined based on the second encoder and the first quantization method; and dequantizing and demodulating the fourth coded information according to the updated first decoder and the first quantization method to obtain channel information corresponding to the fourth coded information.

[0031] In a possible design, the first quantization method is one of multiple preconfigured quantization methods.

[0032] In one possible design, the method further includes: sending the first quantization mode to the second device.

[0033] In a fourth aspect, the present application provides a communication method, comprising: sending second coded information or quantization information corresponding to the second coded information; wherein the second coded information corresponds to first channel information input to a second encoder applied to a second device, and the statistical dependency between the first channel information and the second coded information meets a first condition; the second coded information is used to update a first decoder applied to a first device; wherein the statistical dependency between the input and output of the first encoder applied to the first device meets the first condition, and the first decoder before the update matches the first encoder.

[0034] In a possible design, the first encoder, the first decoder before updating, or the second encoder is obtained through offline training.

[0035] In one possible design, the statistical dependency between the first channel information and the second coded information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or M parameters in the normalized mutual information of the first channel information and the second coded information, where M is a positive integer; and the first condition is a preconfigured condition that has a mapping relationship with the M parameters.

[0036] In a possible design, before sending the second coding information or the quantization information corresponding to the second coding information, it also includes: receiving first request information, where the first request information is used to request the second coding information.

[0037] In one possible design, the method is applied to the second device, and the method also includes: obtaining fourth channel information; encoding and quantizing the fourth channel information according to the second encoder and the first quantization method to obtain fourth encoded information; and sending the fourth encoded information.

[0038] In a possible design, the first quantization method is one of a plurality of preconfigured quantization methods, or the first quantization method is indicated by the first device to the second device.

[0039] In a fifth aspect, the present application provides a communication device, which may be a second device, or a device, module, or chip applied to the second device, or a device that can be used in combination with the second device. In one design, the communication device may include a module that performs the method / operation / step / action described in the first aspect, and the module may be a hardware circuit, or software, or a combination of a hardware circuit and software. In one design, the communication device may include a processing module and a communication module.

[0040] A communication module, configured to obtain first encoding information; wherein the first encoding information corresponds to first channel information input to a first encoder applied to a first device, and a statistical dependency relationship between the first channel information and the first encoding information meets a first condition;

[0041] A processing module, used for determining a statistical dependency between the first coding information and the second coding information; wherein the second coding information corresponds to the first channel information input to the second encoder applied to the second device, and the statistical dependency between the first channel information and the second coding information meets the first condition; and when the statistical dependency between the first coding information and the second coding information meets the second condition, updating the second encoder according to the first coding information.

[0042] In one possible design, a difference between the first encoding information and the third encoding information is smaller than a set threshold, and the third encoding information corresponds to the first channel information input into the updated second encoder.

[0043] In a possible design, the first encoder or the second encoder before updating is obtained through offline training.

[0044] In one possible design, the communication module is further used to send first capability information to the first device before obtaining the first encoding information, where the first capability information is used to indicate support for updating the second encoder.

[0045] In one possible design, the communication module is also used to send the statistical dependency of the first coding information and the second coding information to the first device before the processing module updates the second encoder according to the first coding information; and receive first indication information from the first device, wherein the first indication information is used to indicate that the statistical dependency of the first coding information and the second coding information meets the second condition, or the first indication information is used to trigger an update of the second encoder.

[0046] In one possible design, the statistical dependency relationship between two information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or M parameters in the normalized mutual information of the two information, where M is a positive integer; wherein the two information include the first channel information and the first coding information, or the first channel information and the second coding information, and the first condition is a preconfigured condition that has a mapping relationship with the M parameters.

[0047] In one possible design, the communication module is also used to receive third indication information, and the third indication information is used to indicate a method for determining a statistical dependency relationship between the first coding information and the second coding information, and the second condition corresponding to the determination method.

[0048] In one possible design, the statistical dependency between the first coding information and the second coding information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or N parameters in the normalized mutual information of the first coding information and the second coding information, where N is a positive integer, and there is a mapping relationship between the second condition and the N parameters.

[0049] In one possible design, the communication module is specifically used to receive the first coded information from the first device; or, to receive quantization information corresponding to the first coded information from the first device, and to dequantize the quantization information to obtain the first coded information.

[0050] In one possible design, the communication module is further used to obtain second channel information; the processing module is further used to encode and quantize the second channel information according to the updated second encoder and the first quantization method to obtain third coded information; the communication module is further used to send the third coded information. Optionally, the first quantization method is one of a plurality of preconfigured quantization methods, or the first quantization method is indicated by the first device to the second device.

[0051] In a sixth aspect, the present application provides a communication device, which may be a first device, or a device, module, or chip applied to the first device, or a device that can be used in combination with the first device. In one design, the communication device may include a module that executes the method / operation / step / action described in the second aspect, and the module may be a hardware circuit, or software, or a combination of a hardware circuit and software. In one design, the communication device may include a processing module and a communication module.

[0052] The communication module is used to perform the following operations under the control of the processing module:

[0053] Sending first coded information or quantization information corresponding to the first coded information; wherein the first coded information corresponds to first channel information input to a first encoder applied to a first device, and a statistical dependency relationship between the first channel information and the first coded information meets a first condition; the first coded information is used to update a second encoder applied to a second device; wherein a statistical dependency relationship between an input and an output of the second encoder before the update meets the first condition.

[0054] In a possible design, the first encoder or the second encoder before updating is obtained through offline training.

[0055] In one possible design, the communication module is also used to receive first capability information from the second device before sending the first encoding information or quantization information corresponding to the first encoding information, and the first capability information is used to indicate that the second device supports updating the second encoder.

[0056] In one possible design, the communication module is also used to: receive a statistical dependency between the first coding information and the second coding information; wherein the second coding information corresponds to the first channel information input to the second encoder before updating, and the statistical dependency between the first channel information and the second coding information meets the first condition; and send a first indication information, wherein the first indication information is used to indicate that the statistical dependency between the first coding information and the second coding information meets the second condition, or the first indication information is used to trigger an update of the second encoder.

[0057] In one possible design, the statistical dependency relationship between two information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or M parameters in the normalized mutual information of the two information, where M is a positive integer; wherein the two information include the first channel information and the first coding information, or the first channel information and the second coding information, and the first condition is a preconfigured condition that has a mapping relationship with the M parameters.

[0058] In one possible design, the communication module is also used to send third indication information, and the third indication information is used to indicate a method for determining a statistical dependency relationship between the first coding information and the second coding information, and the second condition corresponding to the determination method.

[0059] In one possible design, the statistical dependency between the first coding information and the second coding information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or N parameters in the normalized mutual information of the first coding information and the second coding information, where N is a positive integer, and there is a mapping relationship between the second condition and the N parameters.

[0060] In a seventh aspect, the present application provides a communication device, which may be a first device, or a device, module, or chip applied to the first device, or a device that can be used in combination with the first device. In one design, the communication device may include a module that executes the method / operation / step / action described in the third aspect, and the module may be a hardware circuit, or software, or a combination of a hardware circuit and software. In one design, the communication device may include a processing module and a communication module.

[0061] A communication module, configured to obtain second encoding information; wherein the second encoding information corresponds to first channel information input to a second encoder applied to a second device, and a statistical dependency relationship between the first channel information and the second encoding information meets a first condition;

[0062] A processing module, used to determine a statistical dependency between first coding information and second coding information; wherein the first coding information corresponds to the first channel information input to a first encoder applied to a first device, and the statistical dependency between the first channel information and the first coding information meets the first condition; when the statistical dependency between the first coding information and the second coding information meets the second condition, the first decoder applied to the first device is updated according to the second coding information, and the first decoder before the update matches the first encoder.

[0063] In one possible design, the difference between the first channel information and the third channel information is less than a set threshold; wherein the third channel information is the output of the updated first decoder, and the third channel information corresponds to the second encoded information input to the updated first decoder.

[0064] In a possible design, the first encoder, the first decoder before updating, or the second encoder is obtained through offline training.

[0065] In a possible design, the statistical dependency relationship between two information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or M parameters in the normalized mutual information of the two information, where M is a positive integer; wherein the two information include the first channel information and the first coding information, or the first channel information and the second coding information; and the first condition is a preconfigured condition that has a mapping relationship with the M parameters.

[0066] In one possible design, the statistical dependency between the first coded information and the second coded information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or N parameters in the normalized mutual information of the first coded information and the second coded information, where N is a positive integer; the second condition is a preconfigured condition that has a mapping relationship with the N parameters.

[0067] In one possible design, the communication module is further used to send a first request message to the second device before obtaining the second coding information, where the first request message is used to request the second coding information.

[0068] In one possible design, the communication module is specifically used to receive the second coded information from the second device; or, to receive quantization information corresponding to the second coded information from the second device, and to dequantize the quantization information to obtain the second coded information.

[0069] In one possible design, the communication module is also used to receive fourth coded information from a second device, where the fourth coded information is determined based on the second encoder and the first quantization method; the processing module is also used to dequantize and demodulate the fourth coded information according to the updated first decoder and the first quantization method to obtain channel information corresponding to the fourth coded information.

[0070] In a possible design, the first quantization method is one of multiple preconfigured quantization methods.

[0071] In one possible design, the communication module is also used to send the first quantization method to the second device.

[0072] In an eighth aspect, the present application provides a communication device, which may be a second device, or a device, module, or chip applied to the second device, or a device that can be used in combination with the second device. In one design, the communication device may include a module that performs the method / operation / step / action described in the fourth aspect, and the module may be a hardware circuit, or software, or a combination of a hardware circuit and software. In one design, the communication device may include a processing module and a communication module.

[0073] The communication module is used to perform the following operations under the control of the processing module:

[0074] Sending second coded information or quantization information corresponding to the second coded information; wherein the second coded information corresponds to first channel information input to a second encoder applied to a second device, and a statistical dependency relationship between the first channel information and the second coded information meets a first condition; the second coded information is used to update a first decoder applied to a first device; wherein the statistical dependency relationship between the input and output of the first encoder applied to the first device meets the first condition, and the first decoder before the update matches the first encoder.

[0075] In a possible design, the first encoder, the first decoder before updating, or the second encoder is obtained through offline training.

[0076] In one possible design, the statistical dependency between the first channel information and the second coded information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or M parameters in the normalized mutual information of the first channel information and the second coded information, where M is a positive integer; and the first condition is a preconfigured condition that has a mapping relationship with the M parameters.

[0077] In one possible design, the communication module is also used to receive first request information before sending the second coding information or quantization information corresponding to the second coding information, where the first request information is used to request the second coding information.

[0078] In one possible design, the communication module is also used to obtain fourth channel information; the processing module is also used to encode and quantize the fourth channel information according to the second encoder and the first quantization method to obtain fourth encoded information; the communication module is also used to send the fourth encoded information.

[0079] In a possible design, the first quantization method is one of a plurality of preconfigured quantization methods, or the first quantization method is indicated by the first device to the second device.

[0080] In a ninth aspect, an embodiment of the present application provides a communication device, wherein the communication device includes a processing unit, for example, the processing unit is a processor, and the processor is used to execute computer program instructions to implement the method described in any one of the first to fourth aspects above. The communication device may also include a communication circuit, and the communication circuit is used for the communication device to communicate with other devices. Exemplarily, the communication interface may include a circuit, a bus, a module, a pin or other types of communication interfaces, or include a transceiver. Optionally, the processor is coupled to a memory, and the memory is used to store the aforementioned computer program instructions. When the processor specifically executes the computer program instructions stored in the memory, the method described in any one of the first to fourth aspects can be implemented. In one implementation, the communication device may also include the memory.

[0081] In a tenth aspect, the present disclosure provides a communication system, comprising a communication device as described in the fifth and sixth aspects; or a communication device as described in the seventh and eighth aspects.

[0082] In an eleventh aspect, the present disclosure further provides a computer program, which, when executed on a computer, enables the computer to execute the method provided in any one of the first to fourth aspects above.

[0083] In a twelfth aspect, the present disclosure further provides a computer program product, comprising instructions, which, when executed on a computer, enable the computer to execute the method provided in any one of the first to fourth aspects above.

[0084] In the thirteenth aspect, the present disclosure further provides a computer-readable storage medium, in which a computer program or instruction is stored. When the computer program or instruction is executed on a computer, the computer executes the method provided in any one of the first to fourth aspects above.

[0085] In a fourteenth aspect, the present disclosure further provides a chip, the chip is used to read a computer program stored in a memory, and execute the method provided in any one of the first to fourth aspects above, or the chip includes a circuit for executing the method provided in any one of the first to fourth aspects above. Optionally, the chip includes a processor and an input / output interface.

[0086] In a fifteenth aspect, the present disclosure further provides a chip system, which includes a processor for supporting a device to implement the method provided in any one of the first to fourth aspects above. In a possible design, the chip system also includes a memory, which is used to store programs and data necessary for the device. The chip system can be composed of a chip, or it can include a chip and other discrete devices.

[0087] The effects of the solutions provided in any of the second to fifteenth aspects above can be referred to the corresponding description in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] Figure 1 A schematic diagram of the structure of a communication system;

[0089] Figure 2A A schematic diagram of the structure of a neuron;

[0090] Figure 2B A schematic diagram of the layer relationship of a neural network;

[0091] Figure 2C This is a schematic diagram of an AI application framework;

[0092] Figure 2D is a schematic diagram of the structure of another communication system;

[0093] Figure 3A to Figure 3D Schematic diagrams of several network architectures;

[0094] Figure 4 A schematic diagram of an offline training framework provided in an embodiment of the present application;

[0095] Figure 5A One of the flow charts of the communication method provided in the embodiment of the present application;

[0096] Figure 5B One of the flow charts of the communication method provided in the embodiment of the present application;

[0097] Figure 6 One of the structural schematic diagrams of the communication device provided in the embodiment of the present application;

[0098] Figure 7 This is one of the structural diagrams of the communication device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0099] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings.

[0100] The at least one (item) involved in the present application as follows indicates one (item) or more (items). More than one (item) refers to two (items) or more than two (items). "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. In addition, it should be understood that although the terms first, second, etc. may be used to describe each object in the present application, these objects should not be limited to these terms. These terms are only used to distinguish each object from each other.

[0101] The terms "including" and "having" and any variations thereof mentioned in the following description of the present application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes other steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices. It should be noted that in the present application, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any method or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other methods or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete manner.

[0102] The technical solution provided in the present application can be applied to various communication systems, such as: the fifth generation (5th generation, 5G) or new radio (new radio, NR) system, long term evolution (long term evolution, LTE) system, LTE frequency division duplex (frequency division duplex, FDD) system, LTE time division duplex (time division duplex, TDD) system, wireless local area network (wireless local area network, WLAN) system, satellite communication system, future communication system, such as the sixth generation (6th generation, 6G) mobile communication system, or a fusion system of multiple systems, etc. The technical solution provided in the present application can also be applied to device to device (D2D) communication, vehicle-to-everything (V2X) communication, machine to machine (M2M) communication, machine type communication (MTC), and Internet of things (IoT) communication system or other communication systems.

[0103] A network element in a communication system can send a signal to another network element or receive a signal from another network element. The signal may include information, signaling or data, etc. The network element may also be replaced by an entity, a network entity, a device, a communication device, a communication module, a node, a communication node, etc. The network element is used as an example for description in this application. For example, a communication system may include at least one terminal device and at least one access network device. The access network device may send a downlink signal to the terminal device, and / or the terminal device may send an uplink signal to the access network device. In addition, it can be understood that if a plurality of terminal devices are included in the communication system, the plurality of terminal devices may also send signals to each other, that is, the signal sending network element and the signal receiving network element may both be terminal devices.

[0104] The communication method provided in this application can be applied to wireless communication systems such as 5G, 6G, and satellite communications. Figure 1 , Figure 1 is a simplified schematic diagram of the wireless communication system provided by the present application. Figure 1As shown, the wireless communication system includes a wireless access network 100. The wireless access network 100 may be a next generation (e.g., 6G or higher) wireless access network, or a traditional (e.g., 5G, 4G, 3G, or 2G) wireless access network. One or more communication devices (120a-120j, collectively referred to as 120) may be connected to each other or to one or more network devices (110a, 110b, collectively referred to as 110) in the wireless access network 100. Optionally, Figure 1 This is just a schematic diagram. The wireless communication system may also include other devices, such as core network devices, wireless relay devices and / or wireless backhaul devices. Figure 1 Not drawn in.

[0105] Optionally, in practical applications, the wireless communication system may include multiple network devices (also referred to as access network devices) at the same time, or may include multiple communication devices at the same time. A network device may serve one or more communication devices at the same time. A communication device may also access one or more network devices at the same time. This application does not limit the number of communication devices and network devices included in the wireless communication system.

[0106] The network device may be an entity on the network side for transmitting or receiving signals. The network device may be an access device for a communication device to access the wireless communication system in a wireless manner, such as a base station. The base station can broadly cover the following various names, or be replaced with the following names, such as: NodeB, evolved NodeB (eNB), next generation NodeB (gNB), access network equipment in open radio access network (O-RAN), relay station, access point, transmission point (transmitting and receiving point, TRP), transmission point (transmitting point, TP), master station MeNB, secondary station SeNB, multi-standard radio (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), radio head (RRH), central unit (CU), distributed unit (DU), radio unit (radio unit, RU), centralized unit control plane (CUcontrol plane, CU-CP) node, centralized unit user plane (CU user plane, CU-UP) node, positioning node, etc. The base station can be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. The network device may also refer to a communication module, a modem or a chip used to be arranged in the aforementioned device or apparatus. The network device may also be a mobile switching center and a device to device (Device-to-Device, D2D), vehicle external connection (vehicle-to-everything, V2X), a device that performs the base station function in machine-to-machine (machine-to-machine, M2M) communication, a network side device in a 6G network, and a device that performs the base station function in a future communication system. The network device may support networks with the same or different access technologies. The embodiments of the present application do not limit the specific technology and specific device form adopted by the network device.

[0107] The network equipment may be fixed or mobile. For example, the base stations 110 a and 110 b are stationary and are responsible for wireless transmission and reception in one or more cells from the communication device 120 . Figure 1 The helicopter or drone 120i shown in the figure can be configured to act as a mobile base station, and one or more cells can move according to the location of the mobile base station 120i. In other examples, the helicopter or drone (120i) can be configured to be used as a communication device for communicating with the base station 110b.

[0108] The network device in the embodiment of the present application may be an integrated base station, or may be a base station including a centralized unit (CU) and / or a distributed unit (DU). A base station including a CU and a DU may also be referred to as a base station with CU and DU separated, such as the base station including a gNB-CU and a gNB-DU. Among them, the CU may also be separated into a CU control plane (CU control plane, CU-CP) and a CU user plane (CU user plane, CU-UP), such as the base station including a gNB-CU-CP, a gNB-CU-UP and a gNB-DU. Alternatively, the network device in the embodiment of the present application may also be an antenna unit (radiounit, RU). Alternatively, the network device in the embodiment of the present application may also be an open radio access network (O-RAN) architecture, etc. The embodiment of the present application does not limit the specific deployment method of the network device. Exemplarily, when the network device is an O-RAN architecture, the network device shown in the embodiment of the present application may be an access network device in O-RAN, such as a combination of one or more of CU, DU, or RU, or a module in an access network device. In the ORAN system, CU may also be referred to as open (O)-CU, CU-CP may also be referred to as O-CU-CP, CU-UP may also be referred to as O-CU-UP, and RU may also be referred to as O-RU.

[0109] In the present application, the communication device used to implement the above access network function can be an access network device, or a network device with some functions of accessing the network, or a device capable of supporting the implementation of the access network function, such as a chip system, a hardware circuit, a software module, or a hardware circuit plus a software module, which can be installed in the access network device or used in combination with the access network device. In the method of the present application, the communication device used to implement the access network device function is an access network device for example.

[0110] The communication device can be an entity on the user side for receiving or transmitting signals, such as a mobile phone. The communication device can be used to connect people, objects and machines. The communication device can communicate with one or more core networks through a network device. The communication device includes a handheld device with a wireless connection function, other processing devices connected to a wireless modem, or a vehicle-mounted device. The communication device can be a portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile device. The communication device 120 can be widely used in various scenarios, such as cellular communication, device-to-device D2D, vehicle-to-all V2X, end-to-end P2P, machine-to-machine M2M, machine type communication MTC, Internet of Things IOT, virtual reality VR, augmented reality AR, industrial control, automatic driving, telemedicine, smart grid, smart furniture, smart office, smart wear, smart transportation, smart city, drone, robot, remote sensing, passive sensing, positioning, navigation and tracking, autonomous delivery and mobility, etc. Some examples of the communication device 120 are: user equipment (UE) of 3GPP standard, fixed equipment, mobile equipment, handheld equipment, wearable equipment, cellular phone, smart phone, session initiation protocol (SIP) phone, laptop computer, personal computer, smart book, vehicle, satellite, global positioning system (GPS) equipment, target tracking equipment, drone, helicopter, aircraft, ship, remote control equipment, smart home equipment, industrial equipment, personal communication service (PCS) phone, wireless local loop (WLL) station, personal digital assistant (PDA), wireless network camera, tablet computer, PDA, mobile internet device (MID), wearable equipment such as smart watch, virtual reality (VR) equipment, augmented reality (AR) equipment, wireless terminal in industrial control, terminal in vehicle networking system, wireless terminal in self-driving, wireless terminal in smart grid, wireless terminal in transportation safety, smart city, etc. The communication device 120 may be a wireless device in the above scenarios or a device used to be set in a wireless device, for example, a communication module, a modem or a chip in the above devices.The communication device may also be referred to as a terminal, terminal equipment, user equipment (UE), mobile station (MS), mobile terminal (MT), etc. The communication device may also be a communication device in a future wireless communication system. The communication device may be used in a dedicated network device or a general device. The embodiments of the present application do not limit the specific technology and specific device form used by the communication device.

[0111] Optionally, the communication device may be used to act as a base station. For example, the UE may act as a scheduling entity that provides sidelink signals between UEs in V2X, D2D, or P2P, etc. Figure 1 As shown, the cell phone 120a and the car 120b communicate with each other using a sidelink signal. The cell phone 120a and the smart home device 120e communicate with each other without relaying the communication signal through the base station 110b.

[0112] In the present application, the communication device used to implement the functions of the communication device may be a terminal device, or a terminal device having some of the functions of the above communication device, or a device capable of supporting the implementation of the functions of the above communication device, such as a chip system, which may be installed in the terminal device or used in combination with the terminal device. In the present application, the chip system may be composed of a chip, or may include a chip and other discrete devices. In the technical solution provided in the present application, the communication device is described as a terminal device or UE as an example.

[0113] Optionally, a wireless communication system is usually composed of cells, and the base station provides management of the cell. The base station provides communication services to multiple mobile stations (MS) in the cell. The base station includes a baseband unit (BBU) and a remote radio unit (RRU). The BBU and RRU can be placed in different places, for example: the RRU is remote and placed in an area with high traffic volume, and the BBU is placed in a central computer room. The BBU and RRU can also be placed in the same computer room. The BBU and RRU can also be different components under one rack. Optionally, a cell can correspond to a carrier or a component carrier.

[0114] It can be understood that the present application can be applied between a network device and a communication device, between a network device and a network device, or between a communication device and a communication device, that is, between a primary device and a secondary device. The primary device can be a network device or a communication device. When the primary device is a network device, the secondary device can be another network device or a communication device. When the primary device is a communication device, the secondary device can be another communication device.

[0115] The following describes the scheme by taking the master device as a network device, such as an access network device, and the slave device as a communication device, such as a terminal device, as an example. The communication direction corresponding to the downlink is the transmission from the master device to the slave device, and the communication direction corresponding to the uplink is the transmission from the slave device to the master device.

[0116] Protocol layer structure between access network equipment and terminal equipment

[0117] The communication between the access network device and the terminal device follows a certain protocol layer structure. The protocol layer structure may include a control plane protocol layer structure and a user plane protocol layer structure. For example, the control plane protocol layer structure may include the functions of the protocol layers such as the radio resource control (RRC) layer, the packet data convergence protocol (PDCP) layer, the radio link control (RLC) layer, the medium access control (MAC) layer and the physical layer. For example, the user plane protocol layer structure may include the functions of the protocol layers such as the PDCP layer, the RLC layer, the MAC layer and the physical layer. In a possible implementation, the service data adaptation protocol (SDAP) layer may also be included above the PDCP layer.

[0118] Optionally, the protocol layer structure between the access network device and the terminal may also include an artificial intelligence (AI) layer for transmitting data related to AI functions.

[0119] Taking data transmission between access network equipment and terminal equipment as an example, data transmission needs to pass through the user plane protocol layer, such as the SDAP layer, PDCP layer, RLC layer, MAC layer, and physical layer. Among them, the SDAP layer, PDCP layer, RLC layer, MAC layer, and physical layer can also be collectively referred to as the access layer. According to the transmission direction of the data, it is divided into sending or receiving, and each of the above layers is divided into a sending part and a receiving part. Taking downlink data transmission as an example, after the PDCP layer obtains data from the upper layer, it transmits the data to the RLC layer and the MAC layer, and then the MAC layer generates a transmission block, and then transmits it wirelessly through the physical layer. The data is encapsulated accordingly in each layer. For example, the data received by a layer from the upper layer of the layer is regarded as the service data unit (SDU) of the layer, which becomes a protocol data unit (PDU) after being encapsulated by the layer, and then passed to the next layer.

[0120] Exemplarily, the terminal device may also have an application layer and a non-access layer. The application layer may be used to provide services to applications installed in the terminal device. For example, downlink data received by the terminal device may be sequentially transmitted from the physical layer to the application layer, and then provided to the application by the application layer; for another example, the application layer may obtain data generated by the application, and sequentially transmit the data to the physical layer and send it to other communication devices. The non-access layer may be used to forward user data, such as forwarding uplink data received from the application layer to the SDAP layer or forwarding downlink data received from the SDAP layer to the application layer.

[0121] Structure of access network equipment

[0122] The access network equipment may include a centralized unit (CU) and a distributed unit (DU). Multiple DUs may be centrally controlled by one CU. As an example, the interface between the CU and the DU may be referred to as an F1 interface. Among them, the control plane (CP) interface may be F1-C, and the user plane (UP) interface may be F1-U. The CU and DU may be divided according to the protocol layers of the wireless network: for example, the functions of the PDCP layer and above are set in the CU, and the functions of the protocol layers below the PDCP layer (such as the RLC layer and the MAC layer, etc.) are set in the DU; for another example, the functions of the protocol layers above the PDCP layer are set in the CU, and the functions of the protocol layers below the PDCP layer are set in the DU.

[0123] It is understandable that the above-mentioned division of the processing functions of CU and DU according to the protocol layer is only an example, and it can also be divided in other ways, for example, CU or DU can be divided into functions with more protocol layers, and for example, CU or DU can also be divided into partial processing functions with protocol layers. In one design, some functions of the RLC layer and the functions of the protocol layers above the RLC layer are set in the CU, and the remaining functions of the RLC layer and the functions of the protocol layers below the RLC layer are set in the DU. In another design, the functions of the CU or DU can also be divided according to the service type or other system requirements, for example, divided by latency, and the functions whose processing time needs to meet the latency requirements are set in the DU, and the functions that do not need to meet the latency requirements are set in the CU. In another design, the CU can also have one or more functions of the core network. Exemplarily, the CU can be set on the network side to facilitate centralized management. In another design, the RU of the DU is set remotely. Among them, the RU has a radio frequency function.

[0124] Optionally, DU and RU can be divided at the physical layer (PHY). For example, DU can implement high-level functions in the PHY layer, and RU can implement low-level functions in the PHY layer. Wherein, when used for sending, the functions of the PHY layer may include adding cyclic redundancy check (CRC) code, channel coding, rate matching, scrambling, modulation, layer mapping, precoding, resource mapping, physical antenna mapping, and / or RF transmission functions. When used for receiving, the functions of the PHY layer may include CRC, channel decoding, rate matching, descrambling, demodulation, layer mapping, channel detection, resource demapping, physical antenna demapping, and / or RF receiving functions. Wherein, the high-level functions in the PHY layer may include a part of the functions of the PHY layer, such as the part of the functions that is closer to the MAC layer, and the low-level functions in the PHY layer may include another part of the functions of the PHY layer, such as the part of the functions that is closer to the RF function. For example, the high-level functions in the PHY layer may include adding CRC code, channel coding, rate matching, scrambling, modulation, and layer mapping, and the low-level functions in the PHY layer may include precoding, resource mapping, physical antenna mapping, and RF transmission functions; or, the high-level functions in the PHY layer may include adding CRC code, channel coding, rate matching, scrambling, modulation, layer mapping and precoding, and the low-level functions in the PHY layer may include resource mapping, physical antenna mapping, and RF transmission functions.

[0125] Exemplarily, the functions of CU can be implemented by one entity, or can be implemented by different entities. For example, the functions of CU can be further divided, that is, the control plane and the user plane are separated and implemented by different entities, namely the control plane CU entity (i.e., CU-CP entity) and the user plane CU entity (i.e., CU-UP entity). The CU-CP entity and the CU-UP entity can be coupled with the DU to jointly complete the functions of the access network device.

[0126] In the above architecture, the signaling generated by the CU can be sent to the terminal device through the DU, or the signaling generated by the terminal device can be sent to the CU through the DU. For example, the signaling of the RRC or PDCP layer will eventually be processed into the signaling of the physical layer and sent to the terminal device, or converted from the received signaling of the physical layer. In this architecture, the signaling of the RRC or PDCP layer can be considered to be sent through the DU, or through the DU and the RU.

[0127] Optionally, any of the above DU, CU, CU-CP, CU-UP and RU can be a software module, a hardware structure, or a software module + hardware structure, without limitation. Among them, the existence forms of different entities can be different, without limitation. For example, DU, CU, CU-CP, CU-UP are software modules, and RU is a hardware structure. These modules and their execution methods are also within the scope of protection of this application.

[0128] It should be understood that Figure 1 The number and type of each device in the communication system shown are for illustration only, and the present application is not limited thereto. In actual applications, the communication system may also include more terminal devices, more access network devices, and other network elements, such as core network devices, network management and / or network elements for implementing artificial intelligence functions. Among them, the network management can also be called operation administration and maintenance (OAM) network element, referred to as OAM. Operations mainly complete the analysis, prediction, planning and configuration of daily networks and services; maintenance mainly involves daily operational activities such as testing and fault management of the network and its services. The network management can detect the network operation status, optimize network connections and performance, improve network operation stability, and reduce network maintenance costs.

[0129] It is understandable that all or part of the functions implemented by one or more of the terminal equipment, access network equipment, or core network equipment can be virtualized, that is, implemented by one or more of the proprietary processors or general-purpose processors and corresponding software modules. Among them, the terminal equipment and access network equipment involve interfaces for air interface transmission, and the transceiver functions of the interfaces can be implemented by hardware. Core network equipment, such as the aforementioned OAM, can be virtualized. Optionally, one or more functions of the virtualized terminal equipment, access network equipment or core network equipment can be implemented by cloud devices, such as cloud devices in over the top (OTT) systems.

[0130] The method provided in the present application can be used for communication between access network equipment and terminal equipment, and can also be used for communication between other communication equipment, such as communication between macro base stations and micro base stations in a wireless backhaul link, communication between two terminal devices in a side link (sidelink, SL), communication between terminal equipment and network management (OAM), communication between terminal equipment and core network equipment, etc., without limitation.

[0131] The method provided in this application involves artificial intelligence (AI). For ease of understanding, some terms related to AI in this application are introduced below in combination with A1 to A4. It is understood that this introduction does not limit this application.

[0132] A1, AI model

[0133] The AI ​​model is a specific implementation of the AI ​​technology function. The AI ​​model represents the mapping relationship between the input and output of the model. The AI ​​model can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q learning model, or other machine learning (ML) models. In this application, the AI ​​model is referred to as the model.

[0134] The embodiments of the present application mainly relate to auto-encoders (AE). An auto-encoder is a neural network for unsupervised learning, which is characterized by taking input data as label data, so an auto-encoder can also be understood as a neural network for self-supervised learning. An auto-encoder can generally refer to a network structure composed of two sub-models. An auto-encoder can also be called a bilateral model, a dual-end model or a collaborative model. For example, an auto-encoder includes an encoder and a decoder. The encoder and the decoder are matched and used, and it can be understood that the encoder and the decoder are matching AI models. An auto-encoder can be used for data compression and recovery. For example, the encoder in the auto-encoder can encode (for example, compress) the original information A to obtain the encoded information B; the decoder in the auto-encoder can decode the encoded information B to obtain the decoded information A' corresponding to the original information A. Optionally, the decoded information A' can also be described as the recovered original information A. It can be understood that the decoder is the inverse operation of the encoder; for example, the encoding process of the encoder includes compression processing, and the decoding process of the decoder includes decompression processing. The encoder and decoder included in the same auto-encoder can be deployed on different devices respectively. In addition, it can be understood that the present application describes the autoencoder including an encoder and a decoder as an example, but the encoder can also be replaced by describing it as a generation model, a generation sub-model or other names, and the decoder can also be replaced by describing it as a reconstruction model, a reconstruction sub-model or other names, and the present application does not limit this.

[0135] The autoencoder can be applied to common scenarios such as channel state information (CSI) compression feedback and CSI prediction. Among them, CSI is a kind of channel information, which is information that can reflect channel characteristics and channel quality. Among them, channel information can also be called channel response. Exemplarily, CSI can be expressed by a channel matrix, for example, CSI includes a channel matrix, or CSI can be composed of a channel eigenvector. Taking CSI compression feedback as an example, in a frequency division duplex (FDD) communication scenario, since the uplink and downlink channels are not reciprocal or the reciprocity of the uplink and downlink channels cannot be guaranteed, the access network device usually sends a downlink reference signal to the terminal device, and the terminal device performs channel measurement and interference measurement based on the received downlink reference signal to estimate the original channel information, and the original channel information includes CSI, and then the terminal device can use the encoder to compress the original channel information and feed it back to the access network device, and the access network device side can use the decoder to reconstruct (or restore) the compressed original channel information to obtain the reconstructed channel information.

[0136] A2, Neural Network

[0137] Neural network is a specific implementation form of AI or machine learning technology. According to the universal approximation theorem, neural network can theoretically approximate any continuous function, thus enabling neural network to learn any mapping.

[0138] The idea of ​​neural network comes from the neuron structure of brain tissue. For example, each neuron in a deep neural network (DNN) performs a weighted sum operation on its input value and outputs the operation result through an activation function. Figure 2A As shown in Figure 1, it is a schematic diagram of the neuron structure. Assume that the input of the neuron is x = [x0, x1, ..., x n ], and the weights corresponding to each input are w=[w,w1,…,w n ], where w i As x i The weight of x i Weighted. The bias of weighted summation of input values ​​according to the weight is, for example, b. The activation function can take many forms. Assuming that the activation function of a neuron is: y = f(c) = max(0, c), the output of the neuron is: For another example, the activation function of a neuron is: y = f(c) = c, then the output of the neuron is: Among them, b, w i 、x iIt can be a decimal, an integer (such as 0, a positive integer or a negative integer), or a complex number. The activation functions of different neurons in a neural network can be the same or different.

[0139] Depending on how the network is constructed, DNNs can include feedforward neural networks (FNNs), convolutional neural networks (CNNs), or recurrent neural networks (RNNs).

[0140] A neural network generally includes multiple layers, and each layer may include one or more neurons. By increasing the depth and / or width of the neural network, the expressive power of the neural network can be improved, providing more powerful information extraction and abstract modeling capabilities for complex systems. Among them, the depth of a neural network can refer to the number of layers included in the neural network, and the number of neurons included in each layer can be called the width of the layer. In one implementation, the neural network includes an input layer and an output layer. The input layer of the neural network processes the received input information through neurons, and passes the processing results to the output layer, and the output layer obtains the output result of the neural network. In another implementation, the neural network includes an input layer, a hidden layer, and an output layer, which can be referenced Figure 2B In the schematic FNN, neurons in adjacent layers of the FNN are fully connected to each other. The input layer of the neural network processes the received input information through neurons and passes the processing results to the middle hidden layer. The hidden layer calculates the received processing results to obtain the calculation results. The hidden layer passes the calculation results to the output layer or the adjacent hidden layer, and finally the output layer obtains the output result of the neural network. Among them, a neural network can include one hidden layer, or include multiple hidden layers connected in sequence, without limitation.

[0141] A3, training data and inference data

[0142] The training data set is used for model training. The training data set may include the input data of the model, or the input data and target output data of the model. The training data set includes one or more training data. The training data may be the data input to the model or the target output data of the model. The target output data may also be referred to as a label, output label data or output label sample. The training data set is one of the important parts of machine learning. Model training is essentially learning certain features of the training data from the training data so that the output data of the model is as close to the target output data as possible, such as the difference between the output data of the model and the target output data is as small as possible. The composition and selection of the training data set can determine the performance of the trained model to a certain extent.

[0143] In addition, during the training process of a model (such as a neural network), a loss function can be defined. The loss function describes the gap or difference between the output value of the model and the target output value. This application does not limit the specific form of the loss function. The training process of the model is to adjust the model parameters of the model so that the value of the loss function is less than the threshold, or the value of the loss function meets the target requirements.

[0144] Among them, the model parameters may include one or more of the following: structural parameters of the model (such as the number of layers and / or weights of the model, etc.), for example, if the model is a neural network, the structural parameters of the neural network include at least one of the following: the number of layers, width, weights of neurons, or parameters in the activation function of neurons of the neural network; input parameters of the model (such as input dimension, number of input ports); output parameters of the model (such as output dimension, number of output ports). It can be understood that the input dimension may refer to the size of an input data, for example, when the input data is a sequence, the input dimension corresponding to the sequence may indicate the length of the sequence. The number of input ports may refer to the number of input data. Similarly, the output dimension may refer to the size of an output data, for example, when the output data is a sequence, the output dimension corresponding to the sequence may indicate the length of the sequence. The number of output ports may refer to the number of output data.

[0145] Inference data can be used as input to a trained model for model reasoning, model performance verification or monitoring. In the model reasoning process, the inference data is input into the model, and the corresponding output, i.e., the reasoning result, can be obtained. Optionally, the input data of the model included in the training data set can also be used as inference data for model reasoning, model performance verification or monitoring.

[0146] A4, AI model design

[0147] The design of AI models mainly includes data collection (such as collecting training data and / or inference data), model training, and model inference. It can also include the application of inference results. Figure 2CAn AI application framework is illustrated. In the aforementioned data collection link, the data source is used to provide training data sets and inference data. In the model training link, an AI model is obtained by analyzing or training the training data provided by the data source. Among them, the AI ​​model represents the mapping relationship between the input and output of the model. Learning the AI ​​model through the model training node is equivalent to learning the mapping relationship between the input and output of the model using the training data. In the model reasoning link, the AI ​​model trained through the model training link is used to reason based on the inference data provided by the data source to obtain the reasoning result. This link can also be understood as: inputting the reasoning data into the AI ​​model, and obtaining the output data through the AI ​​model, and the output data is the reasoning result. The reasoning result may indicate: configuration parameters used (executed) by the execution object, and / or operations performed by the execution object. In this application, an AI model can infer one parameter, or infer multiple parameters. The reasoning results are published in the reasoning result application link. For example, the reasoning results can be uniformly planned by the execution (actor) entity. For example, the execution entity can send the reasoning results to one or more execution objects (for example, core network equipment, access network equipment, terminal equipment or network management, etc.) for execution.

[0148] It is understandable that a network element with artificial intelligence function may be included in the communication system. The above-mentioned AI model design-related links may be performed by one or more network elements with artificial intelligence function. In one possible design, an AI function (such as an AI module or an AI entity) may be configured in an existing network element in the communication system to implement AI-related operations, such as training and / or reasoning of an AI model. For example, the existing network element may be an access network device (such as a gNB), a terminal device, a core network device, or a network management system. The operation mainly completes the analysis, prediction, planning and configuration of daily networks and services; maintenance is mainly daily operation activities such as testing and fault management of the network and its services. The network management system can detect the network operation status, optimize network connection and performance, improve network operation stability, and reduce network maintenance costs. Or in another possible design, an independent network element may also be introduced into the communication system to perform AI-related operations, such as training an AI model. The independent network element may be referred to as an AI network element or an AI node, etc., and this application does not limit this name. The AI ​​network element may be directly connected to the access network device in the communication system, or it may be indirectly connected to the access network device through a third-party network element. The third-party network element may be an authentication management function (AMF) network element, a user plane function (UPF) network element, or other core network equipment, a network management system, a cloud server, or other network elements, without limitation. Figure 2D , illustrating a communication system, the communication system includes a network device 110, a terminal device 120 and a terminal device 130; and, Figure 2D The illustrated communication system also introduces an AI network element 140. Optionally, the AI ​​network element may be located in a server, such as a host or cloud server of the OTT system.

[0149] At present, the training of the autoencoder is usually that the access network device trains the paired encoder and decoder, and then sends the trained encoder or the model parameters of the encoder to the terminal device; or the terminal device trains the paired encoder and decoder, and then sends the trained decoder or the model parameters of the decoder to the access network device, or the terminal device and the network device jointly train the autoencoder, for example, the terminal device side trains the encoder, and the access network device side trains the decoder. During the training process: the terminal device sends the output data of the encoder to the access network device side, the access network device side inputs the output data of the encoder to the decoder, and then indicates the output data of the decoder to the terminal device, and the terminal device iteratively updates the encoder in combination with the gradient information corresponding to the decoder received. According to the above designs, a large amount of information interaction is required between the terminal device and the access network device during the training process of the autoencoder, which will generate a large communication overhead.

[0150] Based on this, an embodiment of the present application provides a training method for an autoencoder, including two stages: offline training and online adaptation. In the offline training stage, the terminal device and the access network device independently train a set of autoencoders according to the same constraints. The terminal device side trains the first autoencoder, and the access network device side trains the second autoencoder. The statistical dependency between the input and output of the encoder in the first autoencoder and the statistical dependency between the input and output of the encoder in the second autoencoder meet the same statistical constraints. In the offline training stage, there is no need for information exchange between the terminal device and the access network device, which can reduce communication overhead. Further in the online adaptation stage, the terminal device and the access network device can complete the adaptation between the encoder on the terminal device side and the decoder on the access network device side through a small amount of data exchange, that is, improve the performance of the model. The schemes in the offline training stage and the online adaptation stage are described in detail below in combination with specific implementation methods.

[0151] First combine Figure 3A to Figure 3D , introduces the network architecture to which the method provided by the present disclosure can be applied.

[0152] like Figure 3A As shown, in a first possible implementation, the access network device includes a near real-time access network intelligent control (RAN intelligent controller, RIC) module for model training and / or reasoning. For example, the near real-time RIC can obtain information on the network side and / or the terminal side from at least one of the CU, DU and RU, and the information may include training data or reasoning data. For example, the near real-time RIC can be used to train the model, and the trained model can also be used for reasoning. In addition, optionally, the near real-time RIC can submit the reasoning result to at least one of the CU, DU and RU. Optionally, the reasoning result can be exchanged between the CU and the DU. Optionally, the reasoning result can be exchanged between the DU and the RU, for example, the near real-time RIC submits the reasoning result to the DU, and the DU submits it to the RU.

[0153] like Figure 3AAs shown, in a second possible implementation, the access network device may include a non-real-time RIC (optionally, the non-real-time RIC may be located in the OAM or in the core network device) for model training and reasoning. For example, the non-real-time RIC may obtain information on the network side and / or the terminal side from at least one of the CU, DU and RU, and the information may include training data or reasoning data. For example, the non-real-time RIC is used to train the model, and the trained model may also be used for reasoning. In addition, optionally, the non-real-time RIC may submit the reasoning result to at least one of the CU, DU and RU. Optionally, the reasoning result may be exchanged between the CU and DU. Optionally, the reasoning result may be exchanged between the DU and RU, for example, the non-real-time RIC submits the reasoning result to the DU, and the DU submits it to the RU.

[0154] like Figure 3A As shown, in the third possible implementation, the access network device includes a near real-time RIC, and the access network device includes a non-real-time RIC (optionally, the non-real-time RIC can be located in the OAM or in the core network device). As with the second possible implementation described above, the non-real-time RIC can be used for model training and / or reasoning; and / or, as with the first possible implementation described above, the near real-time RIC can be used for model training and / or reasoning; and / or, the near real-time RIC can obtain AI model information from the non-real-time RIC, and obtain network side and / or terminal side information from at least one of the CU, DU and RU, and obtain the reasoning result using the information and the AI ​​model information. Optionally, the near real-time RIC can submit the reasoning result to at least one of the CU, DU and RU, and optionally, the CU and DU can exchange reasoning results, and optionally, the DU and RU can exchange reasoning results, for example, the near real-time RIC submits the reasoning result to the DU, and the DU submits it to the RU. For example, the near real-time RIC is used to train the model and use the trained model for reasoning. For example, the non-real-time RIC is used to train the model and use the trained model for reasoning. For example, the non-real-time RIC is used to train the model and submit the trained model to the near-real-time RIC, which uses the trained model for inference.

[0155] Figure 3B The figure shows an example of a network architecture to which the method provided by the present disclosure can be applied. Figure 3A , Figure 3B The CU is separated into CU-CP and CU-UP.

[0156] Figure 3C FIG. 1 is an example diagram of a network architecture to which the method provided by the present disclosure can be applied. Figure 3CAs shown, optionally, the access network device includes one or more AI entities, and the function of the AI ​​entity is similar to the above-mentioned near real-time RIC. Optionally, the OAM includes one or more AI entities, and the function of the AI ​​entity is similar to the above-mentioned non-real-time RIC. Optionally, the core network device includes one or more AI entities, and the function of the AI ​​entity is similar to the above-mentioned non-real-time RIC. When both the OAM and the core network device include AI entities, the models trained by their respective AI entities are different, and / or the models used for reasoning are different.

[0157] In the present disclosure, the difference in models includes at least one of the following differences: structural parameters of the model (such as the number of layers and / or weights of the model, etc.), input parameters of the model, or output parameters of the model.

[0158] Figure 3D The figure shows an example of a network architecture to which the method provided by the present disclosure can be applied. Figure 3C , Figure 3D The access network devices in the embodiment are separated into CU and DU. Optionally, the CU may include an AI entity, and the function of the AI ​​entity is similar to the above-mentioned near real-time RIC. Optionally, the DU may include an AI entity, and the function of the AI ​​entity is similar to the above-mentioned near real-time RIC. When both the CU and the DU include AI entities, the models trained by their respective AI entities are different, and / or the models used for reasoning are different. Optionally, the Figure 3D The CU in the CU is split into CU-CP and CU-UP. Optionally, one or more AI models can be deployed in the CU-CP. And / or, one or more AI models can be deployed in the CU-UP. Optionally, Figure 3C or Figure 3D In the example, the OAM of access network equipment and the OAM of core network equipment can be deployed separately and independently.

[0159] The schemes of the offline training phase and the online adaptation phase are described in detail below.

[0160] 1. Offline training phase

[0161] The access network device side and the terminal device side can independently train a set of autoencoders (including encoders and decoders) based on partially identical, completely identical or different training data sets. Figure 4 Taking the CSI compression feedback scenario as an example, the training data set may include multiple channel information. The first encoder and the first decoder are trained on the access network device side, and the second encoder and the second decoder are trained on the terminal device side. The input of the first encoder and the input of the second encoder may be the same or different channel information.

[0162] It is understandable that the first encoder and the first decoder trained on the access network device side are applied to the access network device. Optionally, the training of the first encoder and the first decoder on the access network device side can be implemented in one of the following two ways: training the first encoder and the first decoder in the access network device, or training the first encoder and the first decoder by an independent AI network element associated with the access network device. Similarly, the training of the first encoder and the first decoder on the terminal device side is applied to the terminal device. Optionally, the training of the second encoder and the second decoder on the terminal device side can be implemented in one of the following three ways: training the second encoder and the second decoder in the terminal device, or training the second encoder and the second decoder by an independent AI network element associated with the terminal device, or training the second encoder and the second decoder by a server to which the terminal device belongs.

[0163] Optionally, the channel information used by the terminal device side / access network device side in the offline training phase can be pre-configured. For example, the access network device can use the channel information collected from one or more terminal devices for offline training of the first encoder and the first decoder, and the terminal device can use some channel information obtained by historical estimation for offline training of the second encoder and the second decoder. For another example, the access network device / terminal device can offline train a specific autoencoder based on the channel information of a specific scenario, configuration, or site to adapt to specific scenario requirements and facilitate improving the compression performance of CSI.

[0164] In an optional implementation, the encoder involved in the embodiment of the present application may have compression and quantization functions, channel information is input to the encoder, and the corresponding output is the information after the channel information is compressed and quantized; accordingly, the decoder may have dequantization and decompression functions, the encoded information is input to the decoder, and the corresponding output decoded information is the information after the encoded information is dequantized and decompressed, that is, the reconstructed channel information. In another optional implementation, the encoder only has a compression function but not a quantization function, the channel information is input to the encoder, and the corresponding output encoded information is the information after the channel information is compressed. Accordingly, the decoder may have a decompression function but not a dequantization function, the encoded information is input to the decoder, and the corresponding output decoded information is the information after the encoded information is decompressed, that is, the reconstructed channel information. Optionally, the channel information is regarded as the input space of the encoder, and the encoded information corresponding to the channel information can also be replaced and described as latent space information.

[0165] In addition, it can be understood that the second decoder trained on the terminal device side and the first encoder trained on the network device side are only used in the offline training process. In the actual reasoning process, the terminal device side can only deploy the second encoder, and the access network device side can only deploy the first decoder; when the second encoder does not have the quantization function and the first decoder does not have the dequantizer function, the terminal device side can also deploy a quantizer after the second encoder to quantize the coded information output by the second encoder to obtain the quantization information corresponding to the coded information, and the access network device side can also deploy a dequantizer before the first decoder to dequantize the quantization information corresponding to the coded information to obtain the input of the first decoder, that is, the coded information.

[0166] Taking the second encoder trained on the terminal device side as an example, after offline training, the statistical dependency between the channel information input and the coding information output by the second encoder meets certain constraints; accordingly, the statistical dependency between the channel information input and the coding information output by the first encoder obtained by offline training on the access network device side also meets the same constraints.

[0167] Optionally, the constraint condition may be predefined; or specified by an entity participating in the model training, such as a terminal device or an access network device specifying the constraint condition; or may be determined by negotiation between the terminal device and the access network device.

[0168] In one possible design, there is a mapping relationship between the constraint and the aforementioned statistical dependency, such as different statistical dependencies correspond to different constraints. Optionally, the constraint corresponding to a statistical dependency may be one or more index requirements set for the statistical dependency. The input of the second encoder is denoted as V, and the output of the second encoder is denoted as z. After offline training, a set of statistical dependencies between V and z of the second encoder may meet the index requirements. For example, a set of statistical dependencies between V and z is greater than a set threshold, denoted as dep i (V,z)≥η i .

[0169] Among them, dep i (V,z) represents the i-th statistical dependency between the input and output of the second encoder, η iis the threshold value corresponding to the i-th statistical dependency, i=1,2,..,m, m is a positive integer, and m represents the number of types of statistical dependencies. Among them, for different statistical dependencies, the specific value of the threshold may be different. Optionally, the statistical method of the statistical dependency may be predefined, or specified by an entity participating in the training, or determined by negotiation between the terminal device side and the network device side. For each specific statistical dependency, the terminal device side and the network device side need to meet the same threshold constraint. Similarly, the threshold may be predefined, or specified by an entity participating in the training (such as an access network device), or determined by negotiation between the terminal device side and the access network device side. In short, the terminal device side and the access network device side use the same calculation method to determine the statistical dependency between the input and output of their respective offline trained encoders, and measure the statistical dependency based on the same constraints (such as thresholds).

[0170] For ease of implementation, the present application provides some examples of calculating statistical dependencies, as follows:

[0171] Example 1: Determine the statistical dependency between V and z based on the linear regression similarity between V and z.

[0172] set up Represents the matrix composed of input space samples, where n represents the number of training samples, d v Represents the feature dimension of the input space sample, n,d v are all positive integers; let Represents the matrix composed of latent space samples, d z is a positive integer, d z Represents the characteristic dimension of the latent space sample, then the similarity of the linear regression from V to z is measured as:

[0173]

[0174] in, is the linear regression matrix to be optimized, min(·) means taking the minimum value, represents the Frobenius norm of the matrix, Q z For z 1:n The orthogonal basis of column vectors can be expressed as or any orthogonal transformation thereof, (·) T Represents the matrix transpose. Similarly, the similarity calculated from the linear regression of z to V is:

[0175]

[0176] Furthermore, the input is and The function to measure the statistical dependence between V and z is depi (V,z), including but not limited to one or more of the following:

[0177] dep i (V,z) is Self or Any non-zero power of ;

[0178] dep i (V,z) is Self or Any non-zero power of ;

[0179] dep i (V,z) is and The arithmetic mean, geometric mean, harmonic mean or weighted mean of the data; wherein the coefficient of the weighted mean may be predefined;

[0180] Any non-zero power and The result of the arithmetic mean, geometric mean, harmonic mean or weighted mean corresponding to the same power; wherein the coefficient of the weighted mean may be predefined.

[0181] Example 2: Based on the canonical correlation coefficient of V and z, the statistical dependence relationship between V and z is determined.

[0182] set up Represents the matrix composed of input space samples, where n represents the number of training samples, d v Represents the feature dimension of the input space sample, n,d v are all positive integers; let Represents the matrix composed of latent space samples, d z is a positive integer, d z represents the characteristic dimension of the latent space sample. First, calculate p typical correlation coefficients ρ1, ρ2, …, ρ p , where p is a positive integer greater than or equal to 1, for example, p is d v and d z The minimum value in . Taking j as an integer from 1 to p as an example, the calculation method of the jth typical correlation coefficient can be understood as follows:

[0183]

[0184]

[0185]

[0186] Among them, max(·) means taking the maximum value, corr(·,·) means the linear correlation coefficient, for example, given n samples of random variables a and b Then the following relationship exists:

[0187]

[0188] in, Represents a 1:n The arithmetic mean of Representative b 1:n The arithmetic mean of .

[0189] w z,j and w V,j Represents the projection vector of the jth typical correlation coefficient, V 1:n and z 1:n Projected onto the jth typical variable V 1:n w V,j and z 1:n w z,j ⊥ represents the orthogonal symbol. In the process of solving the jth canonical correlation coefficient, it is necessary to ensure that the jth canonical variable is orthogonal to 1, 2, …, j-1 canonical variables, that is:

[0190]

[0191] Furthermore, a function whose input is p typical correlation coefficients is used to measure the statistical dependence relationship dep between V and z. i (V,z), including but not limited to one or more of the following:

[0192] dep i (V,z) is the arithmetic mean, geometric mean, or harmonic mean of the p canonical correlation coefficients themselves or any non-zero power of the p canonical correlation coefficients;

[0193] dep i (V,z) is the weighted average of the p canonical correlation coefficients themselves or any non-zero power of the p canonical correlation coefficients, that is: Among them, the weighting coefficient α j is a non-negative real number and It is predefined; s is a power (s≠0), and its value includes but is not limited to 0.5, 1, 2, etc.

[0194] Example 3: Based on the distance correlation coefficient of V and z, the statistical dependence relationship between V and z is determined.

[0195] set up Represents the matrix composed of input space samples, where n represents the number of training samples, dv Represents the feature dimension of the input space sample, n,d v are all positive integers; let Represents the matrix composed of latent space samples, d z is a positive integer, d z Represents the feature dimension of the latent space sample. An input space sample and a latent space sample constitute a sample pair. Taking the calculation of the distance between the x-th sample pair and the y-th sample pair by the terminal device as an example, the following operations can be performed: Calculate the distance a between the input space sample of the x-th sample pair and the input space sample of the y-th sample pair x,y =||V x -V y ||, the distance b between the latent space sample of the x-th sample pair and the latent space sample of the y-th sample pair x,y =||z x -z y ||; where ||·|| represents distance. Some distance metrics include but are not limited to Euclidean distance, Manhattan distance, or Chebyshev distance.

[0196] Calculate the distance A between the xth sample and the yth sample x,y ,B x,y :

[0197]

[0198]

[0199] in, Represents a x,1 ,a x,2 ,…,a x,n The arithmetic mean of Represents a 1,y ,a 2,y ,…,a n,y The arithmetic mean of represent The arithmetic mean of . Similarly, Representative b x,1 ,b x,2 ,…,b x,n The arithmetic mean of Representative b 1,y ,b 2,y ,…,b n,y The arithmetic mean of represent The arithmetic mean of .

[0200] Compute the distance covariance:

[0201]

[0202] Calculate the distance variance:

[0203]

[0204]

[0205] Finally, the distance correlation coefficient is calculated:

[0206]

[0207] Furthermore, the input is the distance correlation coefficient The function to measure the statistical dependence between V and z is dep i (V,z), including but not limited to one or more of the following: dep i (V,z) is the distance correlation coefficient itself, or, dep i (V,z) is any non-zero power of the distance correlation coefficient.

[0208] Example 4, the Hilbert-Schmidt independence test (HSIC) of V and z is calculated as follows:

[0209] set up Represents the matrix composed of input space samples, where n represents the number of training samples, d v Represents the feature dimension of the input space sample, n,d v are all positive integers; let Represents the matrix composed of latent space samples, d z is a positive integer, d z represents the characteristic dimension of the latent space sample. The kernel function defined in V is The kernel function defined at z is For each input sample pair (V r ,V c ) Calculate the kernel function and get the matrix K, where the elements in the rth row and cth column satisfy K r,c = k(V r ,V c ) ; Calculate the kernel function for each latent space sample pair to obtain the matrix L, where the elements in the rth row and cth column satisfy: L r,c = l(z r ,z c ), where r,c=1,2,…,n. The kernel function may include a linear kernel function, a radial basis kernel function or other functions.

[0210] Specifically, the calculation method of HSIC is:

[0211] Among them, C refers to the center matrix, Among them I n represents the n×n unit matrix, 1 represents an n×1 vector whose elements are all 1; tr(·) represents the trace of the matrix.

[0212] The HSIC is normalized to obtain similarity indices such as centered kernel alignment (CKA):

[0213] in,

[0214] After the calculation is completed, the function with input CKA(V,z) is used to measure the statistical dependency dep between V and z i (V,z), including but not limited to one or more of the following: dep i (V,z) is CKA(V,z) itself, or, dep i (V,z) is any non-zero power of CKA(V,z).

[0215] Example 5, the normalized mutual information of V and z is defined as follows:

[0216]

[0217] Among them, MI(V,z) represents the mutual information of V and z, H(V) and H(z) represent the information entropy of V and z respectively, and the information entropy can be Shannon entropy, Rényi entropy or other entropies. MI(V,z), H(V), and H(z) need to be calculated based on the same definition of information entropy. AVG(H(V),H(z)) represents the average value of H(V) and H(z). The specific average value calculation method includes but is not limited to the arithmetic mean, geometric mean, harmonic mean or weighted mean; wherein the weighting coefficient corresponding to the weighted average can be pre-defined. After the calculation is completed, the function with input NMI(V,z) is used to measure the statistical dependency relationship dep between V and z i (V,z), including but not limited to one or more of the following: dep i (V,z) is NMI(V,z) itself, or, dep i (V,z) is any non-zero power of NMI(V,z).

[0218] (II) Online Adaptation Phase

[0219] After the autoencoder on the terminal device side and the access network device side completes offline training, the adaptation between the second encoder on the terminal device side and the first decoder on the access network device side can be achieved through information interaction.

[0220] Method 1: Taking the first decoder trained offline on the access network device side as a benchmark, the terminal device side and the access network device side update the first encoder trained offline on the terminal device side through information interaction, so that the updated first encoder matches the aforementioned first decoder, or it can also be understood that the updated first encoder and the aforementioned first decoder can constitute an autoencoder that can be applied to the reasoning link. In addition, it can be understood that the first encoder and the second encoder described below are both those that have been trained offline.

[0221] For example, Figure 5A Taking the information interaction between the first device and the second device as an example, a communication method for realizing online adaptation is provided. It can be understood that the first device is a device on the access network device side, such as an access network device or an independent AI network element associated with the access network device; the second device is a device on the terminal device side, such as a terminal device, an independent AI network element associated with the terminal device, or a server to which the terminal device belongs. The method mainly includes the following steps:

[0222] S401: The second device sends first capability information to the first device.

[0223] The first capability information is used to indicate that the second device has the capability of online model adaptation. For example, the first capability information indicates that the second encoder is to be updated.

[0224] Optionally, the second device may also include indication information for updating the input samples of the second encoder in the first capability information. For example, in the case where the training data sets used by the terminal device and the access network device in the offline training phase are partially or completely the same, the second device may include an identifier of the first channel information in the first capability information, and the first channel information is the channel information in the training data set of the terminal device and the training set of the access network device. For another example, the second device may also include the first channel information in the first capability information, and the first channel information is used to update the second encoder.

[0225] Optionally, the number of the first channel information may be one or more, and the embodiment of the present application is described below using one as an example.

[0226] S402: The first device sends first coding information to the second device.

[0227] The first coded information corresponds to the first channel information input to the first encoder applied to the first device. In the design where the first encoder has compression and quantization functions, the first device inputs the first channel information into the first encoder, and the corresponding output is the first coded information. In the design where the first encoder has compression but no quantization function, the first device inputs the first channel information into the first encoder, and the corresponding output is the first coded information, which is unquantized information. Further, the first device can quantize the first coded information and send the quantized information corresponding to the first coded information to the second device.

[0228] From the description of the above offline training phase, it can be seen that the statistical dependency relationship between the first channel information and the first coded information meets the first condition. Among them, the statistical dependency relationship between the first channel information and the first coded information is determined based on M parameters in the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or normalized mutual information, where M is a positive integer; that is, one or more parameters can be used to determine the statistical dependency relationship, and the specific parameters used can be pre-configured. The first condition is a pre-configured condition that has a mapping relationship with the M parameters.

[0229] For example, in a design in which the first encoder has compression and quantization functions, the statistical dependency between the first channel information and the first encoding information can be determined based on a distance correlation coefficient. Accordingly, the first condition can include a threshold corresponding to the distance correlation coefficient for measuring the statistical dependency.

[0230] For another example, in a design where the first encoder has compression and quantization functions, the statistical dependency between the first channel information and the first coded information may include a first statistical dependency determined based on an independence check result, and a second statistical dependency determined based on normalized mutual information. Accordingly, the first condition may include a threshold corresponding to the independence check result for measuring the first statistical dependency, and a threshold corresponding to the normalized mutual information for measuring the second statistical dependency.

[0231] S403: The second device determines a statistical dependency relationship between the first coding information and the second coding information.

[0232] Similar to S402, it can be understood that: the second coded information corresponds to the first channel information input to the second encoder applied to the second device. In the design where the first encoder has compression and quantization functions, the second device receives the first coded information; similarly, the second encoder also has compression and quantization functions, and the second device can input the first channel information into the second encoder, and the corresponding output is the second coded information. In the design where the first encoder has compression but not quantization functions, the second device receives the quantization information corresponding to the first coded information; the second device can perform dequantization processing on the quantization information corresponding to the first coded information to obtain the first coded information; similarly, the second encoder has compression but not quantization functions, and the second device can input the first channel information into the second encoder, and the corresponding output is the second coded information, which is unquantized information.

[0233] From the description of the above offline training phase, it can be seen that the statistical dependency relationship between the first channel information and the second coded information meets the first condition. Among them, the statistical dependency relationship between the first channel information and the second coded information is determined based on M parameters in linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or normalized mutual information, where M is a positive integer; that is, one or more parameters can be used to determine the statistical dependency relationship, and the specific parameters used can be pre-configured. The first condition is a pre-configured condition that has a mapping relationship with the M parameters.

[0234] For example, in a design in which the first encoder has compression and quantization functions, the statistical dependency between the first channel information and the second encoded information can be determined based on a distance correlation coefficient. Accordingly, the first condition can include a threshold corresponding to the distance correlation coefficient for measuring the statistical dependency.

[0235] For another example, in a design where the first encoder has compression and quantization functions, the statistical dependency between the first channel information and the second coded information may include a first statistical dependency determined based on an independence check result, and a second statistical dependency determined based on normalized mutual information. Accordingly, the first condition may include a threshold corresponding to the independence check result for measuring the first statistical dependency, and a threshold corresponding to the normalized mutual information for measuring the second statistical dependency.

[0236] Further, the second device can determine whether the statistical dependency relationship between the first coded information and the second coded information meets the second condition. Similarly, the statistical dependency relationship between the first coded information and the second coded information is determined based on N parameters in the linear regression similarity, the typical correlation coefficient, the distance correlation coefficient, the independence check result, or the normalized mutual information, where N is a positive integer; that is, one or more parameters can be used to determine the statistical dependency relationship, and the specific parameters used can be pre-configured. The second condition is a pre-configured condition that has a mapping relationship with the N parameters. It can be understood that M and N can be the same or different.

[0237] Among them, when M and N are the same, the aforementioned M parameters and N parameters may be partially the same, completely the same, or different. For example, M and N are both 2, the M parameters include the distance correlation coefficient and the independence verification result, and the N parameters include the linear regression similarity and the typical correlation coefficient; for example, M and N are both 2, the M parameters include the linear regression similarity and the typical correlation coefficient, and the N parameters include the linear regression similarity and the typical correlation coefficient; for another example, the M parameters include the linear regression similarity and the distance correlation coefficient, and the N parameters include the linear regression similarity and the typical correlation coefficient. When M and N are different, the aforementioned M parameters and N parameters may be partially the same or completely different. For example, M is 1, N is 2, the M parameters include the distance correlation coefficient, and the N parameters include the distance correlation coefficient and the typical correlation coefficient; for another example, M is 1, N is 2, the M parameters include the normalized mutual information, and the N parameters include the distance correlation coefficient and the typical correlation coefficient.

[0238] In a possible implementation, the second device may interact with the first device through S404a to S405a to determine whether the statistical dependency relationship between the first coded information and the second coded information meets the second condition; or the second device may determine whether the statistical dependency relationship between the first coded information and the second coded information meets the second condition by itself; or the second device may interact with the first device through S404b to S405b to determine whether the statistical dependency relationship between the first coded information and the second coded information meets the second condition. S404a to S405a and S404b to S405b are optional steps. Figure 5A Indicated by dotted line.

[0239] S404a: The second device sends the statistical dependency relationship between the first encoding information and the second encoding information to the first device.

[0240] S405a: The second device receives first indication information from the first device.

[0241] Among them, the first indication information is used to indicate that the statistical dependency between the first encoding information and the second encoding information meets the second condition, that is, the first device indicates to the second device in an explicit manner that the statistical dependency between the first encoding information and the second encoding information meets the second condition; or, the first indication information is used to trigger an update of the second encoder, that is, the first device indicates to the second device in an implicit manner that the statistical dependency between the first encoding information and the second encoding information meets the second condition.

[0242] S404b: The first device sends third indication information to the second device.

[0243] The third indication information is used to indicate a method for determining a statistical dependency relationship between the first coding information and the second coding information, and the second condition corresponding to the determination method.

[0244] S405b: The second device determines, based on the third indication information, whether the statistical dependency between the first encoding information and the second encoding information meets a second condition.

[0245] First, the second device can determine the calculation method of the reference statistical dependency based on the third indication information, and then determine the statistical dependency of the first coding information and the second coding information with reference to the specific calculation method in the aforementioned statistical dependency example; finally, the second device can determine whether the statistical dependency of the first coding information and the second coding information meets the second condition in the third indication information.

[0246] Furthermore, after the second device executes S404a to S405a, or executes S404b to S405b, or executes S403 without executing S404a to S405a and S404b to S405b, if the second device determines that the statistical dependency between the first coding information and the second coding information meets the second condition, it may continue to execute the following step S406; if it is determined that the statistical dependency between the first coding information and the second coding information does not meet the second condition, taking the compression feedback of CSI as an example, the second device may perform compression feedback of CSI based on a fixed codebook, or perform compression feedback of CSI based on a second encoder obtained by offline training.

[0247] S406: If the statistical dependency between the first encoding information and the second encoding information meets a second condition, the second device updates the second encoder according to the first encoding information.

[0248] In an optional implementation, the second device may use a preconfigured update method to update some or all of the model parameters involved in the second encoder until the second device inputs the first channel information into the second encoder, and the difference between the corresponding output third encoded information and the first encoded information is less than a set threshold. It is understandable that the third encoded information corresponds to the first channel information of the second encoder after input update. Exemplarily, the update method may be an update method based on gradient descent, such as a traditional gradient update method, a gradient update method based on transfer learning, or a gradient update method based on meta-learning.

[0249] In another possible implementation, the second device can introduce a specific transformation function based on the second encoder to implement the update of the second encoder, such as adding a linear transformation function in the encoder, or separately connecting a linear transformation module after the encoder. The linear change function / linear transformation module can be used to transform the aforementioned second coded information to obtain third coded information, the difference between the third coded information and the first coded information is less than a set threshold, and the third coded information corresponds to the first channel information of the second encoder after input update. Optionally, the linear change function / linear transformation module can be implemented by least squares method, Pugh analysis method or other algorithms. Such a design does not need to change the original model parameters of the second encoder, and the update is relatively simple.

[0250] In summary, the second encoder can be updated through the above steps, thereby completing the online adaptation between the second encoder of the second device and the first decoder of the first device.

[0251] Based on this, the embodiment of the present application takes the compression feedback of CSI as an example, and describes in detail the reasoning process of the updated second encoder and the first decoder after online adaptation through the following steps S407 to S410.

[0252] S407: The second device obtains second channel information.

[0253] For example, when the second device is a terminal device and the first device is an access network device, the second device can receive a downlink reference signal (such as CSI-RS) from the first device, and measure and estimate the second channel information based on the downlink reference signal.

[0254] S408: The second device encodes and quantizes the second channel information according to the updated second encoder and the first quantization method to obtain third encoded information.

[0255] The first quantization method may be one of a plurality of preconfigured quantization methods, or the first quantization method is indicated by the first device to the second device. For example, in a design where the second encoder has compression and quantization functions, the first quantization method is used to determine the length of the third coded information output by the updated second encoder. For another example, in a design where the second encoder has a compression function but does not have a quantization function, the first quantization method is used to determine the length of the quantization information corresponding to the third coded information output by the updated second encoder.

[0256] Through such a design, the quantization bit width (i.e., the length of the quantization information) can be adjusted to achieve different feedback overheads. During implementation, only the model under a certain bit width can be adapted online to complete model adaptation with different feedback overheads, further reducing the model adaptation overhead. Model adaptation here refers to the matching between the encoder and the decoder.

[0257] S409: The second device sends the third coding information to the first device.

[0258] For example, in a design where the second encoder has compression and quantization functions, the second device sends the third coded information to the first device. In another example, in a design where the second encoder has compression function but not quantization function, the second device sends the quantization information corresponding to the third coded information to the first device.

[0259] S410: The first device restores second channel information according to third coding information and a first decoder.

[0260] For example, in a design where the second encoder has compression and quantization functions, the first device receives the third coded information. If the first decoder has dequantization and decompression functions, the first device can input the third coded information into the first decoder, and the corresponding output is the restored second channel information. For another example, in a design where the second encoder has compression function but not quantization function, the first device receives the quantization information corresponding to the third coded information. The first device can first perform dequantization processing on the quantization information corresponding to the third coded information to restore the third coded information; then the first device can input the third coded information into the first decoder, and the corresponding output is the restored second channel information.

[0261] The above-mentioned scheme provided in the embodiment of the present application realizes the training and updating of the encoder on the terminal device side through offline training and online adaptation, which can complete the matching of the encoding on the terminal device side and the decoding on the access network device side while reducing the communication overhead, which is beneficial to improving the performance of the autoencoder used in the communication network.

[0262] Method 2: Taking the second encoder trained offline on the terminal device side as a reference, the terminal device side and the access network device side update the first decoder trained offline on the access network device side through information interaction, so that the second encoder matches the updated first decoder, or it can also be understood that the second encoder and the updated first decoder can constitute an autoencoder that can be applied to the reasoning link. In addition, it can be understood that the first encoder, second encoder, and first decoder described below all refer to those after offline training.

[0263] For example, Figure 5B Taking the information interaction between the first device and the second device as an example, a communication method for realizing online adaptation is provided. It can be understood that the first device is a device on the access network device side, such as an access network device or an independent AI network element associated with the access network device; the second device is a device on the terminal device side, such as a terminal device, an independent AI network element associated with the terminal device, or a server to which the terminal device belongs. The method mainly includes the following steps:

[0264] S501: A first device sends first request information to a second device.

[0265] The first request information is used to request the output of the second encoder of the second device. Optionally, the first device may also include indication information of the input to the second encoder in the first request information. For example, in the case where the training data sets used by the terminal device and the access network device in the offline training phase are partially or completely the same, the first device may include the identifier of the first channel information in the first request information. The first channel information is the channel information in the training data set of the terminal device and the training set of the access network device. The second device may determine the first channel information based on the received first request information, and then input the first channel information into the second encoder to obtain the corresponding encoding information.

[0266] Optionally, the number of the first channel information may be one or more, and the embodiment of the present application is described below using one as an example.

[0267] S502: The second device sends second coded information to the first device.

[0268] The second coded information corresponds to the first channel information input to the second encoder applied to the second device. In the design where the second encoder has compression and quantization functions, the second device inputs the first channel information into the second encoder, and the corresponding output is the second coded information. In the design where the second encoder has compression but no quantization function, the second device inputs the first channel information into the second encoder, and the corresponding output is the second coded information, which is unquantized information. Further, the second device can quantize the second coded information and send the quantized information corresponding to the second coded information to the first device.

[0269] From the description of the above offline training phase, it can be seen that the statistical dependency relationship between the first channel information and the second coded information meets the first condition. Among them, the statistical dependency relationship between the first channel information and the second coded information is determined based on M parameters in linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or normalized mutual information, where M is a positive integer; that is, one or more parameters can be used to determine the statistical dependency relationship, and the specific parameters used can be pre-configured. The first condition is a pre-configured condition that has a mapping relationship with the M parameters.

[0270] For example, in a design in which the second encoder has compression and quantization functions, the statistical dependency between the first channel information and the second encoded information can be determined based on a distance correlation coefficient. Accordingly, the first condition can include a threshold corresponding to the distance correlation coefficient for measuring the statistical dependency.

[0271] For another example, in a design where the second encoder has compression and quantization functions, the statistical dependency between the first channel information and the second coded information may include a first statistical dependency determined based on an independence check result, and a second statistical dependency determined based on normalized mutual information. Accordingly, the first condition may include a threshold corresponding to the independence check result for measuring the first statistical dependency, and a threshold corresponding to the normalized mutual information for measuring the second statistical dependency.

[0272] S503: The first device determines a statistical dependency relationship between the first encoding information and the second encoding information.

[0273] Similar to S502, it can be understood that: the first coded information corresponds to the first channel information input to the first encoder applied to the first device. In the design where the second encoder has compression and quantization functions, the first device receives the second coded information; similarly, the first encoder also has compression and quantization functions, and the first device can input the first channel information into the first encoder, and the corresponding output is the first coded information. In the design where the second encoder has compression but not quantization functions, the first device receives the quantization information corresponding to the second coded information; the first device can perform dequantization processing on the quantization information corresponding to the second coded information to obtain the second coded information; similarly, the first encoder has compression but not quantization functions, and the first device can input the first channel information into the first encoder, and the corresponding output is the first coded information, which is unquantized information.

[0274] From the description of the above offline training phase, it can be seen that the statistical dependency relationship between the first channel information and the second coded information meets the first condition. Among them, the statistical dependency relationship between the first channel information and the second coded information is determined based on M parameters in linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or normalized mutual information, where M is a positive integer; that is, one or more parameters can be used to determine the statistical dependency relationship, and the specific parameters used can be pre-configured. The first condition is a pre-configured condition that has a mapping relationship with the M parameters.

[0275] For example, in a design in which the first encoder has compression and quantization functions, the statistical dependency between the first channel information and the second encoded information can be determined based on a distance correlation coefficient. Accordingly, the first condition can include a threshold corresponding to the distance correlation coefficient for measuring the statistical dependency.

[0276] For another example, in a design where the first encoder has compression and quantization functions, the statistical dependency between the first channel information and the second coded information may include a first statistical dependency determined based on an independence check result, and a second statistical dependency determined based on normalized mutual information. Accordingly, the first condition may include a threshold corresponding to the independence check result for measuring the first statistical dependency, and a threshold corresponding to the normalized mutual information for measuring the second statistical dependency.

[0277] Further, the first device can determine whether the statistical dependency relationship between the first coded information and the second coded information meets the second condition. Similarly, the statistical dependency relationship between the first coded information and the second coded information is determined based on N parameters in the linear regression similarity, the typical correlation coefficient, the distance correlation coefficient, the independence check result, or the normalized mutual information, where N is a positive integer; that is, one or more parameters can be used to determine the statistical dependency relationship, and the specific parameters used can be pre-configured. The second condition is a pre-configured condition that has a mapping relationship with the N parameters. It can be understood that M and N can be the same or different.

[0278] Among them, when M and N are the same, the aforementioned M parameters and N parameters may be partially the same, completely the same, or different. For example, M and N are both 2, the M parameters include the distance correlation coefficient and the independence verification result, and the N parameters include the linear regression similarity and the typical correlation coefficient; for example, M and N are both 2, the M parameters include the linear regression similarity and the typical correlation coefficient, and the N parameters include the linear regression similarity and the typical correlation coefficient; for another example, the M parameters include the linear regression similarity and the distance correlation coefficient, and the N parameters include the linear regression similarity and the typical correlation coefficient. When M and N are different, the aforementioned M parameters and N parameters may be partially the same or completely different. For example, M is 1, N is 2, the M parameters include the distance correlation coefficient, and the N parameters include the distance correlation coefficient and the typical correlation coefficient; for another example, M is 1, N is 2, the M parameters include the normalized mutual information, and the N parameters include the distance correlation coefficient and the typical correlation coefficient.

[0279] Further, if the second device determines that the statistical dependency between the first coding information and the second coding information meets the second condition, the following step S504 may continue to be executed; if it is determined that the statistical dependency between the first coding information and the second coding information does not meet the second condition, taking the compression feedback of CSI as an example, the first device may instruct the second device to perform compression feedback of CSI based on a fixed codebook, or to perform compression feedback of CSI based on a second encoder obtained by offline training.

[0280] S504: If the statistical dependency between the first coding information and the second coding information meets a second condition, the first device updates the first decoder according to the second coding information.

[0281] In an optional implementation, the first device may use a preconfigured update method to update some or all of the model parameters involved in the first decoder until the first device inputs the second encoded information into the first decoder, and the difference between the corresponding output third channel information and the first channel information is less than a set threshold. It is understandable that the third channel information corresponds to the second encoded information input to the updated first decoder. Exemplarily, the update method can be an update method based on gradient descent, such as a traditional gradient update method, a gradient update method based on transfer learning, or a gradient update method based on meta-learning.

[0282] In another possible implementation, the first device can introduce a specific transformation function on the basis of the first decoder to implement the update of the first encoder, such as adding a linear transformation function in the first decoder, or separately connecting a linear transformation module before the first decoder. The linear change function / linear transformation module can be used to transform the aforementioned second coded information and input it into the updated first decoder, and correspondingly output the third channel information, the difference between the third channel information and the first channel information is less than the set threshold, and the third channel information corresponds to the second coded information input into the updated first decoder. Optionally, the linear change function / linear transformation module can be implemented by the least squares method, the Pugh analysis method or other algorithms. Such a design does not require changing the original model parameters of the first decoder, and the update is relatively simple.

[0283] In summary, the first decoder can be updated through the above steps, thereby completing the online adaptation between the second encoder of the second device and the first decoder of the first device.

[0284] Based on this, the embodiment of the present application takes the compression feedback of CSI as an example, and describes in detail the reasoning process of the second encoder and the updated first decoder after online adaptation through the following steps S505 to S508.

[0285] S505: The second device obtains fourth channel information.

[0286] For example, when the second device is a terminal device and the first device is an access network device, the second device can receive a downlink reference signal (such as CSI-RS) from the first device, and measure and estimate the fourth channel information based on the downlink reference signal.

[0287] S506: The second device encodes and quantizes the fourth channel information according to the second encoder and the first quantization method to obtain fourth encoded information.

[0288] The first quantization method may be one of a plurality of preconfigured quantization methods, or the first quantization method is indicated by the first device to the second device. For example, in a design where the second encoder has compression and quantization functions, the first quantization method is used to determine the length of the fourth coded information output by the updated second encoder. For another example, in a design where the second encoder has a compression function but does not have a quantization function, the first quantization method is used to determine the length of the quantization information corresponding to the fourth coded information output by the updated second encoder.

[0289] S507: The second device sends the fourth coding information to the first device.

[0290] For example, in a design where the second encoder has compression and quantization functions, the second device sends the fourth coded information to the first device. In another example, in a design where the second encoder has compression function but not quantization function, the second device sends the quantization information corresponding to the fourth coded information to the first device.

[0291] S508: The first device restores fourth channel information according to the fourth coding information and the updated first decoder.

[0292] For example, in a design where the second encoder has compression and quantization functions, the first device receives the fourth coded information. If the updated first decoder has dequantization and decompression functions, the first device can input the fourth coded information into the first decoder, and the corresponding output is the restored fourth channel information. For another example, in a design where the second encoder has compression function but not quantization function, the first device receives the quantization information corresponding to the fourth coded information. The first device can first perform dequantization processing on the quantization information corresponding to the fourth coded information to restore the fourth coded information; then the first device can input the fourth coded information into the updated first decoder, and the corresponding output is the restored fourth channel information.

[0293] The above-mentioned scheme provided in the embodiment of the present application realizes the training and updating of the decoder on the access network device side through offline training and online adaptation, which can complete the matching of the encoding on the terminal device side and the decoding on the access network device side while reducing the communication overhead, which is beneficial to improving the performance of the autoencoder used in the communication network.

[0294] In addition, the embodiment of the present application is as follows, targeting the scenario of CSI compression feedback in a MIMO-OFDM system in frequency division duplex mode, and exemplifies the offline training and online adaptation process of the autoencoder used in the scenario.

[0295] 1. Offline training phase

[0296] The terminal device and the access network device can use the same or different channel information in the public data set to independently train the dual-end model (i.e., autoencoder) of CSI compression. Among them, the public data set can be channel data based on the UMi model. Optionally, the terminal device and the access network device can use the same or different model structures when training the dual-end model. Some examples of model structures include Transformer, CNN, MLP-Mixer, etc. Optionally, the terminal device and the access network device can use the same or different loss functions when training the dual-end model. Some examples of loss functions include cosine similarity loss, mean square error loss, etc. In addition, in the implementation involving quantization processing, the terminal device and the access network device can use the same or different quantization bit widths.

[0297] Taking the method of determining the statistical dependency described in the above example 3 as an example, the terminal device and the access network device can add the constraint corresponding to the distance correlation coefficient between the input spatial samples and the output latent space samples of the encoders trained by each of them as a regular term to the loss function. Exemplarily, the constraint includes: the distance correlation coefficient between the spatial samples and the output latent space samples is greater than or equal to 0.9.

[0298] The embodiment of the present application is described below by taking the example that the encoder obtained by the terminal device and the access network device through offline training has a compression function but does not have a quantization function, and the decoder has a decompression function but does not have a dequantization function. Among them, the first encoder obtained by the access network device through offline training is denoted as g enc and the first decoder is denoted as g dec The second encoder obtained by the terminal device through offline training is denoted as f enc , the second decoder is denoted as f dec .

[0299] (II) Online Adaptation Phase

[0300] Assume that in a MIMO-OFDM system in frequency division duplex mode, the access network equipment deploys 16 ports, the terminal equipment deploys 4 ports, the number of resource blocks is 52, each resource block has 12 subcarriers, the subcarrier spacing is 15kHz, and the center frequency is 2GHz. The feedback object of the terminal device is the characteristic vector of the subband channel. Each subband has 4 resource blocks, and the highest rank is 4. The original dimension of the channel characteristic vector is The terminal device can stack the real and imaginary parts of the characteristic vector of a sub-band channel into a real vector, and then use the real vectors corresponding to one or more sub-bands as input space samples V.

[0301] Taking the aforementioned public dataset including n input space samples as an example, the public dataset can be represented as a set Based on this, the terminal device / network device side can determine the corresponding latent space samples based on the same input space samples in the public data set. The calculation process can be expressed as: UE =f enc (V comm ),z BS =g enc (V comm ). Among them, z UE Corresponding to the terminal device input to f enc Public data V comm , z BS Corresponding to the access network equipment input to g enc Public data V comm .

[0302] Taking the above solution 1 to update the second encoder of the terminal device as an example, the access network device can BS In a possible implementation, the access network device may first use the quantization bit width indicated in the preconfigured quantization codebook to BS Quantify it and get z BS The corresponding quantitative information e BS , e BS =q(z BS , bits). The quantization codebook includes a quantization mode q(·), and a quantization bit width (bits). Based on this, the access network device can convert the quantization mode q(·), the quantization bit width (bits) and the e BS Sent to the terminal device. Assume that the network side sends e BS The number of e is 100, each e BS The dimension is 128, the quantization bit width is 2 bits, and the total transmission overhead is 3.2 kilobytes. The maximum quantization bit width is set to 4, and it is transmitted as a 32-bit floating point number. The total transmission overhead of the quantization codebook is 64 bytes.

[0303] Furthermore, the terminal device receives e BS , you can first BS After dequantization, the latent space vector Among them, q -1 (·) represents the dequantization function.

[0304] Terminal computing and z UE Statistical dependencies, such as those based on and z UE The distance correlation coefficient is determined by Then, the terminal device uses 0.8 as the threshold to judge Does this equation hold?

[0305] If it is established, the terminal device can perform online adaptation, that is, the terminal device can To update the second encoder f obtained by offline training enc For example, the terminal device uses the least squares method to calculate the linear transformation matrix to align and z UE , that is, calculation Among them, λ is the regularization factor, for example, the value of λ is 0.1, and I is the unit matrix. Finally, the terminal device can enc Based on the introduction of W LR , that is, the updated second encoder can be expressed as W LR *f enc .

[0306] Furthermore, the updated second encoder on the terminal device side is matched with the first decoder on the access network device side to form a dual-end model (i.e., autoencoder) applied to the inference link, i.e., the actual CSI compression feedback.

[0307] Exemplarily, the terminal device may obtain new channel information, and according to the above-mentioned preprocessing design, preprocess the channel information to be fed back to obtain V new , and then V new Input to the updated second encoder to obtain the corresponding latent space z new =W LR *f enc (V new ). The terminal device can use the quantization bit width indicated by the access network device to new Quantify and obtain e new =q(z new , bits), and e new The quantization bit width indicated by the access network device is less than or equal to the maximum quantization bit width supported by the access network device for training the dual-end model, for example, the quantization bit width indicated by the access network device is 2.

[0308] The access network equipment responds to the received channel feedback according to the same quantization bit width. new Perform dequantization operation, and then restore the first decoder model

[0309] Based on the same idea, see Figure 6The embodiment of the present application provides a communication device 600, which includes a processing module 601 and a communication module 602. The communication device 600 can be a terminal device, or a communication device applied to a terminal device or used in combination with a terminal device, and capable of implementing a random access method executed by the terminal device side; or, the communication device 600 can be a network device, or a communication device applied to a network device or used in combination with a network device, and capable of implementing a random access method executed by the network device.

[0310] The communication module may also be referred to as a transceiver module, a transceiver, a transceiver, or a transceiver device, etc. The processing module may also be referred to as a processor, a processing board, a processing unit, or a processing device, etc. Optionally, the communication module is used to perform the sending operation and the receiving operation on the terminal device side or the network device side in the above method, and the device used to implement the receiving function in the communication module may be regarded as a receiving unit, and the device used to implement the sending function in the communication module may be regarded as a sending unit, that is, the communication module includes a receiving unit and a sending unit.

[0311] When the communication device 600 is applied to a terminal device, the processing module 601 can be used to implement Figure 5A or Figure 5B The processing function of the terminal device in the example described above, the communication module 602 can be used to implement Figure 5A or Figure 5B The transceiver function of the terminal device in the example. Optionally, the communication device can also be understood by referring to the third aspect in the content of the invention and the possible designs in the third aspect.

[0312] When the communication device 600 is applied to a network device, the processing module 601 can be used to implement Figure 5A or Figure 5B The processing function of the network device in the example described above, the communication module 602 can be used to implement Figure 5A or Figure 5B The transceiver function of the network device in the example. Optionally, the communication device can also be understood by referring to the fourth aspect of the invention and the possible designs in the fourth aspect.

[0313] In addition, it should be noted that, in one possible design, the aforementioned communication module and / or processing module may be implemented through a virtual module, for example, the processing module may be implemented through a software function unit or a virtual device, and the communication module may be implemented through a software function or a virtual device. In another possible design, the processing module or the communication module may also be implemented through a physical device, for example, if the device is implemented using a chip / chip circuit, the communication module may be an input / output circuit and / or a communication interface, performing input operations (corresponding to the aforementioned receiving operations) and output operations (corresponding to the aforementioned sending operations); the processing module is an integrated processor or microprocessor or integrated circuit.

[0314] The division of modules in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each example of the embodiments of the present application may be integrated into a processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0315] Based on the same technical concept, the embodiment of the present application also provides a communication device 700. For example, the communication device 700 can be a chip or a chip system. Optionally, in the embodiment of the present application, the chip system can be composed of chips, or can include chips and other discrete devices.

[0316] The communication device 700 can be used to implement the functions of any network element in the communication system described in the above examples. The communication device 700 may include at least one processor 710. Optionally, the processor 710 is coupled to a memory, and the memory may be located within the device; or, the memory may be integrated with the processor; or, the memory may be located outside the device. For example, the communication device 700 may also include at least one memory 720. The memory 720 stores the necessary computer programs, computer programs or instructions and / or data for implementing any of the above examples; the processor 710 may execute the computer program stored in the memory 720 to complete the method in any of the above examples.

[0317] The communication device 700 may also include a communication interface 730, and the communication device 700 may exchange information with other devices through the communication interface 730. Exemplarily, the communication interface 730 may be a transceiver, a circuit, a bus, a module, a pin, or other types of communication interfaces. When the communication device 700 is a chip-type device or circuit, the communication interface 730 in the device 700 may also be an input-output circuit, which may input information (or receive information) and output information (or send information), and the processor may be an integrated processor or a microprocessor or an integrated circuit or a logic circuit, and the processor may determine the output information based on the input information.

[0318] The coupling in the embodiment of the present application is an indirect coupling or communication connection between devices, units or modules, which can be electrical, mechanical or other forms, and is used for information exchange between devices, units or modules. The processor 710 may cooperate with the memory 720 and the communication interface 730. The specific connection medium between the above-mentioned processor 710, memory 720 and communication interface 730 is not limited in the embodiment of the present application.

[0319] Optional, see Figure 7The processor 710, the memory 720 and the communication interface 730 are connected to each other via a bus 740. The bus 740 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0320] In the embodiments of the present application, the processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, and may implement or execute the methods, steps, and logic block diagrams of the embodiments of the present application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method of the embodiment applied in conjunction with the embodiment of the present application may be directly embodied as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0321] In the embodiment of the present application, the memory may be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), etc., or a volatile memory (volatile memory), such as a random access memory (RAM). The memory is any other medium that can be used to carry or store a desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory in the embodiment of the present application may also be a circuit or any other device that can implement a storage function, for storing program instructions and / or data.

[0322] In a possible implementation, the communication device 700 can be applied to a terminal device. Specifically, the communication device 700 can be a terminal device, or a device that can support a terminal device and implement the functions of the terminal device in any of the above-mentioned examples. The memory 720 stores a computer program (or instruction) and / or data that implements the functions of the terminal device in any of the above-mentioned examples. The processor 710 can execute the computer program stored in the memory 720 to complete the method executed by the terminal device in any of the above-mentioned examples. If the communication device is applied to a terminal device, the communication interface in the communication device 700 can be used to interact with a network device, send information to the network device, or receive information from the network device.

[0323] In another possible implementation, the communication device 700 can be applied to a network device. Specifically, the communication device 700 can be a network device, or a device that can support a network device and implement the functions of the network device in any of the above-mentioned examples. The memory 720 stores a computer program (or instruction) and / or data that implements the functions of the network device in any of the above-mentioned examples. The processor 710 can execute the computer program stored in the memory 720 to complete the method executed by the network device in any of the above-mentioned examples. If the communication device is applied to a network device, the communication interface in the communication device 700 can be used to interact with a terminal device, send information to the terminal device, or receive information from the terminal device.

[0324] Since the communication device 700 provided in this example can be applied to a network device to complete the method executed by the cell on the network device side, or applied to a terminal device to complete the method executed by the terminal device, the technical effects that can be obtained can refer to the above method examples and will not be repeated here.

[0325] The technical solution provided in the embodiment of the present application can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a terminal device, a network device or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital video disc (DVD)), or a semiconductor medium, etc.

[0326] In the embodiments of the present application, under the premise of no logical contradiction, the examples may reference each other, for example, the methods and / or terms between method embodiments may reference each other, for example, the functions and / or terms between device embodiments may reference each other, for example, the functions and / or terms between device examples and method examples may reference each other.

[0327] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the embodiments of the present application and their equivalents, the embodiments of the present application are also intended to include these modifications and variations.

Claims

1. A communication method, characterized in that: include: Acquire first encoding information; wherein the first encoding information corresponds to first channel information input to a first encoder applied to a first device, and a statistical dependency relationship between the first channel information and the first encoding information meets a first condition; Determining a statistical dependency relationship between the first encoding information and the second encoding information; wherein the second encoding information corresponds to the first channel information input to a second encoder applied to a second device, and the statistical dependency relationship between the first channel information and the second encoding information meets the first condition; When the statistical dependency between the first encoding information and the second encoding information meets a second condition, the second encoder is updated according to the first encoding information.

2. The method according to claim 1, characterized in that The difference between the first encoding information and the third encoding information is smaller than a set threshold, and the third encoding information corresponds to the first channel information input into the updated second encoder.

3. A communication method, characterized in that: include: Sending first coded information; wherein the first coded information corresponds to first channel information input to a first encoder applied to a first device, and a statistical dependency between the first channel information and the first coded information meets a first condition; the first coded information is used to update a second encoder, and the second encoder is applied to a second device; wherein a statistical dependency between an input and an output of the second encoder before the update meets the first condition.

4. The method according to any one of claims 1 to 3, characterized in that: The first encoder or the second encoder before updating is obtained through offline training.

5. A communication method, characterized in that: include: Acquire second encoding information; wherein the second encoding information corresponds to first channel information input to a second encoder applied to a second device, and a statistical dependency relationship between the first channel information and the second encoding information meets a first condition; Determining a statistical dependency relationship between first encoding information and the second encoding information; wherein the first encoding information corresponds to the first channel information input to a first encoder applied to a first device, and the statistical dependency relationship between the first channel information and the first encoding information meets the first condition; When the statistical dependency between the first coding information and the second coding information meets the second condition, the first decoder applied to the first device is updated according to the second coding information, and the first decoder before the update matches the first encoder.

6. The method according to claim 5, characterized in that The difference between the first channel information and the third channel information is less than a set threshold; wherein the third channel information is the output of the updated first decoder, and the third channel information corresponds to the second encoded information input to the updated first decoder.

7. A communication method, characterized in that: include: Sending second coded information; wherein the second coded information corresponds to first channel information input to a second encoder applied to a second device, and a statistical dependency relationship between the first channel information and the second coded information meets a first condition; The second encoding information is used to update the first decoder in the first device, the statistical dependency between the input and output of the first encoder in the first device meets the first condition, and the first decoder and the first encoder before updating match.

8. The method according to any one of claims 5 to 7, characterized in that: The first encoder, the first decoder before updating, or the second encoder is obtained through offline training.

9. The method according to claim 1 or 5, characterized in that: The statistical dependency relationship between the two information is determined based on the linear regression similarity, the typical correlation coefficient, the distance correlation coefficient, the independence check result, or the M parameter in the normalized mutual information of the two information, where M is a positive integer; The two pieces of information include the first channel information and the first coding information, or the first channel information and the second coding information, and the first condition is a preconfigured condition having a mapping relationship with the M parameters.

10. The method according to claim 1 or 5, characterized in that: The statistical dependency between the first coding information and the second coding information is determined based on the linear regression similarity, typical correlation coefficient, distance correlation coefficient, independence check result, or N parameters in the normalized mutual information of the first coding information and the second coding information, where N is a positive integer, and there is a mapping relationship between the second condition and the N parameters.

11. A communication device, characterized in that: The method comprises a processor and a communication circuit, wherein the processor is used to execute computer program instructions to perform the method according to any one of claims 1 to 10.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the computer is enabled to execute the method according to any one of claims 1 to 10.

13. A computer program product, characterized in that The method comprises instructions, which, when executed on a computer, cause the computer to execute the method according to any one of claims 1 to 10.

Images