WO2025098261A1 - Communication method and communication device - Google Patents

Abstract

The present application provides a communication method and a communication device. The communication method comprises: a first communication device acquires first neural network blocks for channel information feedback, wherein neural networks comprising different numbers of first neural network blocks have different functions; and the first communication device sends to a second communication device parameters respectively corresponding to N first neural network blocks, the N first neural network blocks being comprised in a first neural network, and the first neural network being used for coding processing of channel information. Therefore, in a subsequent channel information feedback process, the second communication device can perform coding processing on channel information on the basis of the first neural network, thereby achieving AI-based channel information feedback. In addition, in the technical solution, neural networks having different functions can be obtained by means of neural network blocks of the same structure, and no training for different functions is required to obtain different neural networks, thereby reducing the management overhead and the storage overhead of neural networks.

Description

Communication method and communication device

This application claims the priority of the Chinese patent application filed with the China Patent Office on November 10, 2023, with application number 202311503718.8 and application name “Communication Method and Communication Device”, all contents of which are incorporated by reference in this application.

Technical Field

The present application relates to the field of communication technology, and more specifically, to a communication method and a communication device.

Background Art

Wireless communications are developing rapidly. The sixth-generation wireless (local area) network (wireless fidelity, Wi-Fi) (such as the Institute of Electrical and Electronics Engineers (IEEE) 802.11ax) standard has been commercialized, and the next-generation wireless technology and standardization are also in full swing around the world. Wireless communications have penetrated into all aspects of daily life and work and have become an indispensable part. With the rapid growth of the number of smart terminals and the popularization of Internet of Things (IoT) devices. New wireless technologies, new terminals, and new applications have made wireless networks unprecedentedly complex. It is foreseeable that wireless networks will become more and more complex in the future. In order to combat the high complexity development trend of wireless networks, artificial intelligence (AI) has become an industry consensus as an effective tool for wireless network design and management.

In addition, in order to improve the communication rate, it is necessary to increase the bandwidth and the number of antennas, and adopt beamforming technology. Beamforming technology requires the receiving end to feedback channel information (such as channel state information (CSI)). The current standard uses a channel information feedback method based on Givens rotation. With the increase in bandwidth and the number of antennas, the channel state information feedback overhead based on Givens rotation is greatly increased. With the development of AI technology, how to perform channel information feedback based on AI has become an urgent problem to be solved.

Summary of the invention

The present application provides a communication method for performing channel information feedback based on AI.

In a first aspect, a communication method is provided, which can be executed by a first communication device, or can also be executed by a component (such as a chip or circuit) of the first communication device, without limitation. For ease of description, the following description is based on the first communication device as an example.

The communication method includes: a first communication device obtains a first neural network block, the first neural network block is used for channel information feedback, and neural networks containing different numbers of the first neural network blocks have different functions; the first communication device sends parameters corresponding to N first neural network blocks to a second communication device, the N first neural network blocks are included in a first neural network, and the first neural network is used for encoding processing of channel information, wherein N is a positive integer.

Based on the above technical solution, the first communication device can obtain the first neural network block, and when different neural networks contain different numbers of first neural network blocks, different neural networks implement different functions. For example, the more first neural network blocks a neural network contains, the higher the computational complexity of the neural network, and the compression ratio of the channel information compressed and processed by the neural network is high, the feedback overhead is reduced, and the system throughput is high. Furthermore, the first communication device can provide the second communication device with parameters corresponding to the N first neural network blocks, so that the second communication device can determine the first neural network based on the parameters corresponding to the N received first neural network blocks, and in the subsequent channel information feedback process, the channel information can be encoded and processed based on the first neural network to achieve AI-based channel information feedback.

Moreover, in this technical solution, neural networks with different functions can be obtained through a neural network block of a structure, without the need to train different neural networks for different functions, thereby reducing management overhead and storage overhead of the neural network.

In combination with the first aspect, in certain implementations of the first aspect, before the first communication device sends parameters corresponding to N first neural network blocks to the second communication device, the method also includes: the first communication device receives first indication information from the second communication device, and the first indication information is used to indicate the N.

Based on the above technical solution, the second communication device can notify the first communication device of the number N of the required first neural network blocks through the first indication information, so that the first communication device can send the corresponding number of first neural network blocks according to the needs of the second communication device. The corresponding parameters are used to avoid the situation where the parameters corresponding to the first neural network block provided by the first communication device do not meet the requirements of the second communication device.

In combination with the first aspect, in certain implementations of the first aspect, the method also includes: the first communication device receives second indication information from the second communication device, the second indication information is used to indicate M, and M is a positive integer; the first communication device sends parameters corresponding to M first neural network blocks to the second communication device, wherein the M first neural network blocks and the N first neural network blocks are included in the second neural network.

Based on the above technical solution, when the number of first neural network blocks included in the neural network required by the second communication device changes, the second communication device can request the parameters of the newly added first neural network blocks from the first communication device, without requesting the first communication device to provide the parameters corresponding to all the first neural network blocks included in the updated neural network. For example, if the updated neural network includes Q first neural network blocks, which is M more than the first neural network, then it is sufficient to request the first communication device to provide the parameters corresponding to the M first neural network blocks, without requesting the first communication device to provide the parameters corresponding to the Q first neural network blocks, thereby effectively reducing the transmission overhead of the neural network update.

In combination with the first aspect, in certain implementations of the first aspect, the method also includes: the first communication device determines the maximum number P of first neural network blocks that the neural network can include, P is a positive integer greater than or equal to N; the first communication device sends the parameters corresponding to the N first neural network blocks to the second communication device, including: the first communication device sends the parameters corresponding to the P first neural network blocks to the second communication device.

Based on the above technical solution, the first communication device can broadcast the parameters corresponding to multiple first neural network blocks when the number of first neural network blocks is the largest to the second communication device according to the maximum number of first neural network blocks that the neural network can contain. The first communication device actively sends the parameters corresponding to multiple first neural network blocks by broadcasting, without the need for the second communication device to initiate a request process, thereby reducing the signaling overhead of the second communication device.

In combination with the first aspect, in certain implementations of the first aspect, the method also includes: the first communication device receives first information from the second communication device, the first information includes third indication information and a first vector, the third indication information is used to indicate N, and the first vector is a result of encoding the channel information through the first neural network.

Based on the above technical solution, in the process of feeding back channel information, the second communication device, in addition to feeding back the first vector obtained based on neural network processing of the channel information, also carries third indication information in the feedback information to indicate which structure of the neural network processing is used to obtain the currently fed back first vector, thereby helping the first communication device to select a suitable neural network to parse the first vector and obtain channel information with high accuracy.

In combination with the first aspect, in certain implementations of the first aspect, the parameters of the first neural network block include at least one of the following: weight information, bias information, or activation function information corresponding to the first neural network block.

In combination with the first aspect, in certain implementations of the first aspect, the first neural network block supports at least one of the following neural network structures: a convolutional neural network CNN, a multi-layer perceptron MLP, or a transformer Transformer.

In combination with the first aspect, in certain implementations of the first aspect, when the first neural network includes multiple first neural network blocks, the connection method between the multiple first neural network blocks includes deep connection and/or wide connection.

In combination with the first aspect, in some implementations of the first aspect, the method further includes: the first communication device sending parameters of an output layer of the first neural network to the second communication device.

In combination with the first aspect, in some implementations of the first aspect, the method also includes: the first communication device sends at least one of the following information to the second communication device: information indicating a quantization method, information indicating a number of quantization bits, and information indicating a feature embedding method.

In a second aspect, a communication method is provided, which can be executed by a second communication device, or can also be executed by a component (such as a chip or circuit) of the second communication device, without limitation. For ease of description, the second communication device is used as an example for description below.

The communication method includes: the second communication device obtains parameters corresponding to N first neural network blocks respectively, the first neural network blocks are used for channel information feedback, and the functions of neural networks containing different numbers of the first neural network blocks are different; the second communication device determines the first neural network based on the N first neural network blocks, and the first neural network is used for the second communication device to encode the channel information, wherein N is a positive integer.

In combination with the second aspect, in certain implementations of the second aspect, the second communication device obtains parameters corresponding to the N first neural network blocks respectively, including: the second communication device receives the parameters corresponding to the N first neural network blocks respectively from the first communication device.

In combination with the second aspect, in some implementations of the second aspect, when the second communication device receives the Before the parameters corresponding to the N first neural network blocks respectively, the method also includes: the second communication device determines the number N of first neural network blocks included in the required first neural network based on first information, the first information includes the capabilities of the second communication device and/or the need to process the channel information; the second communication device sends first indication information to the first communication device, and the first indication information is used to indicate the N.

In combination with the second aspect, in certain implementations of the second aspect, after the second communication device obtains the parameters corresponding to N of the first neural network blocks, the method also includes: the second communication device determines, based on second information, that the required second neural network includes Q of the first neural network blocks, where Q is a positive integer greater than N, and the difference between Q and N is M, and the second information includes the capabilities of the second communication device and/or the need to process the channel information; the second communication device sends second indication information to the first communication device, and the second indication information is used to indicate M; the second communication device receives the parameters corresponding to the M of the first neural network blocks from the first communication device.

In combination with the second aspect, in certain implementations of the second aspect, the method also includes: the second communication device receives parameters corresponding to P first neural network blocks from the first communication device, where P is a positive integer greater than or equal to N; the second communication device determines the N first neural network blocks from the P first neural network blocks based on second information, where the second information includes the capabilities of the second communication device and/or the need to process the channel information.

In combination with the second aspect, in certain implementations of the second aspect, the method also includes: the second communication device sends first information to the first communication device, the first information includes third indication information and a first vector, the third indication information is used to indicate N, and the first vector is a result of encoding the channel information through the first neural network.

In combination with the second aspect, in some implementations of the second aspect, the method further includes: the second communication device receiving parameters of the output layer of the first neural network from the first communication device.

The technical effects of the method shown in the above second aspect and its possible design can refer to the technical effects in the first aspect and its possible design.

In a third aspect, a communication device is provided, which is used to execute the method provided in the first and second aspects. Specifically, the communication device may include units and/or modules, such as a processing unit and an acquisition unit, for executing the method provided in any one of the above implementations of the first and second aspects.

In one implementation, the transceiver unit may be a transceiver, or an input/output interface; the processing unit may be at least one processor. Optionally, the transceiver may be a transceiver circuit. Optionally, the input/output interface may be an input/output circuit.

In another implementation, the transceiver unit may be an input/output interface, interface circuit, output circuit, input circuit, pin or related circuit on the chip, chip system or circuit; the processing unit may be at least one processor, processing circuit or logic circuit.

Exemplarily, the communication device is the above-mentioned first communication device or a component (such as a chip or circuit) of the first communication device, and the communication device includes:

A processing unit is used to obtain a first neural network block, wherein the first neural network block is used for channel state information and channel information feedback, and the functions of neural networks containing different numbers of the first neural network blocks are different. A transceiver unit is used to send parameters corresponding to N first neural network blocks to a second communication device, wherein the N first neural network blocks are included in a first neural network, and the first neural network is used for the second communication device to encode channel information, wherein N is a positive integer.

Exemplarily, the communication device is the above-mentioned second communication device or a component (such as a chip or circuit) of the second communication device, and the communication device includes:

A transceiver unit is used to obtain parameters corresponding to N first neural network blocks, wherein the first neural network blocks are used for channel state information and channel information feedback, and neural networks containing different numbers of the first neural network blocks have different functions. A processing unit is used to determine a first neural network based on the N first neural network blocks, wherein the first neural network is used for the second communication device to encode channel information, wherein N is a positive integer.

In a fourth aspect, the present application provides a processor for executing the methods provided in the above aspects.

For the operations such as sending and acquiring/receiving involved in the processor, unless otherwise specified, or unless they conflict with their actual function or internal logic in the relevant description, they can be understood as operations such as processor output, reception, input, etc., or as sending and receiving operations performed by the radio frequency circuit and antenna, and this application does not limit this.

In a fifth aspect, a computer-readable storage medium is provided, which stores a program code for execution by a device, wherein the program code includes a method for executing any one of the implementations of the first and second aspects above.

In a sixth aspect, a computer program product comprising instructions is provided, and when the computer program product is run on a computer, the computer executes the method provided by any one of the implementations of the first and second aspects above.

In a seventh aspect, a chip is provided, the chip comprising a processor and a communication interface, the processor reads data stored in a memory through the communication interface Instructions to execute the method provided by any one of the implementations of the first and second aspects above.

Optionally, as an implementation method, the chip also includes a memory, in which a computer program or instructions are stored, and the processor is used to execute the computer program or instructions stored in the memory. When the computer program or instructions are executed, the processor is used to execute the method provided by any implementation method of the first and second aspects above.

In an eighth aspect, a communication system is provided, comprising a first communication device for executing the method provided in the first aspect and a second communication device for executing the method provided in the second aspect.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic diagram of an application scenario to which an embodiment of the present application is applicable.

FIG2 shows a schematic diagram of a device structure provided in the present application.

FIG. 3 is a schematic diagram of a neural network.

FIG. 4 is a schematic diagram of calculating neuron output.

FIG5 is a schematic diagram of a method for AI-based CSI feedback.

FIG6 is a schematic flowchart of a communication method provided in an embodiment of the present application.

Figure 7 (a) to (d) are schematic diagrams of a neural network provided in an embodiment of the present application.

FIG8 (a) and (b) are schematic diagrams of the structure of a neural network for different requirements provided by an embodiment of the present application.

FIG9 is a schematic diagram of neural network parameters provided in an embodiment of the present application.

FIG10 is another schematic diagram of neural network parameters provided in an embodiment of the present application.

FIG. 11 is a schematic block diagram of a communication device provided in an embodiment of the present application.

FIG. 12 is a schematic diagram of another communication device provided in an embodiment of the present application.

FIG. 13 is a schematic diagram of a chip system provided in an embodiment of the present application.

DETAILED DESCRIPTION

In order to facilitate understanding of the embodiments of the present application, the following points are first explained.

First, in this application, "used for indication" may include being used for direct indication and being used for indirect indication. When describing that a certain indication information is used for indicating A, it may include that the indication information directly indicates A or indirectly indicates A, but it does not mean that the indication information must carry A.

The information indicated by the indication information is called the information to be indicated. In the specific implementation process, there are many ways to indicate the information to be indicated, such as but not limited to, the information to be indicated can be directly indicated, such as the information to be indicated itself or the index of the information to be indicated. The information to be indicated can also be indirectly indicated by indicating other information, wherein there is an association between the other information and the information to be indicated. It is also possible to indicate only a part of the information to be indicated, while the other parts of the information to be indicated are known or agreed in advance. For example, the indication of specific information can also be achieved with the help of the arrangement order of each piece of information agreed in advance (for example, stipulated by the protocol), thereby reducing the indication overhead to a certain extent. At the same time, the common parts of each piece of information can also be identified and indicated uniformly to reduce the indication overhead caused by indicating the same information separately.

Second, "at least one" shown in the present application refers to one or more, and "multiple" refers to two or more. In addition, in the embodiments of the present application, "first", "second" and various digital numbers (for example, "#1", "#2", etc.) are only for the convenience of description, and are not used to limit the scope of the embodiments of the present application. The size of the sequence number of each process below does not mean the order of execution. The execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. It should be understood that the objects described in this way can be interchanged in appropriate circumstances so as to be able to describe the schemes other than the embodiments of the present application. In addition, in the embodiments of the present application, words such as "S610" are only marks made for the convenience of description, and are not used to limit the order of execution steps.

Third, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment 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 embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

Fourth, the "storage" involved in the embodiments of the present application may refer to storage in one or more memories. The one or more memories may be separately set or integrated in an encoder or decoder, a processor, or a communication device. The one or more memories may also be partially separately set and partially integrated in a decoder, a processor, or a communication device. The type of memory may be any form of storage medium, which is not limited by the present application.

Fifth, in the implementation of this application, "protocol" may refer to a standard protocol in the communication field, for example, it may include the NR protocol and related protocols used in future communication systems, and this application does not limit this.

Sixth, in the embodiments of the present application, the terms “of”, “corresponding, relevant”, “corresponding” and “associate” can sometimes be used interchangeably. It should be pointed out that when the distinction between them is not emphasized, the meanings they intend to express are consistent.

Seventh, in the embodiments of the present application, "under the circumstances", "when", and "if" can sometimes be used interchangeably. It should be pointed out that when the distinction between them is not emphasized, the meanings they intend to express are the same.

Eighth, the term "and/or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and/or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character "/" in this article generally indicates that the associated objects before and after are in an "or" relationship.

The technical solution in this application will be described below in conjunction with the accompanying drawings.

The technical solution provided in the embodiments of the present application can be applied to wireless local area network (WLAN) scenarios, for example, supporting the Institute of Electrical and Electronics Engineers (IEEE) 802.11 related standards, such as 802.11a/b/g standards, 802.11n standards, 802.11ac standards, 802.11ax standards, IEEE 802.11ax next-generation Wi-Fi protocols such as 802.11be, Wi-Fi 7. Extremely high throughput (EHT), 802.11ad, 802.11ay or 802.11bf, such as 802.11be next generation, Wi-Fi 8, etc., can also be applied to wireless personal area network systems based on ultra-wide band (UWB), such as the 802.15 series standards, and can also be applied to sensing systems, such as the 802.11bf series standards, and can also be applied to the 802.11bn standard or the ultra-high reliability (UHR) standard. Among them, the 802.11n standard is called high throughput (HT), the 802.11ac standard is called very high throughput (VHT) standard, the 802.11ax standard is called high efficient (HE) standard, and the 802.11be standard is called extremely high throughput (EHT) standard. Among them, 802.11bf includes two major standards: low frequency (for example, sub7GHz) and high frequency (for example, 60GHz). The implementation of sub7GHz mainly relies on 802.11ac, 802.11ax, 802.11be and the next generation standards, and the implementation of 60GHz mainly relies on 802.11ad, 802.11ay and the next generation standards. Among them, 802.11ad can also be called directional multi-gigabit (DMG) standard, and 802.11ay can also be called enhanced directional multi-gigabit (EDMG) standard.

Although the embodiments of the present application are mainly described by taking the deployment of WLAN networks, especially networks using the IEEE 802.11 system standard as an example, it is easy for those skilled in the art to understand that the various aspects involved in the embodiments of the present application can be extended to other networks that adopt various standards or protocols, such as high performance radio local area networks (HIPERLAN), wireless wide area networks (WWAN), wireless personal area networks (WPAN) or other networks now known or developed later. Therefore, regardless of the coverage range and wireless access protocol used, the various aspects provided in the embodiments of the present application can be applied to any suitable wireless network.

The technical solutions of the embodiments of the present application can also be applied to various communication systems, such as: WLAN communication system, wireless fidelity (Wi-Fi) system, long term evolution (LTE) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD), universal mobile telecommunication system (UMTS), worldwide interoperability for microwave access (WiMAX) communication system, fifth generation (5G) system or new radio (NR), next generation communication system, internet of things (IoT) network or vehicle to x (V2X), etc.

The above-mentioned communication system applicable to the present application is only an example for illustration, and the communication system applicable to the present application is not limited to this. A unified description is given here and no further elaboration is given below.

FIG1 is a schematic diagram of an application scenario applicable to an embodiment of the present application. As shown in FIG1 , the communication method provided by the present application is applicable to data communication between an access point (AP) and a station (STA), wherein the station may be a non-access point station (non-AP STA), referred to as a non-AP station or STA. Specifically, the scheme of the present application is applicable to data communication between an AP and one or more non-AP stations (for example, data communication between AP1 and non-AP STA1, non-AP STA2), and also to data communication between APs (for example, data communication between AP1 and AP2), and data communication between non-AP STAs and non-AP STAs (for example, data communication between non-AP STA2 and non-AP STA3).

Among them, the access point can be a node for terminals (such as mobile phones) to enter the wired (or wireless) network. It is mainly deployed in homes, buildings and parks, with a typical coverage radius of tens to hundreds of meters. Of course, it can also be deployed outdoors. The access point is equivalent to a bridge connecting the wired network and the wireless network. Its main function is to connect various wireless network clients together and then connect the wireless network to the Ethernet.

Specifically, the access point can be a terminal or network device with a Wi-Fi chip, and the network device can be a server, a router, a switch, a bridge, a computer, a mobile phone, a relay station, a vehicle-mounted device, a wearable device, a network device in a 5G network, a network device in a future communication network, or a network device in a public land mobile communication network (public land mobile network, PLMN), etc., and the present application The embodiments are not limited. The access point may be a device that supports the Wi-Fi standard. For example, the access point may also support one or more standards of the IEEE 802.11 series, such as 802.11a, 802.11b, 802.11g, 802.11n, 802.11ac, 802.11ax, 802.11be, 802.11ad, 802.11ay, 802.11bn, 802.11bf, etc.

A non-AP site may be a wireless communication chip, a wireless sensor or a wireless communication terminal, etc., and may also be referred to as a user, user equipment (UE), an access terminal, a user unit, a user station, a mobile station, a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, a terminal, a wireless communication device, a user agent or a user device. A non-AP site may be a cellular phone, a cordless phone, a session initiation protocol (SIP) phone, a wireless local loop (WLL) station, a personal digital assistant (PDA), a handheld device with wireless communication function, a computing device or other processing device connected to a wireless modem, a vehicle-mounted device, an Internet of Things device, a wearable device, a terminal device in a 5G network, a terminal device in a future communication network or a terminal device in a PLMN, etc., and the embodiments of the present application are not limited to this. A non-AP site may be a device that supports the WLAN format. For example, a non-AP site can support one or more standards in the IEEE 802.11 series, such as 802.11a, 802.11b, 802.11g, 802.11n, 802.11ac, 802.11ax, 802.11be, 802.11ad, 802.11ay, and 802.11bf.

For example, non-AP sites can be mobile phones, tablets, set-top boxes, smart TVs, smart wearable devices, vehicle-mounted communication equipment, computers, Internet of Things (IoT) nodes, sensors, smart homes such as smart cameras, smart remote controls, smart water and electricity meters, and sensors in smart cities.

The above-mentioned AP or non-AP site may include a transmitter, a receiver, a memory, a processor, etc., wherein the transmitter and the receiver are respectively used for sending and receiving packet structures, the memory is used to store signaling information and store preset values agreed in advance, etc., and the processor is used to parse signaling information, process related data, etc.

For example, FIG2 shows a communication device provided by the present application, and the device shown in FIG2 can be an AP or a non-AP site. Specifically, the communication device involved in this embodiment includes a neural network processing unit (NPU), which is used to train neural network parameters. Optionally, the communication device may also include one or more of the following: a central processing unit, a medium access control (MAC) layer processing module, a transceiver, or an antenna, etc.

As an example and not a limitation, the NPU includes a training module and an inference module. The trained neural network parameters can be fed back to the inference module. The NPU can act on various other modules of the network node. Exemplarily, the NPU can act on the central processing unit, MAC, transceiver, or antenna. Optionally, the NPU can act on the AI tasks of each module. For example, the NPU interacts with the transceiver to decide whether to switch the transceiver on or off for energy saving; for another example, the NPU interacts with the antenna to control the direction of the antenna; for another example, the NPU interacts with the MAC to control channel access, channel selection, etc.

It should be understood that FIG. 2 is only an example of a device provided by the present application and does not constitute a limitation of the present application. For example, the device may also include a controller and/or a scheduler. For another example, the function of the NPU in the device may be implemented by a central processing unit. For another example, the modules included in the device may have other names, which are not illustrated one by one here.

In order to facilitate understanding of the technical solutions of the embodiments of the present application, some terms or concepts that may be involved in the embodiments of the present application are first briefly described.

1. Neural network (NN): It is a machine learning technology that simulates the human brain neural network in order to achieve artificial intelligence-like. A neural network consists of at least 3 layers: an input layer, an intermediate layer (also called a hidden layer), and an output layer. Deeper neural networks may contain more hidden layers between the input layer and the output layer.

For ease of understanding, the internal structure and implementation of the neural network are explained in conjunction with Figure 3. See Figure 3, which is a schematic diagram of a fully connected neural network containing 3 layers. As shown in Figure 3, the neural network includes 3 layers, namely the input layer, the hidden layer and the output layer, wherein the input layer has 3 neurons, the hidden layer has 4 neurons, the output layer has 2 neurons, and each layer of neurons is fully connected to the neurons in the next layer. Each connection between neurons corresponds to a weight, which can be updated through training. Each neuron in the hidden layer and the output layer can also correspond to a bias, which can be updated through training. Updating the neural network means updating these weights and biases. Knowing the structure of the neural network, that is, the number of neurons contained in each layer and how the output of the previous neuron is input into the subsequent neuron (that is, the connection relationship between neurons), plus the parameters of the neural network, that is, the weights and biases, you will know all the information about the neural network.

2. Neural network output: Each neuron may have multiple input connections, and each neuron calculates the output based on the input.

For ease of understanding, we briefly introduce how to calculate the output of a neuron in conjunction with Figure 4. As shown in Figure 4, a neuron contains 3 inputs, 1 output, and 2 calculation functions. The calculation formula for the output can be expressed as:

Output = activation function (input 1 * weight 1 + input 2 * weight 2 + input 3 * weight 3 + bias)      (1-1)

The symbol “*” in formula (1-1) represents the mathematical operation “multiplication” or “times”, which will not be described in detail below.

It should be understood that each neuron may have multiple output connections, and the output of one neuron serves as the input of the next neuron. The input layer has only output connections. Each neuron in the input layer is the value of the input neural network. The output value of each neuron is directly used as the input of all output connections. The output layer has only input connections, and the output is calculated using the above formula (1-1).

Optionally, the output layer may not have the activation function calculated, that is, the above formula (1-1) can be transformed into:

Output = Input 1 * Weight 1 + Input 2 * Weight 2 + Input 3 * Weight 3 + Bias.

For example, the output of a k-layer neural network can be expressed as:

y=fk(fk-1(...(fi(wi*x+bi)))

In formula (1-2), x represents the input of the neural network, y represents the output of the neural network, wi represents the weight of the i-th layer of the neural network, bi represents the bias of the i-th layer of the neural network, and fi represents the activation function of the i-th layer of the neural network. i＝1,2,…,k.

3. Artificial intelligence (AI): Wireless communications are developing rapidly. 5G and the sixth-generation Wi-Fi standards have been commercialized, and the next-generation wireless technology and standardization are in full swing around the world. Wireless communications have penetrated into all aspects of daily life and work and have become an indispensable part. With the rapid growth in the number of smart terminals and the popularity of IoT devices, new wireless applications such as virtual reality, augmented reality, and holographic imaging have emerged. New wireless technologies, new terminals, and new applications have made wireless networks unprecedentedly complex. It is foreseeable that wireless networks will become more and more complex in the future. In order to combat the high complexity of wireless networks, AI has become an industry consensus as an effective tool for wireless network design and management. In wireless networks, the advantages of AI are reflected in four aspects:

1) Solve complex network problems without mathematical models; 2) Solve wireless network management problems with large search spaces; 3) Bring about cross-layer and cross-node network-level global optimization effects; 4) Actively optimize wireless network parameters through AI’s predictive capabilities.

4. CSI feedback based on Givens rotation: The current standard uses a CSI feedback method based on Givens rotation. For example, the AP first sends NDPA, then sends NDP for channel measurement. After the STA estimates the channel, it obtains the current channel H, performs singular value decomposition (SVD) on H to obtain the precoding matrix V, performs Givens rotation on V to convert it into two types of angles, φ and ψ, and then quantizes and feeds back these angles to the AP. The AP recovers from the received angles Used for precoding.

The above text briefly introduces the scenarios in which the communication method provided in the embodiment of the present application can be applied in combination with Figure 1, and introduces the basic concepts that may be involved in the embodiment of the present application, and introduces the connection between AI and wireless communication in the basic concepts, as well as the CSI feedback method based on Givens rotation in the current standard. It should be understood that with the increase in bandwidth and the increase in the number of antennas, continuing to use the CSI feedback method based on Givens rotation for channel state information feedback will greatly increase the overhead of channel state information feedback. And with the development of AI technology, the AI-based CSI feedback scheme has caused extensive discussion in the Wi-Fi standard.

Exemplarily, AI-based CSI feedback solutions include but are not limited to the following two solutions:

Solution 1: Use AI methods to maximize the CSI compression ratio and reduce feedback overhead, thereby improving system throughput.

For example, the specific implementation of solution 1 may be:

In the training phase, the STA performs CSI feedback according to the standard method (e.g., the CSI feedback method based on Givens rotation), and the AP uses the collected data to train the encoder and decoder. After the training is completed, the encoder is sent to the STA. In the inference phase, after the STA performs channel measurement, the precoding matrix V is input into the encoder, and the output vector m is obtained and fed back to the AP. The AP uses the decoder to recover the vector m based on m. Pre-encoding.

For ease of understanding, the AI-based CSI feedback is introduced in conjunction with FIG5 under the situation shown in the first solution.

FIG5 is a schematic diagram of a method for CSI feedback based on AI, comprising the following steps:

S510: The AP sends an NDP to the STA. The NDP is used for channel measurement.

S520, STA performs channel estimation. Specifically, after performing channel estimation based on NDP, STA obtains the current channel H, performs SVD on H to obtain the precoding matrix V, and performs Givens rotation operation on V to convert it into two types of angles, φ and ψ.

S530, the STA sends φ and ψ to the AP.

S540, the AP determines an encoder and a decoder. Specifically, the AP trains the encoder and the decoder using the collected data.

S550: The AP sends the encoder to the STA.

It should be understood that steps S510 to S550 are the aforementioned training phase, during which the STA performs CSI feedback according to a standard method.

S560: The AP sends an NDP to the STA.

S570, STA performs channel estimation. Specifically, after performing channel measurement, STA inputs the precoding matrix V into the encoder to obtain an output result m.

S580, the STA sends m to the AP.

S590, AP determines precoding matrix Specifically, the AP decodes m based on the decoder to obtain the precoding matrix

Solution 2: Use AI methods to reduce the computational complexity of the CSI feedback process. On the basis of ensuring that the feedback amount is no greater than the Givens rotation-based method used in the current standard, a neural network is used to replace the Givens rotation calculation to reduce the computational complexity.

The specific implementation of Scheme 2 may be: after the STA performs channel measurement, the precoding matrix V is input into the encoder (e.g., a fully connected neural network) to obtain the output result m'. It should be understood that in the case shown in Scheme 2, the focus is on reducing the complexity of the CSI feedback calculation process by reducing the computational complexity of determining m' based on the coding matrix V. Therefore, the encoder used by the STA in Scheme 2 has a different structure from the encoder used by the STA in Scheme 1.

The above two solutions correspond to different encoder structures. Different encoder structures will require multiple sets of interconnection and interoperability protocol support, which will cause management overhead. In addition, different encoder structures will also increase the storage overhead of the encoder.

The present application provides a communication method for obtaining a neural network block that can be included in a neural network (or encoder) with different functions to meet the needs of neural networks in different scenarios and improve the performance of channel information feedback based on AI. The communication method will be described in detail below in conjunction with FIG6.

The technical solution provided by the present application will be described in detail below with reference to the accompanying drawings. The embodiments of the present application can be applied to a plurality of different scenarios, including the scenario shown in FIG1 , but are not limited to the scenario. For example, it can also be applied to 5G, next generation communication systems or future communication systems.

It should be understood that the embodiments shown below do not particularly limit the specific structure of the execution subject of the method provided by the embodiments of the present application. As long as it is possible to communicate according to the method provided by the embodiments of the present application by running a program that records the code of the method provided by the embodiments of the present application, for example, the execution subject of the method provided by the embodiments of the present application may be a receiving device or a sending device, or a functional module in the receiving device or the sending device that can call and execute the program.

In the following, without loss of generality, the communication method provided in the embodiment of the present application is described in detail by taking the interaction between the first communication device and the second communication device as an example. The first communication device involved in the embodiment of the present application may be an access point AP, and the second communication device may be a non-access point non-AP (such as STA); or, the first communication device may be a STA, and the second communication device may be an AP; or, the first communication device and the second communication device may be an access point AP; or, the first communication device and the second communication device may be a non-access point non-AP; or, the first communication device may be a network device (or a centralized unit (central unit, CU) or a distributed unit (distributed unit, DU) in the network device), and the second communication device may be a terminal device; or, the first communication device may be a network device in an open radio access network (open radio access network, O-RAN) (or a CU (such as, called O-CU (open CU)) or DU (such as, called O-DU (open DU)) in the network device), and the second communication device may be a terminal device, etc.

In addition, the first neural network block involved in the following embodiments may be referred to as an NN block. The first neural network block is understood as a network structure that can be stacked (or reused). The first neural network block may include one or more layers of convolutional neural networks (CNN), fully-connected networks (FC), or transformers, etc. For example, the network structure of the first neural network block is a network structure including one or more layers of CNN; for another example, the network structure of the first neural network block is a network structure including one or more layers of FC; for another example, the network structure of the first neural network block is a network structure including one or more layers of Transformer; for another example, the network structure of the first neural network block is a network structure including one or more layers of CNN and one or more layers of FC, etc., and examples are not given one by one here.

FIG6 is a schematic flow chart of a communication method provided in an embodiment of the present application, comprising the following steps:

S610, the first communication device obtains a first neural network block.

Specifically, the neural network block (e.g., the first neural network block) involved in this embodiment is used for channel information feedback, and can assist in processing channel information during the channel information feedback process. Neural networks containing different numbers of first neural network blocks have different functions for processing channel information.

The channel information involved in this embodiment includes but is not limited to the above-mentioned CSI, and the CSI includes a precoding matrix indicator (PMI), and the PMI is used to indicate the precoding matrix. In addition, the channel information involved in this embodiment may also include other information that can be used to reflect the channel state, information indicating the channel state between the first communication device and the second communication device, or other information carrying the PMI. For ease of description, the channel information is CSI as an example for explanation below.

Exemplarily, the first neural network block is any one of the at least one neural network block acquired by the first communication device. In this embodiment, there is no limitation on the number of neural network blocks acquired by the first communication device, and different neural network blocks have different structures. For example, the first communication device acquires a first neural network block and a second neural network block, and the structure of the first neural network block is different from the structure of the second neural network block.

Specifically, the different functions of the neural networks containing different numbers of first neural network blocks can be understood as: the first neural network blocks in this embodiment can achieve different functions by stacking. For example, neural network #1 includes N1 first neural network blocks, and neural network #2 includes N2 first neural network blocks, N1 and N2 are unequal positive integers, and the functions achieved by neural network #1 and neural network #2 are different. As an example and not a limitation, neural network #1 can achieve the function of the encoder in the above-mentioned scheme one (improve CSI compression ratio, reduce feedback overhead, and improve system throughput), and neural network #2 can achieve the function of the encoder in the above-mentioned scheme two (reduce the computational complexity in the CSI feedback process). Spend).

When a neural network includes multiple first neural network blocks, the connection between the multiple first neural network blocks can be deep connection or wide connection.

To facilitate understanding, the following briefly introduces four neural network structures obtained by stacking the first neural network blocks in combination with (a) to (d) in Figure 7.

As shown in (a) of Figure 7, a neural network includes multiple first neural network blocks supporting CNN and an output layer, wherein the multiple first neural network blocks supporting CNN are connected via deep connections, the multiple first neural network blocks supporting CNN have the same structure but different corresponding parameters, and the parameters of each first neural network block are related to the position of the first neural network block in the neural network.

As shown in (b) of Figure 7, a neural network includes multiple first neural network blocks supporting FC and an output layer, wherein the multiple first neural network blocks supporting FC are connected via deep connections, and the multiple first neural network blocks supporting FC have the same structure but different corresponding parameters.

As shown in (c) of FIG. 7 , a neural network includes multiple first neural network blocks supporting FC and an output layer, wherein the multiple first neural network blocks supporting FC are connected via width connections, and the multiple first neural network blocks supporting FC have the same structure but different corresponding parameters.

As shown in (d) of FIG. 7 , a neural network includes a plurality of first neural network blocks supporting Transformer and an output layer, wherein the plurality of first neural network blocks supporting Transformer are connected by width connection.

As can be seen from (a) to (d) in FIG7 , the connection mode between different first neural network blocks included in a certain neural network can be a deep connection and/or a wide connection, wherein the deep connection mode between two first neural network blocks means that the output of the first first neural network block of the two first neural network blocks is used as the input of the second first neural network block. The wide connection mode between two first neural network blocks means that the inputs of the two first neural network blocks are the same data (or the outputs of the same neural network block), and the outputs of the two first neural network blocks are output to the same output layer or the next neural network block.

It should be understood that (a) to (d) in FIG. 7 are only illustrative examples of how the first neural network block and the output layer can constitute neural networks for different requirements, and do not constitute any limitation on the scope of protection of the present application. In this embodiment, the first neural network block and the output layer can also constitute neural networks of other modes, for example, neural networks with different numbers of layers, which will not be explained one by one here.

It should be noted that for a deep-connected or wide-connected neural network, the more first neural network blocks it contains, the higher the computational complexity, the fewer feedback bits, and the higher the system throughput. On the contrary, if the fewer first neural network blocks it contains, the lower the computational complexity of the neural network.

For ease of understanding, the following describes the effect of different numbers of first neural network blocks on the computational complexity of the constructed neural network in conjunction with (a) and (b) in FIG8 .

Exemplarily, as shown in FIG8(a), the neural network includes 4 identical first neural network blocks that are deeply connected based on convolutional neural networks. The structure of each of the 4 first neural network blocks is shown in the dotted box in FIG8(a), that is, the dotted box part in FIG8(a) is repeated 4 times (the repeated part is not shown in FIG8(a)). The dotted box in FIG8(a) shows a first neural network block based on convolution.

It should be understood that the first neural network block structure shown in the dotted box in Figure 8 (a) is only an example and does not constitute any limitation on the scope of protection of the present application.

As shown in FIG8(b), the neural network includes a first neural network block with a deep connection based on a convolutional neural network, and the structure of the first neural network block is shown in the dotted box in FIG8(a).

It should be understood that compared with the neural network shown in Figure 8 (a), the neural network shown in Figure 8 (b) includes a smaller number of first neural network blocks (1), so the neural network shown in Figure 8 (b) has the characteristic of low computational complexity and is suitable for the second AI-based CSI feedback solution shown in the previous text.

Compared with the neural network shown in FIG8(b), the number of first neural network blocks (4) included in the neural network shown in FIG8(a) is greater, and the neural network shown in FIG8(b) has the following characteristics:

The characteristics of high computational complexity, fewer feedback bits, and higher system throughput are suitable for solution 1 of the AI-based CSI feedback shown in the previous article.

As an example but not a limitation, in this embodiment, the first communication device may obtain the first neural network block based on the following possible implementations:

As a possible implementation, in this embodiment, the first communication device can determine the first neural network block by itself. For example, the first communication device obtains the first neural network block by training based on the training data. The collected data may be historical communication data between the first communication device and the second communication device, or may be training data provided by the management device, or may be generated by the first communication device for training. In this embodiment, there is no limitation on how to obtain the data for training.

As another possible implementation, in this embodiment, the first communication device may determine the first neural network block in a predefined manner, such as a protocol predetermines the structure of at least one neural network block.

As another possible implementation, in this embodiment, the first communication device may negotiate with the second communication device to determine the first neural network block.

Further, in this embodiment, after the first communication device obtains the first neural network block, the parameters of the corresponding neural network block can be provided to the second communication device with different requirements, so that the second communication device can determine the corresponding neural network based on the parameters of the received neural network block, thereby performing channel state information compression processing based on the determined neural network and performing uplink channel state information feedback. The method flow shown in FIG6 may optionally also include:

S620, the first communication device sends parameters corresponding to the N first neural network blocks to the second communication device, and correspondingly, the second communication device receives parameters corresponding to the N first neural network blocks from the first communication device.

Specifically, N first neural network blocks are included in a first neural network, and the first neural network is used for the second communication device to encode CSI. The parameters corresponding to the N first neural network blocks include: the parameters corresponding to the i-th first neural network block in the first neural network, and i takes values from 1 to N. It should be understood that the N first neural network blocks are the first neural network blocks stacked N times, which is equivalent to the neural network blocks with the same network structure being repeated N times in the first neural network, and the parameters of the first neural network blocks of different layers of the first neural network (such as the weights, biases or activation functions shown in FIG. 4, etc.) are different, so the first communication device needs to send the parameters corresponding to the N first neural network blocks contained in the first neural network to the second communication device.

For example, the first neural network includes two first neural network blocks: the first neural network block #1 and the first neural network block #2. The first neural network block #1 and the first neural network block #2 have the same structure but different parameters. It can be understood that the first neural network block is stacked twice.

Exemplarily, the first communication device may also send the parameters of the output layer of the first neural network to the second communication device. The output layer refers to the neural network used to obtain the final feedback information (such as m or m' shown above) from the output of the last neural network block, and may include one or more layers of FC or CNN, etc.

It should be noted that the above-mentioned first neural network block can also be designed as a neural network block that can serve as an output layer. In this case, the first communication device may not need to provide the parameters of the output layer of the first neural network to the second communication device; or, when the output layer of the first neural network can reuse the output layer of a known neural network, the first communication device may not need to provide the parameters of the output layer of the first neural network to the second communication device.

It should be understood that in this embodiment, the first communication device can provide parameters of the neural network blocks required to form a neural network to one or more second communication devices. For the sake of convenience in description, this embodiment is described by taking an example in which the first communication device provides parameters corresponding to N first neural network blocks contained in the first neural network to a second communication device.

Optionally, the first communication device may also provide the second communication device with parameters such as the quantization method used, the number of quantization bits, and the feature embedding method.

Exemplarily, in this embodiment, the first communication device may provide the second communication device with the parameters corresponding to the N first neural network blocks mentioned above respectively through the following possible implementation methods:

Method 1: The first communication device sends the parameters corresponding to the above-mentioned N first neural network blocks to the second communication device in response to the request of the second communication device.

In the case shown in the first mode, the method flow shown in FIG6 further includes:

S621, the second communication device determines that the number of first neural network blocks included in the first neural network is N according to the second information.

Specifically, the second communication device can determine that the number of first neural network blocks included in the required first neural network is N based on the second information, and the second information includes the capabilities of the second communication device and/or the need to process CSI.

For example, if the second communication device is a device that meets certain requirements (for example, the second communication device is a device with sufficient computing power or high performance, or the second communication device needs to perform high compression on CSI to reduce feedback overhead and improve the effective throughput of the system), then the second communication device can request the first communication device for N first neural network blocks included in the first neural network based on the second information, where N is greater than a threshold.

For another example, if the second communication device is a device that does not meet certain requirements (for example, the second communication device is a device with insufficient computing power, poor performance, or insufficient power, or the second communication device needs to perform compression processing on the CSI with low computational complexity to reduce processing complexity), the second communication device can request the first communication device for the N first neural network blocks included in the first neural network based on the second information, where N is less than a threshold.

S622: The second communication device sends first indication information to the first communication device, and correspondingly, the first communication device receives the first indication information from the second communication device.

Specifically, the first indication information is used to indicate the above-mentioned N. It indicates that the number of first neural network blocks required by the second communication device is N.

For ease of understanding, the following describes how the second communication device requests the parameters of the required neural network block in the case described in the first method with reference to a specific example:

Example 1: Assume that the neural network (or encoder) required by the second communication device is a neural network structure including a maximum of 4 neural network blocks.

When the computing power of the second communication device is sufficient, model 1 (hereinafter referred to as model 1) containing three neural network blocks is requested according to the second information; when the computing power of the second communication device is insufficient (for example, the second communication device is an IoT terminal or the second communication device is low on power), the second communication device requests model 2 (hereinafter referred to as model 2) containing one neural network block according to the second information.

Optionally, the second communication device may use 4 bits to indicate whether the parameters of 4 neural network blocks are needed. Therefore, the second communication device may use 0111 to request model 1 and 0001 to request model 2. The first communication device sends the parameters of the corresponding neural network block to the second communication device according to the received request.

Exemplarily, the parameters provided by the first communication device may be as shown in Figure 9. It should be understood that Figure 9 is only an example, and the parameters provided by the first communication device may not include the parameters of the output layer, which will not be described one by one here.

Method 2: The first communication device broadcasts parameters corresponding to multiple first neural network blocks.

In the case shown in the second mode, the method flow shown in FIG6 further includes:

S623, the first communication device determines the maximum value P of the number of first neural network blocks.

Specifically, the first communication device determines the maximum value P of the number of first neural network blocks that the neural network can include, where P is a positive integer greater than or equal to N.

In the case shown in the second method, the above-mentioned first communication device sends parameters corresponding to N first neural network blocks to the second communication device, including: the first communication device sends parameters corresponding to P first neural network blocks to the second communication device.

After the second communication device receives the parameters broadcast by the first communication device, the second communication device may select different parameters according to its own capabilities or needs. In the case shown in the second mode, the method flow shown in FIG. 6 further includes:

S624, the second communication device determines N first neural network blocks based on the second information.

For example, if the second communication device is a device that meets certain requirements (for example, the second communication device is a device with sufficient computing power or high performance, or the second communication device needs to perform high compression on CSI to reduce feedback overhead and improve the effective throughput of the system), then the second communication device can select N first neural network blocks from P first neural network blocks based on the second information, where N is greater than a threshold.

For example, if the second communication device is a device that does not meet certain requirements (for example, the second communication device is a device with insufficient computing power, poor performance, or insufficient power, or the second communication device needs to perform compression processing on the CSI with low computational complexity to reduce processing complexity), the second communication device can select N first neural network blocks from P first neural network blocks based on the second information, where N is less than a threshold.

Furthermore, when the neural network required by the second communication device needs to be updated, the second communication device may request the parameters corresponding to the first neural network block contained in the updated neural network through the second indication information. The need to update the neural network required by the second communication device may be due to a change in the capability or demand of the second communication device. For example, the second communication device is initially low on power and uses the above-mentioned model 2; after the power is restored, it is desired to use the above-mentioned model 1; for example, the second communication device initially requires low-computational-complexity compression processing of the CSI to reduce processing complexity, and after a period of time, the requirement is updated to high compression of the CSI to reduce feedback overhead.

It should be understood that the conditions for the second communication device to determine that the neural network needs to be updated are only examples, and this embodiment does not limit the reasons why the second communication device determines that the neural network needs to be updated.

As a possible implementation manner, when the number of first neural network blocks included in the neural network required by the second communication device changes, the second communication device may request parameters of the newly added first neural network blocks from the first communication device.

For example, the second communication device knows the parameters corresponding to the N first neural network blocks included in the first neural network. When the neural network required by the second communication device is updated from the first neural network to the second neural network, and the second neural network includes Q first neural network blocks, Q-N = M. Then the second communication device can send a second indication message to the first communication device, and the second indication message indicates M, requesting the parameters of the newly added M first neural network blocks.

As a possible implementation manner, when the number of first neural network blocks included in the neural network required by the second communication device changes, the second communication device may request the first communication device to update the parameters of the first neural network blocks included in the neural network.

For example, the second communication device knows the parameters corresponding to N first neural network blocks included in the first neural network. When the neural network required by the second communication device is updated from the first neural network to the second neural network, and the second neural network includes Q first neural network blocks, the second communication device can request the parameters of the Q first neural network blocks from the first communication device.

For ease of understanding, the following describes how the second communication device updates the neural network with reference to specific examples.

Example 2: The second communication device is initially low on power and uses the above-mentioned model 2; after the power is restored, it is desired to use the above-mentioned model 1.

The second communication device can request the parameters of the additional first neural network blocks required to update from model 2 (1 first neural network block, such as 0001) to model 1 (3 first neural network blocks, such as 0111), i.e., 2 first neural network blocks (e.g., 0110), effectively reducing the transmission overhead of the neural network update.

Exemplarily, the parameters provided by the first communication device may be as shown in Figure 10. It should be understood that Figure 10 is only an example, and in the case where the neural network required by the second communication device is updated, the parameters provided by the first communication device may also be the parameters of the first neural network block included in the updated neural network, which will not be illustrated one by one here.

In this embodiment, after the second communication device obtains the parameters corresponding to the N first neural network blocks, the first neural network can be determined, and CSI processing is performed based on the first neural network. The method flow shown in FIG6 also includes:

S630: The second communication device performs CSI processing.

Specifically, in this embodiment, the second communication device processes the CSI based on the determined first neural network. For example, the second communication device compresses the precoding matrix V or other channel state information based on the first neural network.

It should be understood that the premise for the second communication device to perform channel processing in this embodiment is that the second communication device receives measurement information from the first communication device (such as the NDP shown above, or other measurement reference signals), and the second communication device can perform channel estimation based on the measurement information to obtain the current channel H, and perform SVD on H to obtain the precoding matrix V. This embodiment mainly involves compression processing of V based on the first neural network, and other processes are not limited. For example, the process of the second communication device obtaining the parameters of the first neural network block included in the first neural network and the process of the second communication device receiving the measurement information are independent processes.

When the number of first neural network blocks included in the above-mentioned first neural network is greater than a certain threshold (for example, greater than 4 first neural network blocks), the computational complexity of the first neural network is high, and the second communication device can improve the CSI compression ratio based on the first neural network, reduce feedback overhead, and improve system throughput.

When the number of first neural network blocks included in the above-mentioned first neural network is less than a certain threshold (e.g., less than 2 first neural network blocks), the computational complexity of the first neural network is low, and the second communication device can reduce the computational complexity in the CSI feedback process based on the first neural network.

S640: The second communication device sends first information to the first communication device, and correspondingly, the first communication device receives the first information from the second communication device.

Specifically, the first information includes third indication information and a first vector, the third indication information is used to indicate N, and the first vector is the result of CSI being encoded by the first neural network. It should be understood that the third indication information is used to indicate the structure of the neural network based on which the first vector currently fed back by the second communication device is processed, thereby helping the first communication device to select a suitable neural network to parse the first vector to obtain high-accuracy channel information.

Optionally, the third indication information can be carried in a multiple input multiple output (MIMO) control field, or in a MIMO compressed beamforming report (CBR) field, as shown in Figure 10.

In the communication method shown in FIG6 , the first communication device can obtain the first neural network block, and when different neural networks contain different numbers of first neural network blocks, different neural networks implement different functions. For example, the more first neural network blocks a neural network contains, the higher the computational complexity of the neural network, and the compression ratio of the CSI compressed and processed by the neural network is high, the feedback overhead is reduced, and the system throughput is high. Furthermore, the first communication device can provide the second communication device with parameters corresponding to the N first neural network blocks, so that the second communication device can determine the first neural network based on the parameters corresponding to the N received first neural network blocks, and can encode the CSI based on the first neural network in the subsequent CSI feedback process to implement AI-based CSI feedback.

Moreover, in this technical solution, neural networks with different functions can be obtained through a neural network block of a structure, without the need to train different neural networks for different functions, thereby reducing management overhead and storage overhead of the neural network.

It should be understood that the sequence numbers of the above processes do not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

It should also be understood that in the various embodiments of the present application, unless otherwise specified or there is a logical conflict, the terms and/or descriptions between different embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form new embodiments according to their internal logical relationships.

It should also be understood that in some of the above embodiments, the devices in the existing network architecture are mainly used as examples for exemplary description, and it should be understood that the embodiments of the present application do not limit the specific form of the devices. For example, devices that can achieve the same function in the future are applicable to the embodiments of the present application.

It can be understood that in the above-mentioned various method embodiments, the methods and operations implemented by the device (such as the first communication device and the second communication device) can also be implemented by components that can be used in the device (such as chips or circuits).

It can also be understood that some optional features in the embodiments of the present application may not depend on other features in some scenarios, or may be combined with other features in some scenarios, without limitation.

The communication method provided by the embodiment of the present application is described in detail above in conjunction with FIG6. The above communication method is mainly introduced from the perspective of the first communication device and the second communication device. It can be understood that in order to implement the above functions, the first communication device and the second communication device include hardware structures and/or software modules corresponding to the execution of each function.

Those skilled in the art should be aware that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is performed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

The communication device provided in the embodiment of the present application is described in detail below in conjunction with Figures 11 to 13. It should be understood that the description of the device embodiment corresponds to the description of the method embodiment, so the content not described in detail can refer to the above method embodiment, and for the sake of brevity, some content will not be repeated.

The embodiment of the present application can divide the functional modules of the transmitting end device or the receiving end device according to the above method example. For example, each functional module can be divided corresponding to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical functional division. There may be other division methods in actual implementation. The following is an example of dividing each functional module corresponding to each function.

FIG11 is a schematic block diagram of a communication device 10 provided in an embodiment of the present application. The device 10 includes a transceiver module 11 and a processing module 12. The transceiver module 11 can implement corresponding communication functions, and the processing module 12 is used to perform data processing, or in other words, the transceiver module 11 is used to perform operations related to receiving and sending, and the processing module 12 is used to perform other operations besides receiving and sending. The transceiver module 11 can also be called a communication interface or a communication unit.

Optionally, the device 10 may further include a storage module 13, which may be used to store instructions and/or data. The processing module 12 may read the instructions and/or data in the storage module so that the device implements the actions of the devices in the aforementioned method embodiments.

In one design, the device 10 may correspond to the first communication device in the above method embodiment, or a component (such as a chip) of the first communication device.

The device 10 can implement the steps or processes executed by the first communication device in the above method embodiment, wherein the transceiver module 11 can be used to execute the transceiver related operations of the first communication device in the above method embodiment, and the processing module 12 can be used to execute the processing related operations of the first communication device in the above method embodiment.

In a possible implementation, the processing module 12 is used to obtain a first neural network block, the first neural network block is used for channel information feedback, and the functions of neural networks containing different numbers of the first neural network blocks are different. The transceiver module 11 is used to send parameters corresponding to N first neural network blocks to the second communication device, the N first neural network blocks are included in the first neural network, and the first neural network is used for encoding processing of channel information, wherein N is a positive integer.

When the device 10 is used to execute the method in Figure 6, the transceiver module 11 can be used to execute the steps of sending and receiving information in the method, such as steps S622, S620 and S640, and the processing module 12 can be used to execute the processing steps in the method, such as steps S610 and S623.

It should be understood that the specific process of each unit executing the above corresponding steps has been described in detail in the above method embodiment, and for the sake of brevity, it will not be repeated here.

In another design, the device 10 may correspond to the second communication device in the above method embodiment, or be a component (such as a chip) of the second communication device.

The device 10 can implement steps or processes corresponding to those performed by the second communication device in the above method embodiment, wherein the transceiver module 11 can be used to perform transceiver-related operations of the second communication device in the above method embodiment, and the processing module 12 can be used to perform processing-related operations of the second communication device in the above method embodiment.

In a possible implementation, the transceiver module 11 is used to obtain parameters corresponding to N first neural network blocks, respectively, where the first neural network blocks are used for channel information feedback, and the functions of neural networks containing different numbers of the first neural network blocks are different. The processing module 12 is used to determine a first neural network based on the N first neural network blocks, where the first neural network is used for the second communication device to encode the channel information, wherein N is a positive integer.

When the device 10 is used to execute the method in FIG. 6 , the transceiver module 11 may be used to execute the steps of sending and receiving information in the method, such as step S622, S620 and S640, the processing module 12 can be used to execute the processing steps in the method, such as steps S621, S624 and S630.

It should be understood that the specific process of each unit executing the above corresponding steps has been described in detail in the above method embodiment, and for the sake of brevity, it will not be repeated here.

It should also be understood that the device 10 here is embodied in the form of a functional module. The term "module" here may refer to an application specific integrated circuit (ASIC), an electronic circuit, a processor (such as a shared processor, a dedicated processor or a group processor, etc.) and a memory for executing one or more software or firmware programs, a merged logic circuit and/or other suitable components that support the described functions. In an optional example, those skilled in the art can understand that the device 10 can be specifically the mobile management network element in the above-mentioned embodiment, and can be used to execute the various processes and/or steps corresponding to the mobile management network element in the above-mentioned method embodiments; or, the device 10 can be specifically the terminal device in the above-mentioned embodiment, and can be used to execute the various processes and/or steps corresponding to the terminal device in the above-mentioned method embodiments. To avoid repetition, it will not be repeated here.

The device 10 of each of the above schemes has the function of implementing the corresponding steps performed by the device (such as the first communication device) in the above method. The function can be implemented by hardware, or by hardware executing the corresponding software implementation. The hardware or software includes one or more modules corresponding to the above functions; for example, the transceiver module can be replaced by a transceiver (for example, the sending unit in the transceiver module can be replaced by a transmitter, and the receiving unit in the transceiver module can be replaced by a receiver), and other units, such as the processing module, can be replaced by a processor to respectively perform the transceiver operations and related processing operations in each method embodiment.

In addition, the transceiver module 11 may also be a transceiver circuit (for example, may include a receiving circuit and a sending circuit), and the processing module may be a processing circuit.

FIG12 is a schematic diagram of another communication device 20 provided in an embodiment of the present application. The device 20 includes a processor 21, and the processor 21 is used to execute a computer program or instruction stored in a memory 22, or read data/signaling stored in the memory 22 to execute the method in each method embodiment above. Optionally, there are one or more processors 21.

Optionally, as shown in FIG12 , the device 20 further includes a memory 22, and the memory 22 is used to store computer programs or instructions and/or data. The memory 22 can be integrated with the processor 21, or can also be separately arranged. Optionally, the memory 22 is one or more.

Optionally, as shown in Fig. 12, the device 20 further includes a transceiver 23, and the transceiver 23 is used for receiving and/or sending signals. For example, the processor 21 is used to control the transceiver 23 to receive and/or send signals.

As a solution, the device 20 is used to implement the operations performed by the first communication device or the second communication device in the above various method embodiments.

It should be understood that the processor mentioned in the embodiments of the present application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

It should also be understood that the memory mentioned in the embodiments of the present application may be a volatile memory and/or a non-volatile memory. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM). For example, a RAM may be used as an external cache. By way of example and not limitation, RAM includes the following forms: static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

It should be noted that when the processor is a general-purpose processor, DSP, ASIC, FPGA or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, the memory (storage module) can be integrated into the processor.

It should also be noted that the memory described herein is intended to include, but is not limited to, these and any other suitable types of memory.

FIG13 is a schematic diagram of a chip system 30 provided in an embodiment of the present application. The chip system 30 (or also referred to as a processing system) includes a logic circuit 31 and an input/output interface 32.

The logic circuit 31 may be a processing circuit in the chip system 30. The logic circuit 31 may be coupled to a storage unit and call instructions in the storage unit so that the chip system 30 can implement the methods and functions of the various embodiments of the present application. The input and output circuits in the chip system 30 output information processed by the chip system 30 , or input data or signaling information to be processed into the chip system 30 for processing.

As a solution, the chip system 30 is used to implement the operations performed by the first communication device or the second communication device in the above method embodiments.

For example, the logic circuit 31 is used to implement the processing-related operations performed by the first communication device or the second communication device in the above method embodiment; the input/output interface 32 is used to implement the sending and/or receiving-related operations performed by the terminal device in the above method embodiment.

An embodiment of the present application also provides a computer-readable storage medium on which computer instructions for implementing the methods executed by the device in the above-mentioned method embodiments are stored.

For example, when the computer program is executed by a computer, the computer can implement the method performed by the first communication device or the second communication device in each embodiment of the above method.

An embodiment of the present application further provides a computer program product, comprising instructions, which, when executed by a computer, implement the methods performed by the first communication device or the second communication device in the above-mentioned method embodiments.

An embodiment of the present application also provides a communication system, including the aforementioned first communication device and second communication device.

The explanation of the relevant contents and beneficial effects of any of the above-mentioned devices can be referred to the corresponding method embodiments provided above, which will not be repeated here.

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

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

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

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

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

If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims (24)

A communication method, characterized by comprising: The first communication device obtains a first neural network block, where the first neural network block is used for channel information feedback, and neural networks containing different numbers of the first neural network blocks have different functions; The first communication device sends parameters corresponding to N first neural network blocks to the second communication device, wherein the N first neural network blocks are included in a first neural network, and the first neural network is used for encoding processing of the channel information. Wherein, N is a positive integer. The method according to claim 1 is characterized in that before the first communication device sends the parameters corresponding to the N first neural network blocks to the second communication device, the method further comprises: The first communication device receives first indication information from the second communication device, where the first indication information is used to indicate the N. The method according to claim 1 or 2, characterized in that the method further comprises: The first communication device receives second indication information from the second communication device, where the second indication information is used to indicate M, where M is a positive integer; The first communication device sends the parameters corresponding to the M first neural network blocks to the second communication device, Among them, the M first neural network blocks and the N first neural network blocks are included in the second neural network. The method according to any one of claims 1 to 3, characterized in that the method further comprises: The first communication device determines a maximum value P of the number of first neural network blocks that the neural network can include, where P is a positive integer greater than or equal to N; The first communication device sends parameters corresponding to N first neural network blocks to the second communication device, including: The first communication device sends the parameters corresponding to the P first neural network blocks to the second communication device. The method according to any one of claims 1 to 4, characterized in that the method further comprises: The first communication device receives first information from the second communication device, the first information includes third indication information and a first vector, the third indication information is used to indicate N, and the first vector is a result obtained by processing the channel information through the first neural network encoding. The method according to any one of claims 1 to 5, characterized in that the parameters of the first neural network block include at least one of the following: Weight information, bias information, or activation function information corresponding to the first neural network block. The method according to any one of claims 1 to 6, characterized in that the first neural network block supports at least one of the following neural network structures: Convolutional Neural Network CNN, Multi-layer Perceptron MLP, or Transformer Transformer. The method according to any one of claims 1 to 7 is characterized in that when the first neural network includes multiple first neural network blocks, the connection method between the multiple first neural network blocks includes deep connection and/or width connection. The method according to any one of claims 1 to 8, characterized in that the method further comprises: The first communication device sends parameters of an output layer of the first neural network to the second communication device. A communication method, comprising: The second communication device obtains parameters corresponding to N first neural network blocks respectively, where the first neural network blocks are used for channel information feedback, and neural networks containing different numbers of the first neural network blocks have different functions; The second communication device determines a first neural network based on the N first neural network blocks, and the first neural network is used by the second communication device to encode the channel information, wherein N is a positive integer. The method according to claim 10 is characterized in that the second communication device obtains parameters corresponding to the N first neural network blocks respectively, including: The second communication device receives parameters corresponding to the N first neural network blocks respectively from the first communication device. The method according to claim 11, characterized in that before the second communication device receives the parameters corresponding to the N first neural network blocks respectively from the first communication device, the method further comprises: The second communication device determines the number N of first neural network blocks included in the required first neural network according to second information, wherein the second information includes the capability of the second communication device and/or the requirement for processing the channel information; The second communication device sends first indication information to the first communication device, where the first indication information is used to indicate the N. The method according to any one of claims 10 to 12, characterized in that after the second communication device obtains the parameters corresponding to the N first neural network blocks respectively, the method further comprises: The second communication device determines, according to the second information, that the required second neural network includes Q first neural network blocks, where Q is a positive integer greater than N, and a difference between Q and N is M, and the second information includes a capability of the second communication device and/or a requirement for processing the channel information; The second communication device sends second indication information to the first communication device, where the second indication information is used to indicate the M; The second communication device receives parameters corresponding to the M first neural network blocks from the first communication device. The method according to any one of claims 10 to 13, characterized in that the method further comprises: The second communication device receives parameters corresponding to P of the first neural network blocks from the first communication device, where P is a positive integer greater than or equal to N; The second communication device determines the N first neural network blocks from the P first neural network blocks based on second information, and the second information includes the capabilities of the second communication device and/or the need to process the channel information. The method according to any one of claims 10 to 14, characterized in that the method further comprises: The second communication device sends first information to the first communication device, where the first information includes third indication information and a first vector, where the third indication information is used to indicate N, and the first vector is a result of encoding the channel information through the first neural network. The method according to any one of claims 10 to 15, characterized in that the parameters of the first neural network block include at least one of the following: Weight information, bias information, or activation function information corresponding to the first neural network block. The method according to any one of claims 10 to 16, characterized in that the first neural network block supports at least one of the following neural network structures: Convolutional Neural Network CNN, Multi-layer Perceptron MLP, or Transformer Transformer. The method according to any one of claims 10 to 17 is characterized in that when the first neural network includes multiple first neural network blocks, the connection method between the multiple first neural network blocks includes deep connection and/or width connection. The method according to any one of claims 10 to 18, characterized in that the method further comprises: The second communication device receives parameters of the output layer of the first neural network from the first communication device. A communication device, comprising: A processor, configured to execute a computer program stored in a memory, so that the apparatus performs the method according to any one of claims 1 to 9. A communication device, comprising: A processor, configured to execute a computer program stored in the memory, so that the apparatus performs the method according to any one of claims 10 to 19. A communication system, characterized by comprising at least one communication device according to claim 20 and at least one communication device according to claim 21. A chip, characterized in that it comprises: a processor and an interface, used to call and run a computer program stored in a memory from the memory to execute the method as described in any one of claims 1 to 19. A computer-readable storage medium, characterized in that it is used to store a computer program, wherein the computer program includes instructions for implementing the method as described in any one of claims 1 to 19.