WO2023126007A1 - Channel information transmission method and apparatus - Google Patents

Abstract

Provided in the present disclosure is a channel information transmission method, used for saving communication resources. The method comprises: a terminal device sends channel feedback information to an access network device, the channel feedback information being used for indicating sparse representation information of first channel information, wherein the sparse representation information comprises M elements, the M elements comprise K non-zero elements and M-K zero elements, and M and K are positive integers; and, on the basis of the sparse representation information and a channel reconstruction model, the access network device recovers the first channel information.

Description

Channel information transmission method and device

Cross References to Related Applications

This application claims the priority of a Chinese patent application filed with the Intellectual Property Office of the People's Republic of China on December 31, 2021, with application number 202111660657.7 and application name "Channel Information Transmission Method and Device", the entire contents of which are incorporated herein by reference. Applying.

technical field

The present application relates to the technical field of communications, and in particular to a channel information transmission method and device.

Background technique

In a communication system, such as a fifth generation (5 ^th generation, 5G) mobile communication system, higher requirements are placed on system capacity and spectrum efficiency. The application of massive multiple input multiple output (MIMO) technology plays an important role in improving system capacity and spectrum efficiency. For example, using massive MIMO technology, access network equipment can provide high-quality services for more terminal equipment at the same time. In order to apply the massive MIMO technology, an important link is that the sending end precodes the data to be sent, and sends the precoded data to the receiving end. Precoding can realize spatial multiplexing (spatial multiplexing) of multiple data streams to reduce interference between different data streams, so it can improve the signal-to-interference-plus-noise ratio (SINR) at the receiving end. SINR), thereby improving the system throughput. In order to perform precoding accurately, the premise is that channel information, for example, channel state information (channel state information, CSI), etc., can be accurately obtained. Therefore, how to obtain channel information is a technical problem worthy of research.

Contents of the invention

The present application provides a channel information transmission method and device, aiming at saving communication resources.

In a first aspect, a channel information transmission method is provided. The method can be implemented on the side of the access network device, or on the side of other devices for recovering channel information, without limitation. The method includes: receiving channel feedback information from a terminal device, where the channel feedback information is used to indicate sparse representation information of the first channel information, wherein the sparse representation information includes M elements, and the M elements include K non-zero elements and M-K zero elements, wherein M and K are positive integers; determine the first channel information according to the channel reconstruction model, wherein the input of the channel reconstruction model is determined according to the sparse representation information.

Through the above-mentioned method with strong generalization ability, communication resources can be saved, that is, channel information can be transmitted more accurately in various communication scenarios through a channel reconstruction model. For example, the number and/or positions of the K non-zero elements can be changed to adapt to various communication scenarios.

In a possible design, the channel feedback information is used to indicate values of the K non-zero elements and positions of the K non-zero elements.

Through this method, signaling overhead can be saved, for example, there is no need to feed back the positions and values of M-K zero elements. Optionally, the position of feeding back K non-zero elements may be replaced by the position of feeding back M-K zero elements, and the two are equivalent.

In a possible design, the channel feedback information is used to indicate a first pattern, and the first pattern indicates the positions of the K non-zero elements, where the first pattern is one of multiple candidate patterns one.

Through this method, signaling overhead can be further saved, mainly the overhead when feeding back the positions of K non-zero elements can be saved.

In a possible design, the method further includes: determining at least one of the following according to the first channel information: a precoding matrix indicator PMI, a rank indicator RI, or a channel quality indicator CQI.

In a possible design, the channel feedback information is also used to indicate a scaling factor of the first channel information relative to the second channel information, where the first channel information is a normalized channel information.

In a possible design, the method further includes: determining at least one of the following according to the second channel information: PMI, RI, or CQI.

Through this method, the transmission parameters when the access network equipment and the terminal equipment perform MIMO communication can be obtained, so that MIMO transmission can be performed to improve the system throughput.

In a possible design, the ratio of K and N is a first compression ratio, where N is a positive integer, and N represents a dimension of the first channel information.

In a possible design, the method further includes: sending information indicating that the first compression ratio is one of multiple candidate compression ratios to the terminal device. Optionally, the multiple candidate compression ratios are stipulated in a protocol, or indicated by a signaling sent to the terminal device.

In a possible design, the method further includes: sending information indicating that K is one of multiple candidate values to the terminal device. Optionally, the multiple candidate values are stipulated in the protocol, or indicated by a signaling sent to the terminal device.

Through this method, the compression ratio of the channel information can be flexibly configured according to the requirements of the actual communication scenario, so as to meet the requirements of the communication scenario.

In a second aspect, a channel information transmission method is provided. The method may be implemented on the side of the terminal device, or on the side of other devices for feeding back channel information, without limitation. The method includes: determining sparse representation information of the first channel information according to the first channel information and the channel reconstruction model, wherein the sparse representation information includes M elements, and the M elements include K non-zero elements and M-K zero elements, where M and K are positive integers; sending channel feedback information to the access network device, where the channel feedback information is used to indicate the sparse representation information.

For the introduction of the channel feedback information, please refer to the first aspect, which will not be repeated here.

In a possible design, the determining the sparse representation information of the first channel information according to the first channel information and the channel reconstruction model includes:

The sparse representation information of the first channel information is determined according to the following objective function:

Among them, ‖x‖ ₀ ≤ K, x represents the sparse representation information of the first channel information, H _w represents the first channel information, f _de ( ) represents the channel reconstruction model, ‖ ‖ ₂ represents the L2 norm, ‖ ‖ ₀ represents L0 norm; or,

The sparse representation information of the first channel information is determined according to the following objective function:

Among them, ‖x‖ ₀ ≤ K, x represents the sparse representation information of the first channel information, H _w represents the first channel information, f _de ( ) represents the channel reconstruction model, f _W ( ) represents the precoding generation model, f _C (,) represents the channel capacity calculation model.

In a possible design, the ratio of K to N is a first compression ratio, and N represents a dimension of the first channel information.

In a possible design, the method further includes: receiving information indicating that the first compression ratio is one of multiple candidate compression ratios from the access network device. Optionally, the multiple candidate compression ratios are stipulated by a protocol, or indicated by a signaling from an access network device.

In a possible design, the method further includes: receiving information indicating that K is one of multiple candidate values from an access network device. Optionally, the multiple candidate values are stipulated in the protocol, or indicated by signaling from the access network device.

In a third aspect, an apparatus is provided for implementing the method in the first aspect. The device may be an access network device, or a device configured in the access network device, or a device that can be matched and used with the access network device. In one design, the device includes a one-to-one unit for performing the method/operation/step/action described in the first aspect, and the unit may be a hardware circuit, or software, or a combination of hardware circuit and software.

Exemplarily, the apparatus may include a processing unit and a communication unit, and the processing unit and the communication unit may perform corresponding functions in the first aspect above. For example:

The communication unit is configured to receive channel feedback information from the terminal device, where the channel feedback information is used to indicate sparse representation information of the first channel information, wherein the sparse representation information includes M elements, and the M elements include K non-zero elements and M-K zero elements, wherein M and K are positive integers; the processing unit is used to determine the first channel information according to the channel reconstruction model, wherein the input of the channel reconstruction model is according to the sparse representation Information is determined.

For the introduction of the channel feedback information, please refer to the first aspect, which will not be repeated here.

In a possible design, the processing unit is configured to determine at least one of the following according to the first channel information: a precoding matrix indicator PMI, a rank indicator RI, or a channel quality indicator CQI.

In a possible design, the channel feedback information is also used to indicate a scaling factor of the first channel information relative to the second channel information, where the first channel information is a normalized channel information.

In a possible design, the processing unit is configured to determine at least one of the following according to the second channel information: PMI, RI, or CQI.

In a possible design, the ratio of K and N is a first compression ratio, where N is a positive integer, and N represents a dimension of the first channel information.

In a possible design, the communication unit is configured to send information indicating that the first compression ratio is one of multiple candidate compression ratios to the terminal device. Optionally, the multiple candidate compression ratios are stipulated in a protocol, or indicated by a signaling sent to the terminal device.

In a possible design, the communication unit is configured to send information indicating that K is one of multiple candidate values to the terminal device. Optionally, the multiple candidate values are stipulated in the protocol, or indicated by a signaling sent to the terminal device.

Exemplarily, the above device includes a memory, configured to implement the method described in the above first aspect. The apparatus may also include memory for storing instructions and/or data. The memory is coupled to the processor, and when the processor executes the program instructions stored in the memory, the method described in the first aspect above can be implemented. The apparatus may also include a communication interface for the apparatus to communicate with other devices. Exemplarily, the communication interface may be a transceiver, circuit, bus, module, pin or other types of communication interface. In one possible design, the device includes:

memory for storing program instructions;

Communication Interface;

A processor, configured to use a communication interface to: receive channel feedback information from a terminal device, where the channel feedback information is used to indicate sparse representation information of the first channel information, wherein the sparse representation information includes M elements, and the M The elements include K non-zero elements and M-K zero elements, wherein M and K are positive integers;

The processor is configured to determine first channel information according to a channel reconstruction model, wherein an input of the channel reconstruction model is determined according to the sparse representation information.

For other implementation details, please refer to the first aspect, which will not be repeated here.

In a fourth aspect, a device is provided for implementing the method in the second aspect. The device may be a terminal device, or a device configured in the terminal device, or a device that can be matched with the terminal device. In one design, the device includes a one-to-one unit for performing the method/operation/step/action described in the second aspect, and the unit may be a hardware circuit, or software, or a combination of hardware circuit and software.

Exemplarily, the apparatus may include a processing unit and a communication unit, and the processing unit and the communication unit may perform corresponding functions in the second aspect above. For example:

The processing unit is configured to determine the sparse representation information of the first channel information according to the first channel information and the channel reconstruction model, wherein the sparse representation information includes M elements, and the M elements include K non-zero elements and M-K Among them, M and K are positive integers;

The communication unit is configured to send channel feedback information to the access network device, where the channel feedback information is used to indicate the sparse representation information.

For the introduction of the channel feedback information, please refer to the second aspect, which will not be repeated here.

In one possible design, the processing unit is used for:

The sparse representation information of the first channel information is determined according to the following objective function:

Among them, ‖x‖ ₀ ≤ K, x represents the sparse representation information of the first channel information, H _w represents the first channel information, f _de ( ) represents the channel reconstruction model, ‖ ‖ ₂ represents the L2 norm, ‖ ‖ ₀ represents L0 norm; or,

The sparse representation information of the first channel information is determined according to the following objective function:

Among them, ‖x‖ ₀ ≤ K, x represents the sparse representation information of the first channel information, H _w represents the first channel information, f _de ( ) represents the channel reconstruction model, f _W ( ) represents the precoding generation model, f _C (,) represents the channel capacity calculation model.

In a possible design, the ratio of K to N is a first compression ratio, and N represents a dimension of the first channel information.

In a possible design, the communication unit is configured to receive information indicating that the first compression ratio is one of multiple candidate compression ratios from the access network device. Optionally, the multiple candidate compression ratios are stipulated by a protocol, or indicated by a signaling from an access network device.

In a possible design, the communication unit is configured to receive information indicating that K is one of multiple candidate values from the access network device. Optionally, the multiple candidate values are stipulated in the protocol, or indicated by signaling from the access network device.

Exemplarily, the above device includes a memory, configured to implement the method described in the above second aspect. The apparatus may also include memory for storing instructions and/or data. The memory is coupled to the processor, and when the processor executes the program instructions stored in the memory, the method described in the second aspect above can be implemented. The device may also include a communication interface for the device to communicate with other devices. Exemplarily, the communication interface may be a transceiver, circuit, bus, module, pin or other types of communication interface. In one possible design, the device includes:

memory for storing program instructions;

Communication Interface;

A processor, configured to determine sparse representation information of the first channel information according to the first channel information and the channel reconstruction model, wherein the sparse representation information includes M elements, and the M elements include K non-zero elements and M-K zero elements where M and K are positive integers;

The processor uses the communication interface to: send channel feedback information to the access network device, where the channel feedback information is used to indicate the sparse representation information.

For other implementation details, please refer to the second aspect, which will not be repeated here.

In a fifth aspect, a model training method is provided, comprising: operation 1, determining a set of training data in the training data set; operation 2, for each training data in the set of training data, determining a sparse representation of the training data Information, according to the sparse representation information and the current channel reconstruction model to determine the model output corresponding to the training data; operation 3, for the set of training data, if the loss function meets the performance requirements, the training ends, otherwise, update the channel reconstruction model, And perform operation 1 again.

In a possible design, determining the sparse representation information of the training data includes: determining the sparse representation information of the training data according to a sparse representation algorithm and a current channel reconstruction model.

In a possible design, determining the sparse representation information of the training data includes: determining the sparse representation information of the training data according to the current sparse representation model, and if the loss function does not meet the performance requirements, the method further includes: updating The sparse representation model.

In a possible design, for the set of training data, the loss function meets the performance requirements, including: the average value of the loss function of all training data in the set of training data (or use each loss function of all training data through other The value calculated by the method) meets the threshold requirement, or, the loss function of all the training data in the set of training data meets the threshold requirement.

A sixth aspect provides an apparatus for realizing the method of the fifth aspect. In one design, the device includes a one-to-one unit for performing the method/operation/step/action described in the fifth aspect. The unit may be a hardware circuit, or software, or a combination of hardware circuit and software.

Exemplarily, the apparatus may include a processing unit, and the processing unit may perform corresponding functions in the fifth aspect above. For example:

The processing unit is used for: operation 1, determining a group of training data in the training data set; operation 2, for each training data in the group of training data, determining the sparse representation information of the training data, according to the sparse representation information and The channel reconstruction model determines the model output corresponding to the training data; operation 3, for the set of training data, if the loss function meets the performance requirements, the training ends, otherwise, update the channel reconstruction model, and perform operation 1 again.

Please refer to the fifth aspect for specific details, and details will not be repeated here.

Optionally, the device may also include a communication unit, configured to acquire the training data set.

A seventh aspect provides a communication system, including the device of the third aspect and the device of the fourth aspect; or, including the device of the third aspect, the device of the fourth aspect, and the device of the sixth aspect.

In an eighth aspect, there is provided a computer-readable storage medium, including instructions, which, when run on a computer, cause the computer to execute the method of the first aspect, the second aspect, or the fifth aspect.

In a ninth aspect, a computer program product is provided, including instructions, which, when run on a computer, cause the computer to execute the method of the first aspect, the second aspect, or the fifth aspect.

In a tenth aspect, a chip system is provided, and the chip system includes a processor, and may further include a memory, for implementing the method of the first aspect, the second aspect, or the fifth aspect. The system-on-a-chip may consist of chips, or may include chips and other discrete devices.

Description of drawings

Figure 1 shows an example diagram of the architecture of the communication system;

Figure 2A shows an example diagram of the structure of a neuron;

Figure 2B shows a structural example diagram of a neural network;

3A to 3E are schematic diagrams of the network architecture;

FIG. 4 and FIG. 5 are schematic flowcharts of channel information transmission methods;

FIG. 6A, FIG. 6B and FIG. 7 are schematic flow charts of model training;

Figure 8 and Figure 9 are diagrams showing an example of the structure of the device.

Detailed ways

FIG. 1 is an example diagram of the architecture of a communication system 1000 to which the present disclosure can be applied. As shown in FIG. 1 , the communication system includes a radio access network (radio access network, RAN) 100 and a core network (core network, CN) 200. Optionally, the communication system 1000 may also include the Internet 300 . The radio access network 100 may include at least one access network device (or may be called RAN device, such as 110a and 110b in FIG. 1 ), and may also include at least one terminal device (such as 120a-120j in FIG. 1). The terminal device is connected to the access network device in a wireless manner. Access network devices are connected to core network devices in a wireless or wired manner. The core network device and the access network device can be independent and different physical devices; or can be the same physical device that integrates the functions of the core network device and the access network device; or can be other possible situations, such as a The function of the access network device and some functions of the core network device can be integrated on the physical device, and another physical device realizes the rest of the functions of the core network device. The present disclosure does not limit the physical existence form of the core network device and the access network device. Terminal devices may be connected to each other in a wired or wireless manner. The access network device and the access network device may be connected to each other in a wired or wireless manner. FIG. 1 is only a schematic diagram, and is not intended to limit the present disclosure. For example, the communication system may also include other network devices, such as wireless relay devices and wireless backhaul devices.

The core network 200 may include one or more core network elements. For example, the core network may include at least one of the following network elements: access and mobility management function (access and mobility management function, AMF) network element, session management function (session management function, SMF) network element, user plane function (user plane function (UPF) network element, policy control function (policy control function, PCF) network element, unified data management (unified data management, UDM) network element, application function (application function, AF) network element, or location management function ( location management function, LMF) network element, etc. These core network elements may be a hardware structure, a software module, or a hardware structure plus a software module. The implementation forms of different network elements may be the same or different, and are not limited. Different core network elements may be different physical devices (or may be called core network devices), or multiple different core network elements may be integrated on one physical device, that is, the physical device has the multiple core network elements function.

In the present disclosure, the device used to realize the function of the core network device may be a core network device; it may also be a device capable of supporting the core network device to realize the function, such as a chip system, a hardware circuit, a software module, or a hardware circuit plus a software module , the device can be installed in the core network equipment or can be matched with the core network equipment. In the present disclosure, the technical solution provided by the present disclosure is described by taking the core network device as an example for realizing the functions of the core network device. In the present disclosure, a system-on-a-chip may be composed of chips, and may also include chips and other discrete devices.

A terminal device may also be called a terminal, a user equipment (user equipment, UE), a mobile station, or a mobile terminal, etc. Terminal devices can be widely used in various scenarios for communication. For example, the scenario includes but is not limited to at least one of the following: enhanced mobile broadband (enhanced mobile broadband, eMBB), ultra-reliable low-latency communication (ultra-reliable low-latency communication, URLLC), large-scale machine type communication ( massive machine-type communications (mMTC), device-to-device (D2D), vehicle to everything (V2X), machine-type communication (MTC), Internet of things (internet of things) , IOT), virtual reality, augmented reality, industrial control, automatic driving, telemedicine, smart grid, smart furniture, smart office, smart wear, smart transportation, or smart city, etc. The terminal device can be a mobile phone, a tablet computer, a computer with wireless transceiver function, a wearable device, a vehicle, a drone, a helicopter, an airplane, a ship, a robot, a mechanical arm, or a smart home device, etc. The present disclosure does not limit the specific technology and specific device form adopted by the terminal device.

In the present disclosure, the device for realizing the function of the terminal device may be a terminal device; it may also be a device capable of supporting the terminal device to realize the function, such as a chip system, a hardware circuit, a software module, or a hardware circuit plus a software module. It can be installed in the terminal equipment or can be matched with the terminal equipment. For ease of description, the technical solution provided is described below by taking the terminal device as an example of the apparatus for realizing the functions of the terminal device, and optionally taking the terminal device as an example.

The access network device can be a base station (base station), Node B (Node B), evolved Node B (evolved NodeB, eNodeB or eNB), transmission reception point (transmission reception point, TRP), fifth generation (5 ^th generation , 5G) mobile communication system next generation Node B (next generation NodeB, gNB), open radio access network (open radio access network, O-RAN or open RAN) in the access network equipment, the sixth generation (6 ^th generation, 6G) mobile communication system next-generation base station, wireless fidelity (wireless fidelity, WiFi) system access node, or future mobile communication system base station, etc. Optionally, the access network device may be a module or unit that completes some functions of the access network device, for example, it may be a centralized unit (central unit, CU), a distributed unit (distributed unit, DU), a centralized unit control plane (CU control plane, CU-CP) module, centralized unit user plane (CU user plane, CU-UP) module, or radio unit (radio unit, RU). The access network device may be a macro base station (such as 110a in Figure 1), a micro base station or an indoor station (such as 110b in Figure 1), or a relay node or a donor node. The specific technology and specific device form adopted by the access network device are not limited in the present disclosure. Among them, 5G can also be called new radio (new radio, NR).

In the present disclosure, the device for implementing the function of the access network device may be the access network device; or, it may be a device capable of supporting the access network device to realize the function, such as a chip system, a hardware circuit, a software module, or a hardware circuit Adding software modules, etc., the device can be installed in the access network equipment or matched with the access network equipment. For ease of description, the technical solution provided is described below by taking the access network device as an example of the apparatus for realizing the function of the access network device, and optionally taking the access network device as an example as a base station.

The communication between the access network device and the terminal device may follow a certain protocol layer structure. Exemplarily, the protocol layer structure may include a control plane protocol layer structure and a user plane protocol layer structure. For example, the control plane protocol layer structure may include at least one of the following: a radio resource control (radio resource control, RRC) layer, a packet data convergence protocol (packet data convergence protocol, PDCP) layer, a radio link control (radio link control, RLC) layer, media access control (media access control, MAC) layer, or physical (physical, PHY) layer, etc. For example, the user plane protocol layer structure may include at least one of the following: a service data adaptation protocol (service data adaptation protocol, SDAP) layer, a PDCP layer, an RLC layer, a MAC layer, or a physical layer.

The above protocol layer structure between the access network device and the terminal device can be regarded as an access stratum (access stratum, AS) structure. Optionally, on top of the AS, there may also be a non-access stratum (non-access stratum, NAS), which is used for the access network device to forward information from the core network device to the terminal device, or for the access network device to The core network device forwards the information from the terminal device. At this point, it can be considered that there is a logical interface between the terminal device and the core network device. Optionally, the access network device may forward information between the terminal device and the core network device through transparent transmission. For example, the NAS message may be mapped to or included in RRC signaling as an element of RRC signaling.

Optionally, the protocol layer structure between the access network device and the terminal device may further include an artificial intelligence (AI) layer, which is used to transmit data related to the AI function.

Access network devices may include CUs and DUs. This design can be called CU and DU separation. Multiple DUs can be centrally controlled by one CU. As an example, the interface between CU and DU is called F1 interface. Wherein, the control plane (control panel, CP) interface may be F1-C, and the user plane (user panel, UP) interface may be F1-U. The present disclosure does not limit the specific names of the interfaces. CU and DU can be divided according to the protocol layer of the wireless network: for example, the functions of the PDCP layer and above protocol layers (such as RRC layer and SDAP layer, etc.) etc.) functions are set in the DU; for another example, the functions of the protocol layers above the PDCP layer are set in the CU, and the functions of the PDCP layer and the protocol layers below are set in the DU, without restriction.

The foregoing division of the processing functions of the CU and DU according to the protocol layer is only an example, and may also be divided in other ways. For example, the CU or DU may be divided into functions having more protocol layers, and for example, the CU or DU may be divided into part processing functions having protocol layers. For example, some functions of the RLC layer and functions of the protocol layers above the RLC layer are set in the CU, and the remaining functions of the RLC layer and functions of the protocol layers below the RLC layer are set in the DU. For another example, the functions of the CU or DU can be divided according to the service type or other system requirements, for example, according to the delay, the functions that need to meet the delay requirement are set in the DU, and the functions that do not need to meet the delay requirement are set in the CU.

Optionally, the CU may have one or more functions of the core network. The CU can be set on the network side to facilitate centralized management.

Optionally, the wireless unit (radio unit, RU) of the DU is remotely set. Wherein, the RU has a radio frequency function. Exemplarily, DUs and RUs can be divided at the PHY layer. For example, the DU can implement high-level functions in the PHY layer, and the RU can implement low-level functions in the PHY layer. Wherein, when used for sending, the functions of the PHY layer may include at least one of the following: adding a cyclic redundancy check (cyclic redundancy check, CRC) bit, channel coding, rate matching, scrambling, modulation, layer mapping, precoding, Resource mapping, physical antenna mapping, or radio frequency transmission functions. When used for reception, the functions of the PHY layer may include at least one of the following: CRC check, channel decoding, de-rate matching, descrambling, demodulation, de-layer mapping, channel detection, resource de-mapping, physical antenna de-mapping, or RF receiving function. Wherein, the high-level functions in the PHY layer may include part of the functions of the PHY layer, which are closer to the MAC layer; the lower-level functions in the PHY layer may include another part of the functions of the PHY layer, for example, this part of functions is closer to the radio frequency function. For example, high-level functions in the PHY layer may include adding CRC bits, channel coding, rate matching, scrambling, modulation, and layer mapping, and low-level functions in the PHY layer may include precoding, resource mapping, physical antenna mapping, and radio transmission functions; alternatively, high-level functions in the PHY layer can include adding CRC bits, channel coding, rate matching, scrambling, modulation, layer mapping, and precoding, and low-level functions in the PHY layer can include resource mapping, physical antenna mapping, and radio frequency send function. For example, the high-level functions in the PHY layer may include CRC check, channel decoding, de-rate matching, decoding, demodulation, and de-mapping, and the low-level functions in the PHY layer may include channel detection, resource de-mapping, physical antenna de-mapping, and RF receiving functions; or, the high-level functions in the PHY layer may include CRC check, channel decoding, de-rate matching, decoding, demodulation, de-layer mapping, and channel detection, and the low-level functions in the PHY layer may include resource de-mapping , physical antenna demapping, and RF receiving functions.

Optionally, the functions of the CU may be further divided, and the control plane and the user plane may be separated and implemented by different entities. The separated entities are the control plane CU entity (ie, CU-CP entity) and the user plane CU entity (ie, CU-UP entity). The CU-CP entity and the CU-UP entity can be connected to the DU respectively. In the present disclosure, an entity may be understood as a module or unit, and its existence form may be a hardware structure, a software module, or a hardware structure plus a software module, without limitation.

Optionally, any one of the foregoing CU, CU-CP, CU-UP, DU, and RU may be a software module, a hardware structure, or a software module plus a hardware structure, without limitation. Wherein, the existence forms of different entities may be the same or different. For example, CU, CU-CP, CU-UP and DU are software modules, and RU is a hardware structure. For the sake of brevity, all possible combinations are not listed here. These modules and the methods performed by them are also within the protection scope of the present disclosure. For example, when the method of the present disclosure is executed by an access network device, it may specifically be executed by at least one of CU, CU-CP, CU-UP, DU, RU, or near real-time RIC described below.

In the present disclosure, the access network device and/or the terminal device may be fixed or mobile. Access network equipment and/or terminal equipment can be deployed on land, including indoors or outdoors, hand-held or vehicle-mounted; or can be deployed on water; or can be deployed on aircraft, balloons and artificial satellites in the air. The present disclosure does not limit the environment/scene where the access network device and the terminal device are located. Access network devices and terminal devices can be deployed in the same or different environments/scenarios, for example, access network devices and terminal devices are deployed on land at the same time; or, access network devices are deployed on land and terminal devices are deployed on water First class, no more examples one by one.

The roles of the access network device and the terminal device may be relative. For example, the helicopter or drone 120i in FIG. 1 can be configured as a mobile access network device. For those terminal devices 120j that access the wireless access network 100 through 120i, the terminal device 120i is an access network device. For the access network device 110a, 120i may be a terminal device, that is, communication between 110a and 120i may be performed through a wireless access network air interface protocol. Alternatively, 110a and 120i communicate through an interface protocol between access network devices. At this time, relative to 110a, 120i is also an access network device. Therefore, both the access network device and the terminal device can be collectively referred to as a communication device (or communication device), 110a and 110b in FIG. It is called a communication device with terminal equipment function.

Between an access network device and a terminal device, between an access network device and an access network device, or between a terminal device and a terminal device: communication may be performed through licensed spectrum, or communication may be performed through unlicensed spectrum, or both Communicate over licensed spectrum and unlicensed spectrum; and/or, may communicate over spectrum below 6 gigahertz (GHz), or may communicate over spectrum above 6 GHz, or may use both spectrum below 6 GHz and 6 GHz above the frequency spectrum for communication. The present disclosure does not limit spectrum resources used by wireless communications.

In a communication system, if the data sending end can know the channel information of the channel between the data sending end and the data receiving end, the data transmission efficiency can be improved. For example, in a MIMO-based system, if the data sender can obtain the channel information, it can obtain transmission parameters such as precoding matrix, and can use the precoding matrix to precode the data to be sent, so that the data sender can pass the same The resource (for example, the same time-frequency resource) sends multiple space-division multiplexed data streams to the same data receiving end, and/or, sends data to multiple data receiving ends through the same resource. In a possible implementation, the channel information can be estimated by the data receiving end, and the channel information can be sent to the data sending end; the data sending end determines the precoding matrix based on the channel information, and uses the precoding matrix to precode the data to be sent. And send the precoded data to the data receiving end. In the present disclosure, estimated channel information may also be described as measured channel information or other names, without limitation. Optionally, the data receiving end is a terminal device, and the data sending end is an access network device; or, the data receiving end is an access network device, and the data sending end is a terminal device; or, the data receiving end is a first access network device , the data sending end is the second access network device, which is not limited. For ease of understanding, the following descriptions will be made by taking the data receiving end as a terminal device and the data sending end as an access network device as an example.

In order to improve the feedback efficiency of channel information and enable the system to realize the feedback of channel information intelligently, the present disclosure introduces artificial intelligence (AI) technology into the communication system. In this method, the feedback of channel information is realized or assisted by AI technology, and the fed back channel information can better match the actual channel environment, so the fed back channel information is more accurate.

Artificial intelligence can make machines have human intelligence, for example, it can make machines use computer software and hardware to simulate certain intelligent behaviors of humans. To achieve artificial intelligence, machine learning methods can be employed. In the machine learning method, the machine uses the training data to learn (or train) to obtain a model, and applies the model to reason (or predict). Inference results can be used to solve practical problems. Machine learning methods include but are not limited to at least one of the following: neural network (neural network, NN), probabilistic graphical model, sparse coding/dictionary learning method, variational auto-encoder (variational auto-encoder, VAE), or generate confrontation Networks (generative adversarial networks, GAN), etc., are not limited.

A neural network is a concrete implementation of machine learning techniques and AI models. According to the general approximation theorem, the neural network can theoretically approximate any continuous function, so that the neural network has the ability to learn any mapping. Traditional communication systems need to rely on rich expert knowledge to design communication modules, while deep learning communication systems based on neural networks can automatically discover hidden pattern structures from a large number of data sets, establish mapping relationships between data, and achieve better results than traditional communication systems. The performance of the modeling method.

再例如，一个神经元的激活函数为：y＝f(z)＝z，则该神经元的输出为：

其中，b可以是小数、整数(例如0、正整数或负整数)、或复数等各种可能的类型。神经网络中不同神经元的激活函数可以相同或不同。 The idea of a neural network is derived from the neuronal structure of brain tissue. For example, each neuron performs a weighted sum operation on its input values, and outputs the operation result through an activation function. As shown in FIG. 2A , it is a schematic diagram of a neuron structure. Suppose the input of the neuron is x=[x ₀ ,x ₁ ,…,x _n ], and the weights corresponding to each input are w=[w,w ₁ ,…,w _n ], where n is a positive integer , w _i and _xi may be various possible types such as decimals, integers (such as 0, positive or negative integers, etc.), or complex numbers. w _i is used as the weight of _xi to weight _xi . The bias for performing weighted summation of the input values according to the weights is, for example, b. There are many forms of the activation function. Assuming that the activation function of a neuron is: y=f(z)=max(0,z), the output of the neuron is:

For another example, the activation function of a neuron is: y=f(z)=z, then the output of the neuron is:

Wherein, b may be various possible types such as decimals, integers (such as 0, positive integers or negative integers), or complex numbers. The activation functions of different neurons in a neural network can be the same or different.

A neural network generally includes multiple layers, each layer may include one or more neurons. By increasing the depth and/or width of the neural network, the expressive ability of the neural network can be improved, providing more powerful information extraction and abstract modeling capabilities for complex systems. Wherein, the depth of the neural network may refer to the number of layers included in the neural network, and the number of neurons included in each layer may be referred to as the width of the layer. In one implementation, a neural network includes an input layer and an output layer. The input layer of the neural network processes the received input information through neurons, and passes the processing result to the output layer, and the output layer obtains the output result of the neural network. In another implementation manner, the neural network includes an input layer, a hidden layer and an output layer, refer to FIG. 2B . The input layer of the neural network processes the received input information through neurons, and passes the processing results to the middle hidden layer. The hidden layer calculates the received processing results to obtain the calculation results, and the hidden layer transmits the calculation results to the output layer or The next adjacent hidden layer finally gets the output of the neural network from the output layer. Wherein, a neural network may include one hidden layer, or include multiple hidden layers connected in sequence, without limitation.

The neural network involved in the present disclosure is, for example, a deep neural network (DNN). Depending on how the network is constructed, DNNs can include feedforward neural networks (FNN), convolutional neural networks (CNN) and recurrent neural networks (RNN). The type of the model involved in the present disclosure may be DNN, for example, FNN, CNN or RNN, without limitation.

During model training, a loss function can be defined. The loss function describes the gap or difference between the output value of the model and the ideal target value. The present disclosure does not limit the specific form of the loss function. The model training process can be regarded as the following process: by adjusting some or all parameters of the model, the value of the loss function is less than the threshold value or meets the target requirements.

A model may also be called an AI model, a rule, or other names without limitation. The AI model can be considered as a specific method to realize the AI function. The AI model represents the mapping relationship or function between the input and output of the model. AI functions may include at least one of the following: data collection, model training (or model learning), model information release, model inference (or called model inference, inference, or prediction, etc.), model monitoring or model verification, or inference results release etc. AI functions may also be referred to as AI (related) operations, or AI-related functions.

In the present disclosure, an independent network element (such as AI network element, AI node, or AI device, etc.) may be introduced into the aforementioned communication system shown in FIG. 1 to implement some or all AI-related operations. The AI network element can be directly connected to the access network device, or can be indirectly connected through a third-party network element and the access network device. Optionally, the third-party network element may be a core network element. Alternatively, AI entities may be configured or set in other network elements in the communication system to implement AI-related operations. Wherein, the AI entity may also be called an AI module, an AI unit or other names, and is mainly used to realize some or all AI functions, and the disclosure does not limit its specific name. Optionally, the other network element may be an access network device, a core network device, a cloud server, or a network management (operation, administration and maintenance, OAM), etc. In this case, the network element performing AI-related operations is a network element with a built-in AI function. Since both AI network elements and AI entities implement AI-related functions, for the convenience of description, the AI network elements and network elements with built-in AI functions are collectively described as AI function network elements.

In this disclosure, OAM is used to operate, manage and/or maintain core network equipment (network management of core network equipment), and/or is used to operate, manage and/or maintain access network equipment (network management of access network equipment) . For example, the present disclosure includes a first OAM and a second OAM, the first OAM is the network management of the core network equipment, and the second OAM is the network management of the access network equipment. Optionally, the first OAM and/or the second OAM includes an AI entity. For another example, the present disclosure includes a third OAM, and the third OAM is the network manager of the core network device and the access network device at the same time. Optionally, the AI entity is included in the third OAM.

Optionally, in order to match and support AI functions, an AI entity may be integrated in a terminal or a terminal chip.

FIG. 3A is an example diagram of an application framework of AI in a communication system. In Figure 3A, the data source is used to store training data and inference data. The model training host obtains the AI model by training or updating the training data provided by the data source, and deploys the AI model in the model inference host. Among them, the AI model represents the mapping relationship between the input and output of the model. Learning the AI model through the model training node is equivalent to using the training data to learn the mapping relationship between the input and output of the model. The model inference node uses the AI model to perform inference based on the inference data provided by the data source, and obtains the inference result. This method can be realized as follows: the model inference node inputs the inference data into the AI model, and obtains an output through the AI model, and the output is the inference result. The inference result may indicate: configuration parameters used (executed) by the execution object, and/or operations performed by the execution object. The reasoning result can be uniformly planned by the execution (actor) entity, and sent to one or more execution objects (for example, a network element of the core network, a base station, or a UE, etc.) for execution. Optionally, the model reasoning node can feed back its reasoning results to the model training node. This process can be called model feedback. The fed back reasoning results are used for the model training node to update the AI model, and the updated AI model is deployed on the model Inference node. Optionally, the execution object can feed back the collected network parameters to the data source. This process can be called performance feedback, and the fed back network parameters can be used as training data or inference data.

In the present disclosure, the application framework shown in FIG. 3A can be deployed on the network element shown in FIG. 1 . For example, the application framework in FIG. 3A may be deployed on at least one of the terminal device, access network device, core network device, or independently deployed AI network element (not shown) in FIG. 1 . For example, the AI network element (which can be regarded as a model training node) can analyze or train the training data (training data) provided by the terminal device and/or the access network device to obtain a model. At least one of the terminal device, the access network device, or the core network device (which can be regarded as a model reasoning node) can use the model and reasoning data to perform reasoning and obtain the output of the model. Wherein, the reasoning data may be provided by the terminal device and/or the access network device. The input of the model includes inference data, and the output of the model is the inference result corresponding to the model. At least one of the terminal device, the access network device, or the core network device (which can be regarded as an execution object) can perform a corresponding operation according to the reasoning result. Wherein, the model inference node and the execution object may be the same or different, without limitation.

The network architecture to which the method provided in the present disclosure can be applied is introduced as an example below with reference to FIGS. 3B to 3E .

As shown in FIG. 3B , in the first possible implementation, the access network device includes a near real-time access network intelligent controller (RAN intelligent controller, RIC) module for model training and reasoning. For example, near real-time RIC can be used to train an AI model and use that AI model for inference. For example, the near real-time RIC can obtain network-side and/or terminal-side information from at least one of CU, DU, RU or terminal equipment, and the information can be used as training data or inference data. Optionally, the near real-time RIC may submit the reasoning result to at least one of CU, DU, RU or terminal device. Optionally, the inference results can be exchanged between the CU and the DU. Optionally, the reasoning results can be exchanged between the DU and the RU, for example, the near real-time RIC submits the reasoning result to the DU, and the DU forwards it to the RU.

As shown in FIG. 3B, in the second possible implementation, a non-real-time RIC is included outside the access network (optionally, the non-real-time RIC can be located in the OAM, in the cloud server, or in the core network device) for performing Model training and inference. For example, non-real-time RIC is used to train an AI model and use that model for inference. For example, the non-real-time RIC can obtain network-side and/or terminal-side information from at least one of CU, DU, RU, or terminal equipment. This information can be used as training data or inference data, and the inference results can be submitted to CU, RU, or terminal equipment. At least one of DU, RU, or terminal equipment. Optionally, the inference results can be exchanged between the CU and the DU. Optionally, the reasoning results can be exchanged between the DU and the RU, for example, the non-real-time RIC submits the reasoning result to the DU, and the DU forwards it to the RU.

As shown in Figure 3B, in the third possible implementation, the access network device includes a near real-time RIC, and the access network device includes a non-real-time RIC (optionally, the non-real-time RIC can be located in the OAM, cloud server in, or in core network equipment). Like the second possible implementation above, non-real-time RIC can be used for model training and inference. And/or, like the above first possible implementation, the near real-time RIC can be used for model training and reasoning. And/or, the non-real-time RIC performs model training, and the near-real-time RIC can obtain AI model information from the non-real-time RIC, and obtain network-side and/or terminal-side information from at least one of CU, DU, RU, or terminal equipment , using the information and the AI model information to obtain an inference result. Optionally, the near real-time RIC may submit the reasoning result to at least one of CU, DU, RU or terminal device. Optionally, the inference results can be exchanged between the CU and the DU. Optionally, the reasoning results can be exchanged between the DU and the RU, for example, the near real-time RIC submits the reasoning result to the DU, and the DU forwards it to the RU. For example, near real-time RIC is used to train model A and use model A for inference. For example, non-real-time RIC is used to train Model B and utilize Model B for inference. For example, the non-real-time RIC is used to train the model C, and the information of the model C is sent to the near-real-time RIC, and the near-real-time RIC uses the model C for inference.

FIG. 3C is an example diagram of a network architecture to which the method provided in the present disclosure can be applied. Compared with FIG. 3B , in FIG. 3B CU is separated into CU-CP and CU-UP.

FIG. 3D is an example diagram of a network architecture to which the method provided by the present disclosure can be applied. As shown in FIG. 3D , optionally, the access network device includes one or more AI entities, and the functions of the AI entities are similar to the near real-time RIC described above. Optionally, the OAM includes one or more AI entities, and the functions of the AI entities are similar to the non-real-time RIC described above. Optionally, the core network device includes one or more AI entities, and the functions of the AI entities are similar to the above-mentioned non-real-time RIC. When both the OAM and the core network equipment include AI entities, the models trained by their respective AI entities are different, and/or the models used for reasoning are different.

In the present disclosure, the different models include at least one of the following differences: the structural parameters of the model (such as the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of neurons, the activation function of neurons, or the at least one of the bias), the input parameters of the model (such as the type of the input parameter and/or the dimension of the input parameter), or the output parameters of the model (such as the type of the output parameter and/or the dimension of the output parameter).

FIG. 3E is an example diagram of a network architecture to which the method provided in the present disclosure can be applied. Compared with Fig. 3D, the access network devices in Fig. 3E are separated into CU and DU. Optionally, the CU may include an AI entity, and the function of the AI entity is similar to the above-mentioned near real-time RIC. Optionally, the DU may include an AI entity, and the function of the AI entity is similar to the above-mentioned near real-time RIC. When both the CU and the DU include AI entities, the models trained by their respective AI entities are different, and/or the models used for reasoning are different. Optionally, the CU in FIG. 3E may be further split into CU-CP and CU-UP. Optionally, one or more AI models may be deployed in the CU-CP. Optionally, one or more AI models can be deployed in CU-UP.

In FIG. 3D or FIG. 3E , the OAM of the access network device and the OAM of the core network device are shown as unified deployment. Alternatively, as described above, in FIG. 3D or FIG. 3E , the OAM of the access network device and the OAM of the core network device may be deployed separately and independently.

In the present disclosure, a model can be inferred to obtain an output, and the output includes one parameter or multiple parameters. The learning process or training process of different models can be deployed in different devices or nodes, or can be deployed in the same device or node. Inference processes of different models can be deployed in different devices or nodes, or can be deployed in the same device or node. This disclosure is not limited to these implementations.

In the present disclosure, the involved network element may perform some or all of the steps or operations related to the network element. These steps or operations are just examples, and the present disclosure may also perform other operations or modifications of various operations. In addition, various steps may be performed in a different order than presented in the disclosure, and it may not be necessary to perform all operations in the disclosure.

In each example of the present disclosure, if there is no special explanation and logical conflict, terms and/or descriptions between different examples can be referred to each other, and technical features in different examples can be combined to form a new example according to their inherent logical relationship .

In the present disclosure, at least one (item) can also be described as one (item) or multiple (items), and multiple (items) can be two (items), three (items), four (items) or more Multiple (items), without limitation. "/" can indicate that the associated objects are an "or" relationship, for example, A/B can indicate A or B; "and/or" can be used to describe that there are three relationships between associated objects, for example, A and / Or B, can mean: A alone exists, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. To facilitate the description of the technical solution of the present disclosure, words such as "first", "second", "A", or "B" may be used to distinguish technical features with the same or similar functions. The words "first", "second", "A", or "B" do not limit the quantity and execution order. Moreover, words such as "first", "second", "A", or "B" are not necessarily different. Words such as "exemplary" or "such as" are used to indicate examples, illustrations or illustrations, and any design described as "exemplary" or "such as" should not be construed as being more preferred or better than other design solutions. Advantage. The use of words such as "exemplary" or "for example" is intended to present related concepts in a specific manner for easy understanding.

The network architecture and business scenarios described in this disclosure are to illustrate the technical solution of this disclosure more clearly, and do not constitute a limitation to the technical solution provided by this disclosure. Those skilled in the art know that with the evolution of network architecture and new business The technical solutions provided in this disclosure are also applicable to similar technical problems when the scene arises.

In order to implement or assist in implementing channel information feedback by using AI technology, the present disclosure provides a method for channel information feedback, which can save communication resources. In this method, the terminal device determines the sparse representation information of the first channel information by using the channel reconstruction model. The sparse representation information includes M elements, and the M elements include K non-zero elements and M-K zero elements. Wherein, K is an integer greater than or equal to 1, and M is an integer greater than or equal to K. The terminal device indicates the sparse representation information to the access network device through channel feedback information. The access network device can restore (or describe as: reconstruct) the first channel information by using the sparse representation information and the channel reconstruction model.

Through the above method, resources of the communication system can be saved. Due to the complex and changeable environment in the communication system, the transmission requirements may also be varied. In order to adapt to the diversity of communication systems, a possible approach is to design a transmission parameter of channel information for each transmission requirement and each channel environment. For example, for each transmission requirement and each channel environment, design a set of matching channel information encoder and channel information decoder. Wherein, the channel information encoder and the channel information decoder may be AI models. The terminal equipment uses the channel information encoder to encode the channel information to obtain encoded information. The terminal device sends the encoded information to the access network device. The access network device decodes the coded information and the channel information decoder to obtain the channel information. The disadvantage of this method is that in order to adapt to the diversity of communication systems, multiple sets of encoders and decoders need to be designed. However, in the method provided in the present disclosure, the terminal device sends sparse representation information of channel information to the access network device, the sparse representation information includes zero elements and non-zero elements, and the number of zero elements and non-zero elements and/or Locations can be designed to accommodate a wide variety of possible communication needs. For example, for different channel environments, the number and/or positions of non-zero elements may be independently set to meet communication requirements in different channel environments. Through this method, the design can be simplified, for example, multiple requirements can be met through one signaling. The sparse representation information is obtained by the terminal device side under the condition of knowing the channel reconstruction model, so that the access network device side can use the sparse representation information and the channel reconstruction model to recover more accurate original channel information. Therefore, the above method can save communication resources and has strong generalization ability, that is, through a channel reconstruction model, channel information can be transmitted more accurately in various communication scenarios.

Optionally, the channel reconfiguration model may also be called: a channel restoration model, a channel solution model or other names, without limitation.

Optionally, the structure and/or parameters of the channel reconfiguration model used by the terminal device side and the channel reconfiguration model used by the access network device side may be different. For example, the channel reconfiguration model used by the terminal device side is the first channel reconfiguration model, and the channel reconfiguration model used by the access network device side is the second channel reconfiguration model. Although the specific parameters of the first channel reconstruction model and the second channel reconstruction model can be different, the input dimensions of the two are the same; the output dimensions are the same; and, given the same input, the outputs of the two are the same or approximately the same, For example, the difference between the two outputs is smaller than the threshold value. The purpose of this design is that when the processing capability of the terminal equipment is limited, a channel reconstruction model that has almost the same function as that of the access network equipment side or the output error is within the allowable range can be deployed on the terminal equipment side, but the model structure is simpler. Save processing resources of terminal equipment. For ease of description, the description below takes the first channel reconfiguration model and the second channel reconfiguration model as the same reconfiguration model as an example.

FIG. 4 is a schematic diagram of a channel information feedback method provided by the present disclosure, and the method includes operations S401 to S403.

In operation S401, the terminal device determines sparse representation information of the first channel information according to the first channel information and the channel reconstruction model. Wherein, the sparse representation information includes M elements, the M elements include K non-zero elements and M-K zero elements, K is an integer greater than or equal to 1, and M is an integer greater than or equal to K.

The first channel information is information about a channel between the access network device and the terminal device. The first channel information may be channel information corresponding to one or more time units. Wherein, the time unit may be a symbol, a time slot, a subframe or other possible time units, without limitation. The present disclosure does not limit the type and acquisition manner of the first channel information, for example, the information may be time domain information or frequency domain information, without limitation. For example, the first channel information is the downlink channel response estimated by the terminal device, or the information obtained after preprocessing the downlink channel response, which is not limited. Wherein, the downlink channel response may also be called a downlink channel matrix or other names, without limitation. Optionally, as shown in FIG. 5 , the preprocessing includes at least one of the following operations: channel whitening, channel normalization, or quantization. The present disclosure does not exclude that the preprocessing may also include other possible operations.

For example, the access network device sends a downlink reference signal, such as a downlink synchronization signal or a channel state information reference signal (channel state information reference signal, CSI-RS), to the terminal device. The terminal device knows the sequence value of the downlink reference signal, for example, the sequence value is stipulated in the protocol or the access network device notifies the terminal device in advance, then the terminal device can estimate (measure) the downlink channel response based on the received downlink reference signal H. For example, H is the frequency domain channel response, then H satisfies: Y=H×S+n ₀ , where S is the transmitted downlink reference signal, Y is the received signal, and n ₀ represents noise. Optionally, in a communication system based on Orthogonal Frequency Division Multiplexing (OFDM), H can be expressed as a 3-dimensional matrix, for example, the dimension of H is N _C ×N _Tx ×N _Rx , where, the first The length of the first dimension is equal to the frequency domain bandwidth, for example, equal to the number N _C of frequency domain subcarriers, the length of the second dimension is equal to the number N _Tx of antenna ports at the transmitting end, and the length of the third dimension is equal to the number N _Rx of antenna ports at the receiving end. Wherein, N _C , N _Tx and _NRx are integers. Optionally, the order of these three dimensions can be exchanged. For example, the length of the first dimension of H is equal to N _Tx , the length of the second dimension is equal to N _Rx , and the length of the third dimension is equal to N _C . Hereinafter, the dimension of H is N _C ×N _Tx ×N _Rx as an example for illustration. Record the elements in H as h _i,j,z , h _i,j,z are complex numbers, i takes a value from 0 to N _C -1, j takes a value from 0 to N _Tx -1, and z takes a value of 0 to N _Rx -1. h _{i, j, z} represent the channel response of the channel between antenna port j at the transmitting end and antenna port z at the receiving end on subcarrier i.

Optionally, the first channel information is H.

Optionally, the first channel information is H1, and H1 is a matrix obtained by whitening H (which may be called the second channel information) by using the interference noise covariance matrix Ruu. The method can be described as: the first channel information is the whitened channel information of the second channel information.

For example, the access network device sends a zero power channel state information reference signal (ZP CSI-RS) to the terminal device, and the terminal device can estimate (measure) the interference and noise based on the received reference signal , denoted as I. I contains a plurality of subcarrier information, where the dimension of I(k) of the kth subcarrier is N _Rx ×1, and the covariance matrix of interference and noise can be obtained as follows:

R _uu is the covariance matrix of interference and noise, I ^H (k) represents the conjugate transpose of I(k), and its dimension is N _Rx ×N _Rx . From it, a whitening matrix P can be generated with a dimension of N _Rx ×N _Rx , where P satisfies:

或者，P ^H×P＝R _uu，

Alternatively, P ^H × P = R _uu ,

where ^PH represents the conjugate transpose of P.

For the kth subcarrier, the channel information matrix is multiplied by the whitening matrix to complete the channel whitening:

H _whiten (k) = P × H (k)

The dimension after channel whitening remains unchanged, and the dimension of H _whiten (k) is still N _Rx ×N _Tx . Combining the H _whiten (k) of each subcarrier can obtain the above second channel information H1, and the dimension of H1 is N _C ×N _Tx ×N _Rx .

Optionally, the first channel information is H2, and H2 is a matrix obtained after normalizing H or H1. After normalization, the scaling factor and H2 can be obtained. Wherein, the scaling factor can also be referred to as a signal-to-noise ratio (signal-to-noise, SNR), and the matrix before normalization (which can be referred to as the second channel information, for example, the second channel information is H or H1) is divided by the scaling factor Equal to H2, indicating that the scaling factor is the scaling factor of the second channel information relative to the first channel information, or the matrix before normalization (which can be called the second channel information, for example, the second channel information is H or H1) multiplied by The scaling factor is equal to H2, indicating that the scaling factor is a scaling factor of the first channel information relative to the second channel information. After normalization, the values of the real and imaginary parts of H2 can be located in the interval [0,1]. The value of the scaling factor may be a decimal or an integer, for example, a number less than 1 or a number greater than or equal to 1, without limitation. The method can be described as: the first channel information is the normalized channel information of the second channel information. In this method, the terminal device may also send the scaling factor to the access network device, for example, send the scaling factor to the access network device through channel feedback information in S402 below. As shown in FIG. 5 , the access network device may perform channel scaling on the recovered first channel information according to the scaling factor. Optionally, when the terminal device sends the scaling factor to the access network device, the sent scaling factor may be an original value or a quantized value. For example, when the feedback scaling factor is a quantized value, the value of the feedback scaling factor is one of 2 ^U candidate values, and the information field used to carry the scaling factor includes U bits to feed back the scaling factor value Which one of the 2 ^U candidate values is the value. Wherein, U is an integer, such as 1, 2, 3, 4, 5, 6 or other integers, without limitation. The 2 ^U candidate values may be stipulated in the protocol, or pre-configured by the access network device to the terminal device through signaling, and are not limited.

Optionally, the first channel information is a matrix H3 obtained by quantizing H, H1, or H2 (which may be called second channel information). The method can be described as: the first channel information is the quantized channel information of the second channel information.

The terminal device obtains the sparse representation information of the first channel information by using the channel reconstruction model. The sparse representation information of the first channel information includes M elements. Wherein, the M elements include K non-zero elements and M-K zero elements, K is an integer greater than or equal to 1, and M is an integer greater than or equal to K. For each of the K non-zero elements, the value of the element is not limited, for example, may be a real number or a complex number, may be a positive number or a negative number, and/or may be a decimal number or an integer. The channel reconstruction model can be used to restore and obtain the first channel information according to the sparse representation information of the first channel information. Wherein, the value of M is equal to the input dimension of the channel reconstruction model. After the channel reconstruction model is given, the value of M is given. K non-zero elements indicate that the values of the K elements can be non-zero, and it is whatever is calculated. That is to say, during actual calculation, the value of one or some of the K non-zero elements may actually be equal to zero, but even if they are equal to zero, the terminal device will report to the access network according to the rules for reporting non-zero elements. The device reports the values of these elements. For the M-K non-zero elements, the terminal device does not need to report their values to the access network device, because the access network device will assume that the values of these elements are zero by default.

Optionally, different values and/or different positions of K may be set to adapt to different network requirements. For example, various possible feedback compression ratios can be adapted by setting different values of K. The compression ratio of the first channel information can be expressed as the ratio of K and N, where N represents the dimension of the first channel information, and the dimension is N _C ×N _Tx ×N _Rx or 2×N _C ×N _Tx ×N _Rx , N is a positive integer, and 2 means that when the real part and the imaginary part of the first channel information are respectively considered, there are 2 dimensions in total. The following description takes the dimension of the first channel information as N _C × N _Tx × N _Rx as an example. When the dimension is 2 × N _C × N _Tx × N _Rx , those skilled in the art can replace the method described below Get the corresponding method. Wherein, the compression ratio of the first channel information may also be referred to as a feedback compression ratio of the first channel information, a first compression ratio, or other names, without limitation.

Optionally, the value of K or the first compression ratio may be stipulated in the protocol; or, the access network device notifies the terminal device in advance; or, is sent by the terminal device to the access network device through signaling, for example It is sent to the access network device through the channel feedback information in the following S402.

个比特指示第一压缩比的索引或者K的索引，其中，

表示上取整，L为正整数。 Optionally, the terminal device may be notified of multiple candidate compression ratios or multiple candidate values of K by agreement or by the access network device in advance. Further, the access network device notifies the terminal device which of the multiple candidate compression ratios the first compression ratio is, or which of the multiple candidate values the value of K is. Alternatively, the terminal device feeds back which of the multiple candidate compression ratios the first compression ratio is to the access network device through signaling (such as the channel feedback information in S402 below), or the value of K is multiple candidate compression ratios. which of the values. For example, if there are a total of L candidate compression ratios or a total of L candidate values, then the access network device or terminal device can pass a value greater than or equal to

bits indicate the index of the first compression ratio or the index of K, where,

Indicates rounding up, and L is a positive integer.

For example, suppose the dimension N of the first channel information is 4096, where N _C is 64, N _Tx is 16, and N _Rx is 4, and a total of 4 candidate compression ratios are configured, which are: 1/64, 1/128, 1 /256 and 1/512, then K can have 4 values, respectively 64, 32, 16 and 8; or, a total of 4 candidate values of K are configured, respectively: 64, 32, 16 and 8, then There are 4 candidate compression ratios: 1/64, 1/128, 1/256 and 1/512. Assuming that the value of M is 256, the sparse representation information of the first channel has four possible forms, which are: including 256 elements, including 64 non-zero elements and 256-64=192 zero elements; including 256 elements, including 32 non-zero elements and 256-32=224 zero elements; including 256 elements, including 16 non-zero elements and 256-16=240 zero elements; including 256 elements, including 8 non-zero elements and 256-8=248 zero elements. The access network device or the terminal device may indicate the first compression ratio or the specific value of K through 2 bits. Wherein, the value of the 2 bits and the first compression ratio indicated by each value are shown in Table 1A, and the value of the 2 bits and the value of K indicated by each value are shown in Table 1B.

Table 1A

2 bit value first compression ratio 00 1/64 01 1/128 10 1/256

11 1/512

Table 1B

2 bit value K 00 64 01 32 10 16 11 8

In the above method, after a channel reconstruction model is given, various possible feedback compression ratios can be adapted by setting different values of K or different values of the first compression ratio for a channel reconstruction model. Since the input of the channel reconstruction model is sparse information, the value of the number K of non-zero elements in the sparse information can be set to be different to adapt to different compression ratios, so one model can be used to realize each channel information This kind of feedback needs, so that communication resources can be saved, such as no need to train multiple different models for different compression ratios.

For example, given the value of K, the positions of K non-zero elements can be set among the M elements to adapt to different channel environments. Wherein, the positions of the K non-zero elements represent the positions of the K non-zero elements in the M elements. For example, given M and K, a total of U kinds of positions are set. U is a positive integer. For example, four positions are set, corresponding to: the first channel environment, the second channel environment, the third channel environment and the fourth channel environment. The position of the K non-zero elements will be described in more detail in operation S402 below.

The channel reconstruction model can be stipulated in the agreement, such as agreed in the agreement after offline training; or by the network side, such as AI functional network element, OAM, access network equipment, or core network equipment, etc., and sent to the terminal equipment after training or downloaded by the terminal device from a third-party network; or obtained by training the terminal device; there is no limit.

In the above method, when the training node trains to obtain the channel reconstruction model, it can train and obtain the channel reconstruction model according to the training data in the training data set. Wherein, the training data set includes one or more training data. The form of the training data is the same as the form of the above-mentioned first channel information. Exemplarily, the training data is the channel information collected by the terminal device in history (optionally, when the training node is not a terminal device, the terminal device can send the training data to the training node); or, the training data is the access network device Historically collected channel information (optionally, when the training node is not an access network device, the access network device can send the training data to the training node); or, the training data is channel information generated according to a known channel model ; The present disclosure does not limit the way of obtaining or determining the training data.

In a possible implementation, the training node may use the method shown in FIG. 6A or FIG. 6B to train and obtain the channel reconstruction model. The channel reconstruction model obtained through training satisfies: when the input of the channel reconstruction model is sparse representation information of channel information, the channel information can be reconstructed and recovered as accurately as possible.

The training method may include: operation 1, the training node determines a set of training data in the training data set; operation 2, for each training data in the set of training data, determine the sparse representation information of the training data, according to the sparse representation information Determine the model output corresponding to the training data with the current channel reconstruction model; operation 3, for the set of training data, if the loss function meets the performance requirements, the training ends, otherwise, update the current channel reconstruction model, and perform operation 1 again.

The first possible model training method is introduced below in conjunction with FIG. 6A . In this method, the sparse representation information of the training data is determined according to the sparse representation algorithm and the current channel reconstruction model.

Before training, the training node determines the input dimension of the channel reconstruction model, the characteristics of the input data, the output dimension, and the initial model parameters of the channel reconstruction model (for example, the channel reconstruction model is a neural network, and the initial model parameters include: Structural parameters). Wherein, the characteristics of the input data include: the input data includes M elements, and the M elements include K non-zero elements and M-K zero elements. Optionally, K may be equal to any one of multiple candidate values. That is, the channel reconstruction model obtained through training can be applicable to the case where K is equal to any one of the multiple candidate values, that is, the channel reconstruction model can be applicable to various compression ratios.

During the first iterative training, the current channel reconstruction model in the following operation 6A-1 is the initial channel reconstruction model.

Operation 6A-1: The training node determines a set of training data from the training data set, for example, the first set of training data, and for each training data in the set of training data, respectively perform operation 6A-1-1 and operation 6A-1- 2.

In the present disclosure, a set of training data may include one or more training data, for example, may include part or all of the data in the training data set. The number of training data included in different sets of training data may be the same or different. There may or may not be an intersection between different sets of training data, which is not limited.

Operation 6A-1-1: The training node uses one training data A of the training data and the current channel reconstruction model f _de ( ), and obtains the sparse representation information x of the training data A according to the sparse representation algorithm.

Optionally, the sparse representation algorithm includes determining the sparse representation information x of the training data A according to the objective function -.

Among them, H _w represents the training data A, ‖ ‖ ₂ represents the L2 norm, ‖ ‖ ₀ represents the L0 norm, f _de (x) represents the inference result obtained when the input of the channel reconstruction model is x, where x includes M elements, the M elements include K non-zero elements and MK zero elements.

Optionally, the sparse representation algorithm may be any method for solving the sparse reconstruction problem, without limitation. For example, it can be an iterative shrinkage-thresholding algorithm (ISTA), a fast iterative shrinkage-thresholding algorithm (FISTA), or an alternating direction method of multipliers (ADMM) , or an orthogonal matching pursuit (OMP) method; not limited.

Operation 6A-1-2: After the training node obtains the sparse representation information x of the training data A according to the sparse representation algorithm, the training node inputs the sparse representation information x into the current channel reconstruction model f _de ( ), and obtains the reconstruction of the training data A by reasoning Data f _de (x).

Operation 6A-2: For each training data in the set of training data, the training node calculates the loss between each training data, for example, denoted as training data A, and the reconstructed data f _de (x) corresponding to the training data function value. Among them, the loss function is ‖H _w -f _de (x)‖ ₂ . If the average of the loss functions of all the training data in the group of training data (or the value calculated by other methods using the loss functions of all the training data) is less than or equal to the first threshold, or, all the training data in the group If the loss function of the training data is less than or equal to the first threshold, it is considered that the current channel reconstruction model is a reconstruction model obtained through training, and the model training process ends. Otherwise, update the parameters of the channel reconstruction model, such as using the gradient descent method to update the parameters of the channel reconstruction model, use the updated channel reconstruction model as the current channel reconstruction model, and use another set of training data in the training data set, such as For the second set of training data, perform operations 6A-1 and 6A-2 again.

For example, using the E1 group of training data in the training data set, the above operations 6A-1 and 6A-2 can be performed for E2 iterations until the value of the loss function calculated according to the current channel reconstruction model is less than or equal to the first threshold. At the end of the process, the current channel reconstruction model is used as the channel reconstruction model obtained through training. Wherein, E1 and E2 are positive integers. Optionally, E1 is equal to E2, or E1 is smaller than E2, that is, the same training data can be used for repeated iteration training.

Optionally, K may be any one of multiple candidate values. For the same set of training data, each of the candidate values can be used as the value of K to perform operation 6A-1 to operation 6A-2 respectively, so that the channel reconstruction model obtained through training can be applied to various compression ratios. For example, using the E1 group of training data in the training data set, the above operations 6A-1 and 6A-2 can be performed for E2*L iterations until the value of the loss function calculated according to the current channel reconstruction model is less than or equal to the first threshold, It is considered that the training process is over, and the current channel reconstruction model is used as the channel reconstruction model obtained through training. Wherein, E1 and E2 are positive integers, and L is the number of candidate values of K.

Optionally, the objective function 1 in the training process involved in the above-mentioned FIG. 6A can be replaced by the following objective function 2, the loss function can be replaced by f _C (H _w , f _W (f _de (x)), and The training end condition is replaced by the value of the loss function being greater than or equal to the second threshold, and the channel reconstruction model is obtained through training.

Among them, H _w represents the training data A, f _W () represents the precoding generation model, that is, it represents the precoding operation on f _de (x), and f _C (,) represents the channel capacity calculation model.

Optionally, f _W ( ) represents performing singular value decomposition (singular value decomposition, SVD) (it can also be described as SVD precoding). f _C (,) means to calculate the channel capacity calculation, for example:

为f _W(f _de(x))，

为

的共轭转置，I为单位阵，维度为N _Tx×N _Tx；k为子载波索引；det[ ]表示求方阵的行列式。 in,

is f _W (f _de (x)),

for

The conjugate transpose of , I is the identity matrix, the dimension is N _Tx ×N _Tx ; k is the subcarrier index; det[ ] means to find the determinant of the square matrix.

Optionally, in the model training method of the present disclosure, the trained model can also be tested using test data. When the test result reaches the target, for example, when the loss function obtained by using the model for one or more test data meets the performance requirements , the model is considered usable, otherwise the model needs to be retrained. The type of test data is the same as the training data, and will not be repeated here.

Optionally, the iterative process of the above-mentioned sparse representation algorithm can be expanded into a multi-layer neural network by adopting the method of deep network expansion. Expand. Wherein, Q is a positive integer. Each layer of the unfolded network is based on the operation of the channel reconstruction model. As shown in Figure 7, the sparse representation algorithm (also called the sparse representation model) formed by the unfolded network is trained end-to-end together with the channel reconstruction model, and the final channel reconstruction model can be obtained through multiple iterations of training. Alternatively, other model-based methods based on the channel reconstruction model can be used to construct the sparse representation model; no limitation is imposed.

The second possible model training method is introduced below in conjunction with FIG. 6B . According to the current sparse representation model, the sparse representation information of the training data is determined.

Before training, in addition to determining the relevant parameters of the channel reconstruction model, the training node also needs to determine the relevant parameters of the sparse representation model. Wherein, the relevant parameters of the channel reconstruction model are the same as those described above for FIG. 6A . The relevant parameters of the sparse representation model include: the input dimension of the sparse representation model, the output dimension, the characteristics of the output data, and the initial model parameters of the sparse representation model (for example, the sparse representation model is a neural network, and the initial model parameters include: the structural parameters of the model ). Wherein, the characteristics of the output data include: the output data includes M elements, and the M elements include K non-zero elements and M-K zero elements. Optionally, K may be equal to any one of multiple candidate values.

During the first iteration training, the current sparse representation model in the following operation 6B-1 is the initial sparse representation model, and the current channel reconstruction model is the initial channel reconstruction model.

Operation 6B-1: The training node determines a set of training data from the training data set, such as the first set of training data, and performs operation 6B-1-1 and operation 6B-1- for each training data in the training data set respectively 2.

Operation 6B-1-1: The training node inputs the training data A in the set of training data into the current sparse representation model, and obtains the sparse representation information x of the training data A by reasoning.

Operation 6B-1-2: The training node inputs the sparse representation information x into the current channel reconstruction model f _de ( ), and obtains the reconstruction data f _de (x) of the training data A by reasoning.

Operation 6B-2: For each training data in the set of training data, the training node calculates the loss function value between each training data, such as training data A, and the reconstructed data f _de (x) corresponding to the training data . Among them, the loss function is ‖H _w -f _de (x)‖ ₂ . _Hw represents the training data A. If the average of the loss functions of all the training data in the group of training data (or the value calculated by other methods using the loss functions of all the training data) is less than or equal to the first threshold, or, all the training data in the group If the loss function of the training data is less than or equal to the first threshold, it is considered that the current channel reconstruction model is a reconstruction model obtained through training, and the model training process ends. Otherwise, update the parameters of the sparse representation model, such as using gradient descent to update the parameters of the sparse representation model, and use the updated sparse representation model as the current sparse representation model; and/or update the parameters of the channel reconstruction model, such as using gradient descent The method updates the parameters of the channel reconfiguration model, and uses the updated channel reconfiguration model as the current channel reconfiguration model. Using another set of training data in the training data set, for example, the second set of training data, operations 6B-1 and 6B-2 are performed again.

Similar to the introduction in Figure 6A above, after E2 iterations 6B-1 and 6B-2, or E2*L iterations 6B-1 and 6B-2, until the value of the loss function calculated according to the current channel reconstruction model is less than or equal to When the first threshold is reached, the training process is considered to be over. At this point, the current sparse representation model is used as the trained sparse representation model, and the current channel reconstruction model is used as the trained channel reconstruction model.

Similar to the previous introduction to Figure 6A, the loss function in the training process involved in the above Figure 6B can be replaced by f _C (H _w , f _W (f _de (x)), and the training end condition can be replaced by the value of the loss function Greater than or equal to the second threshold, the training obtains a sparse representation model and a channel reconstruction model.Wherein, _Hw represents the training data A, and f _W ( ) represents the precoding generation model, which means that f _de (x) is carried out to the precoding operation, f _C (,) represents the channel capacity calculation model.

Optionally, during the above model training process in FIG. 6A or FIG. 6B , the features of x may be specified. For example, the characteristics of x include: for each of the K non-zero elements, the value of the element is one of multiple candidate values. Optionally, if G bits are used to feed back the value of the element, the multiple candidate values include 2 ^G candidate values. When feeding back the value of the element, the G bits may be used to feed back the index of the value of the element in the 2 ^G candidate values. Among them, G is a positive integer. The value of G may be stipulated in the protocol, or notified to the terminal device by the access network device in advance, and is not limited. For example, G=4, the interval of the value of each non-zero element is [0,1), then the value of each non-zero element is one of 16 candidate values, and the 16 candidate values can be A multiple of 1/16 is added with an offset value, wherein the multiples of different candidate values are different but the offset value is the same, for example, the offset value can be 0, 0.1 or other possible values, without limitation. This method is equivalent to agreeing that the non-zero elements in the sparse representation information are quantized values. Alternatively, it may not be agreed that the non-zero elements in the sparse representation information are quantized values, and data quantization is performed after obtaining the sparse representation information in the actual application process. It can be understood that, in the method of the present disclosure, the values of the non-zero elements may be quantized to save signaling overhead, or may not be quantized to simplify the calculation process, without limitation.

After obtaining the channel reconstruction model, for example, according to the protocol agreement, training to obtain the channel reconstruction model, or receiving the information of the channel reconstruction model from the network side, the terminal device can use the channel reconstruction model to obtain the sparse representation information of the first channel information . Wherein, the method for training a node, such as a terminal device or a network side device, to train a channel reconstruction model may be the method described in FIG. 6A, FIG. 6B or FIG. 7, or other possible methods, without limitation. For example, the channel reconstruction model is a neural network, and the information of the channel reconstruction model includes at least one of the following: structural parameters of the model (such as the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of neurons, the neuron At least one of the activation function of the unit, or the bias in the activation function), the input parameters of the model (such as the type of the input parameter and/or the dimension of the input parameter), or the output parameters of the model (such as the type of the output parameter and / or the dimensions of the output parameter).

The terminal device determines the sparse representation information of the first channel information according to the first objective function or the second objective function.

The method for the terminal device to determine the sparse representation information of the first channel information according to the objective function 1 or the objective function 2 is similar to the method for the training node to determine the sparse representation of the training data A according to the objective function 1 or objective function 2 in the above operation 6A-1. method, which will not be repeated here. The difference between the two is that operation 6A-1 is to use the current channel reconstruction model to obtain the sparse representation of training data A. In the training process shown in Figure 6A, the parameters of the channel reconstruction model can be adjusted; in method A1, it is The channel reconstruction model that has been trained is used for reasoning, and the sparse representation algorithm is used to sparsely represent the first channel information, so as to obtain the sparse representation information of the first channel information. During the reasoning process, the parameters of the channel reconstruction model remain unchanged.

Similar to the introduction of the method related to FIG. 6A above, the method for the terminal device to solve the first channel information according to the objective function may be any method for solving the sparse reconstruction problem, such as the ISTA, FISTA, ADMM, or OMP method. Optionally, the solution process of these algorithms can adopt the method of deep network expansion.

Optionally, when the above sparse representation algorithm solves the sparse representation information of the first channel information through an iterative process, the initial value of the sparse representation information may be a random value, or an output obtained by inferring the first channel information using a sparse representation model . Wherein, the terminal device can obtain the sparse representation model, train the sparse representation model, or receive the information of the sparse representation model from the network side according to the manner stipulated in the protocol. Wherein, the method for training a node, such as a terminal device or a network side device, to train a sparse representation model may be the method described in FIG. 6B above, or other possible methods, which are not limited.

Optionally, similar to the foregoing, in order to further reduce the signaling overhead in the following S402, the terminal device determines the sparse representation information of the first channel information according to objective function 1 or objective function 2, and x can be constrained to satisfy: for K Each of the non-zero elements is one of multiple candidate values. For details, please refer to the previous article, and will not repeat them here. Or, equivalently, when the terminal device determines the sparse representation information of the first channel information according to objective function 1 or objective function 2, x may not be constrained to be a quantized value, and after calculating the sparse representation information, the quantization operation is performed on Each of the K non-zero elements is quantized separately, so that the value of each element is one of the above multiple candidate values. It can be understood that the values of different non-zero elements may be the same or different, without limitation. In the disclosed method, the values of the non-zero elements may be quantized to save signaling overhead, or may not be quantized to simplify the calculation process, without limitation.

In operation S402, the terminal device indicates the sparse representation information to the access network device through channel feedback information. That is, the terminal device sends channel feedback information to the access network device. Wherein, the channel feedback information is used to indicate sparse representation information.

The terminal device may send the sparse representation information to the access network device through channel feedback information through any one of the following methods B1 to B2.

Method B1: the channel feedback information includes sparse representation information.

Exemplarily, the sparse representation information is a matrix [0, 0.25, 0.9375, 0.125, 0, 0, 0.5, 0]. That is, the sparse representation information includes 8 elements, among which 4 are zero elements and 4 are non-zero elements, then the channel feedback information includes [0,0.25,0.9375,0.125,0,0,0.5,0].

Method B2: The channel feedback information is used to indicate the values of the K non-zero elements and the positions of the K non-zero elements of the sparse representation information.

The channel feedback information may indicate the positions of the K non-zero elements through any of the following examples.

Example 1, the channel feedback information indicates the positions of the K non-zero elements through a bitmap (bitmap). Wherein, the bitmap includes M bits, the value of each bit is 0 or 1, and each bit corresponds to an element in the sparse representation information. That is, there is a one-to-one correspondence between M elements in the sparse representation information and M bits in the bitmap. When the value of a bit in the bitmap is 1, it means that the element corresponding to the bit in the sparse representation information is a non-zero element; when the value of a bit in the bitmap is 0, it means that the element corresponding to the bit in the sparse representation information element is zero element.

For example, the sparse representation information is a matrix [0,0.25,0.9375,0.125,0,0,0.5,0], which includes 4 non-zero elements, and the bitmap in the channel feedback information is [0,1,1 ,1,0,0,1,0]. In addition, the channel feedback information also indicates that the values of the four non-zero elements are 0.25, 0.9375, 0.125, and 0.5, respectively. Then, after the access network device receives the channel feedback information from the terminal device, the sparse representation information can be obtained as [0,0.25,0.9375,0.125,0,0,0.5,0].

位的信息分别指示该K个非零元素在M个元素中的位置。 Example 2, the channel feedback information indicates the position of each non-zero element among the K non-zero elements in the M elements. Specifically, the channel feedback information is passed through K greater than or equal to

The bit information respectively indicates the positions of the K non-zero elements in the M elements.

位信息指示该4个非零元素的位置分别为001,010,011和110。此外，信道反馈信息还指示4个非零元素的值分别为0.25,0.9375,0.125,0.5。则，接入网设备接收到来自终端设备的反馈信息后，可以获得稀疏表示信息为[0,0.25,0.9375,0.125,0,0,0.5,0]。 For example, the sparse representation information is a matrix [0,0.25,0.9375,0.125,0,0,0.5,0], which includes 8 elements, 4 of which are non-zero elements, and the positions of the 4 non-zero elements are respectively is 1, 2, 3 and 6, the channel feedback information passes through 4

The bit information indicates that the positions of the four non-zero elements are 001, 010, 011 and 110, respectively. In addition, the channel feedback information also indicates that the values of the four non-zero elements are 0.25, 0.9375, 0.125, and 0.5, respectively. Then, after the access network device receives the feedback information from the terminal device, the sparse representation information can be obtained as [0,0.25,0.9375,0.125,0,0,0.5,0].

Example 3, the channel feedback information indicates the first pattern, and the first image indicates the positions of K non-zero elements among the M elements. Wherein, the first pattern is one of multiple patterns (set of candidate patterns). Through this method, the overhead of channel feedback information can be reduced.

Different patterns in the plurality of patterns indicate different positions of the K non-zero elements. For example, at least one non-zero element indicated by pattern A is a zero element in pattern B. For each pattern in the plurality of patterns, as the example given in the following Table 2A, its form can be the bitmap of the above-mentioned Example 1; or as the example given in the following Table 2B, its form can be the non- Zero element position, no constraints.

For example, based on Table 2A or Table 2B, the channel feedback information indicates that the index of the first pattern is 0, and indicates that the values of the K non-zero elements are 0.25, 0.9375, 0.125, and 0.5. Then, after receiving the feedback information from the terminal device, the access network device can obtain the sparse representation information [0,0.25,0.9375,0.125,0,0,0.5,0].

Optionally, as shown in Table 2C and Table 2D, different candidate pattern sets may be set for different K, and each candidate pattern set is numbered independently. The corresponding pattern can be determined according to the value of K (or pattern set index) and the pattern index. The channel feedback information indicates K=4 or indicates that the index of the candidate pattern set where the first pattern is located is 0, the index of the first pattern is 0, and the values indicating K non-zero elements are 0.25, 0.9375, 0.125, 0.5. Then, after receiving the feedback information from the terminal device, the access network device can obtain the sparse representation information [0,0.25,0.9375,0.125,0,0,0.5,0].

Table 2A

the index (id) pattern value pattern one 0 [0,1,1,1,0,0,1,0] pattern two 1 [0,0,0,1,0,0,1,0] pattern three 2 [1,0,1,1,1,0,0,0] pattern four 3 [0,0,0,1,0,0,0,1]

Table 2B

the index (id) pattern value pattern one 0 001,010,011,110 pattern two 1 011,110 pattern three 2 000,010,011,100 pattern four 3 011,111

Table 2C (K=4, candidate pattern set 0)

the index (id) pattern value pattern one 0 001,010,011,110 pattern two 1 000,010,011,100

Table 2D (K=2, candidate pattern set 1)

the index (id) pattern value pattern one 0 011,110 pattern two 1 011,111

The above multiple patterns may be stipulated in the agreement, or the access network device notifies the terminal device in advance through signaling, without limitation.

The terminal device can directly indicate the values of the K non-zero elements; or, as mentioned above, in order to save signaling overhead, if the values of the K non-zero elements are quantized values, the channel feedback information indicates the value of each non-zero element value, may indicate the index of the value of the non-zero element among multiple candidate values.

For example, in each of the above examples, if the values of the K non-zero elements are quantized values, for example, there are 16 candidate values, which are respectively multiples of 1/16, namely: 0 (index: 0000), 0.0625 (index: 0001), 0.125 (Index: 0010), 0.1875 (Index: 0011), 0.25 (Index: 0100), 0.3125 (Index: 0101), 0.375 (Index: 0110), 0.4375 (Index: 0111), 0.5 (Index: 1000 ), 0.5625 (Index: 1001), 0.625 (Index: 1010), 0.6875 (Index: 1011), 0.75 (Index: 1100), 0.8125 (Index: 1101), 0.875 (Index: 1110), 0.9375 (Index: 1111) . When the channel feedback information indicates that the values of the four non-zero elements are 0.25, 0.9375, 0.125, and 0.5, the index of the element can be indicated by 4 bits for each element, and the channel feedback information indicates: 0100, 1111, 0010, 1000.

In operation S403, the access network device recovers the first channel information by using the sparse representation information and the channel reconstruction model.

Exemplarily, after receiving the channel feedback information, the access network device obtains sparse representation information according to the channel feedback information, inputs the sparse representation information into a channel reconstruction model, and obtains reconstructed (restored) first channel information by reasoning.

Exemplarily, the access network device sets the initial input of the channel reconstruction model as M zeros. That is, the input dimension of the channel reconstruction model is M dimension. The access network device uses the positions and values of the K non-zero elements indicated by the feedback information to replace the K 0 elements in the initial input of the channel reconstruction model with the K non-zero elements indicated by the feedback information, and then according to the channel reconstruction The reconstructed (recovered) first channel information is obtained by reasoning the structural model.

Optionally, if the first channel information is normalized channel information of the second channel information, as described in S401, the terminal device may also send the scaling factor of the second channel information to the access network device. The access network uses the scaling factor to scale the first channel information to obtain the second channel information. The scaling factor can be in the linear domain or the logarithmic domain. Wherein, there is no restriction on the base of the logarithm, for example, it may be 10, 2, a natural constant e or other possible values, without restriction. For example, taking the scaling factor as an example of the scaling factor of the second channel information relative to the first channel information, and the first channel information is denoted as H ₁ , then the second channel information H ₂ =H ₁ *T, where T represents the scaling factor The value of the linear domain of the factor, T is a number less than 1 or a number greater than or equal to 1. For another example, taking the scaling factor as an example of the scaling factor of the second channel information relative to the first channel information, the first channel information is denoted as H ₁ , and the base of the logarithm is 10, then the second channel information H ₂ =H ₁ * 10 ^T , where T represents the value in the logarithmic domain of the scaling factor, and T is 0, a positive number or a negative number.

The method of the present disclosure can be understood as, for various communication scenarios, such as various channel environments and/or various compression ratios, a channel reconstruction model on the side of the fixed access network device. Since the terminal device uses the channel reconstruction model to obtain the sparse representation of channel information, there are no constraints on the solution method on the terminal device side, which can meet the needs of various communication scenarios and simplify the implementation of the terminal device side. The channel reconstruction model on the access network device side can be regarded as a dictionary network, which can restore channel information through sparse representation information.

It can be understood that in practical applications, one channel reconfiguration model can be used to apply to all communication scenarios; multiple channel reconfiguration models can also be used, and each channel reconfiguration model can be applied to multiple communication scenarios to relatively save communication resources.

Optionally, in operation S404, the access network device may determine transmission parameters according to the recovered first channel information or second channel information, for data transmission with the terminal device.

For example, the access network device determines channel quality information (channel quality indicator, CQI) according to the recovered first channel information or second channel information. The CQI is used by the access network device to schedule a physical downlink shared channel (PDSCH), that is, it is used by the access network device to determine the PDSCH (time domain and/or frequency domain) resources, and/or the modulation and coding mechanism ( modulation and coding scheme, MCS) and other transmission parameters. Optionally, in a system where the uplink channel and the downlink channel are reciprocal, the CQI can also be used by the access network device to schedule a physical uplink shared channel (PUSCH), that is, for the access network device to determine PUSCH (time domain and/or frequency domain) resources, and/or transmission parameters such as MCS.

For example, the access network device determines the precoding matrix indicator (precoding matrix indicator, PMI) and/or rank indicator (rank indicator, RI) of PDSCH and/or PUSCH according to the recovered first channel information or second channel information . Wherein, the PMI and RI of the PDSCH and/or PUSCH are used for multi-antenna transmission by the terminal device and the access network device, for example, massive MIMO transmission.

For example, the access network device may send the PMI and RI of the PDSCH to the terminal device for the terminal device to decode data carried on the PDSCH. And/or, the access network device may send the PMI and RI of the PUSCH to the terminal device, so that the terminal device determines the data carried on the PUSCH according to the PMI and RI.

It can be understood that, in order to realize the functions in the above method, the access network device, the terminal device, and the network element with AI function include hardware structures and/or software modules corresponding to each function. Those skilled in the art should easily realize that, in combination with the units and method steps of each example described in the present disclosure, the present disclosure can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed by hardware or computer software drives the hardware depends on the specific application scenario and design constraints of the technical solution.

FIG. 8 and FIG. 9 are schematic structural diagrams of possible communication devices provided by the present disclosure. These communication devices can be used to realize the functions of the access network device, the terminal device, and the AI functional network element in the above method, and thus can also realize the beneficial effects of the above method.

As shown in FIG. 8 , a communication device 800 includes a processing unit 810 and a communication unit 820 . The communication device 800 is used to implement the method shown above.

When the communication apparatus 800 is used to realize the function of the access network equipment, the communication unit 820 is used to receive channel feedback information from the terminal equipment, where the channel feedback information is used to indicate the sparse representation information of the first channel information, wherein the sparse representation information It includes M elements, and the M elements include K non-zero elements and M-K zero elements, where M and K are positive integers; the processing unit 810 is used to determine the first channel information according to the channel reconstruction model, where the The input of the channel reconstruction model is determined according to the sparse representation information.

When the communication device 800 is used to realize the functions of the terminal equipment: the processing unit 810 is used to determine the sparse representation information of the first channel information according to the first channel information and the channel reconstruction model, wherein the sparse representation information includes M elements, and the M The elements include K non-zero elements and M-K zero elements, where M and K are positive integers; the communication unit 820 is used to send channel feedback information to the access network device, and the channel feedback information is used to indicate the sparse representation information .

A more detailed description about the processing unit 810 and the communication unit 820 may refer to the related description in the foregoing method, and details are not repeated here.

As shown in FIG. 9 , a communication device 900 includes a processor 910 and an interface circuit 920 . The processor 910 and the interface circuit 920 are coupled to each other. It can be understood that the interface circuit 920 may be a transceiver, a pin, an input/output interface or other communication interfaces. Optionally, the communication device 900 may further include a memory 930 for storing at least one of the following: instructions executed by the processor 910, input data required by the processor 910 to execute the instructions, or data generated after the processor 910 executes the instructions.

When the communication device 900 is used to implement the above method, the processor 910 is used to implement the functions of the processing unit 810 , and the interface circuit 920 is used to implement the functions of the communication unit 820 .

When the above communication device is a chip applied to a terminal device, the terminal device chip implements the functions of the terminal device in the above method. The terminal device chip receives information from other modules in the terminal device (such as radio frequency modules or antennas), and the information is sent to the terminal device by the access network device; or, the terminal device chip sends information to other modules in the terminal device (such as radio frequency module or antenna) to send information, the information is sent by the terminal device to the access network device and so on.

When the above communication device is a module applied to access network equipment, the access network equipment module implements the functions of the access network equipment in the above method. The access network equipment module receives information from other modules (such as radio frequency modules or antennas) in the access network equipment, and the information is sent to the access network equipment by terminal equipment; or, the access network equipment module sends information to the access network equipment Other modules (such as radio frequency modules or antennas) in the network equipment send information, and the information is sent by the access network equipment to the terminal equipment and so on. The access network equipment module here can be the baseband chip of the access network equipment, or it can be near real-time RIC, CU, DU or other modules. The near real-time RIC, CU and DU here may be the near real-time RIC, CU and DU under the O-RAN architecture.

In the present disclosure, a processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, and may realize or execute the present disclosure Each method, step and logical block diagram of the method. A general purpose processor may be a microprocessor or any conventional processor or the like. The steps combined with the method of the present disclosure may be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

In the present disclosure, the memory may be a non-volatile memory, such as a hard disk (hard disk drive, HDD) or a solid-state drive (solid-state drive, SSD), etc., or a volatile memory (volatile memory), such as random access Memory (random-access memory, RAM). A memory is, but is not limited to, any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The memory in the present disclosure may also be a circuit or any other device capable of implementing a storage function for storing program instructions and/or data.

The methods in the present disclosure may be fully or partially implemented by software, hardware, firmware or any combination thereof. When implemented using software, it may be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer programs or instructions. When the computer programs or instructions are loaded and executed on the computer, the processes or functions described in this application are executed in whole or in part. The computer may be a general computer, a dedicated computer, a computer network, an access network device, a terminal device, a core network device, an AI function network element, or other programmable devices. The computer program or instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer program or instructions may be downloaded from a website, computer, A server or data center transmits to another website site, computer, server or data center by wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or a data center integrating one or more available media. The available medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; it may also be an optical medium, such as a digital video disk; or it may be a semiconductor medium, such as a solid state disk. The computer readable storage medium may be a volatile or a nonvolatile storage medium, or may include both volatile and nonvolatile types of storage media.

Claims (26)

A channel information transmission method, characterized in that, comprising: Receive channel feedback information from the terminal device, where the channel feedback information is used to indicate sparse representation information of the first channel information, where the sparse representation information includes M elements, and the M elements include K non-zero elements And M-K zero elements, where M and K are positive integers; The first channel information is determined according to a channel reconstruction model, wherein an input of the channel reconstruction model is determined according to the sparse representation information. The method according to claim 1, wherein the channel feedback information is used to indicate values of the K non-zero elements and positions of the K non-zero elements. The method according to claim 1 or 2, wherein the channel feedback information is used to indicate a first pattern, and the first pattern indicates the positions of the K non-zero elements, wherein the first The pattern is one of a plurality of candidate patterns. The method according to any one of claims 1-3, further comprising: Determine at least one of the following items according to the first channel information: a precoding matrix indicator PMI, a rank indicator RI, or a channel quality indicator CQI. The method according to any one of claims 1-3, wherein the channel feedback information is further used to indicate a scaling factor of the first channel information relative to the second channel information, wherein the first channel The information is normalized channel information of the second channel information. The method according to claim 5, further comprising: Determine at least one of the following according to the second channel information: PMI, RI, or CQI. The method according to any one of claims 1-6, wherein the ratio of K and N is the first compression ratio, where N is a positive integer, and N represents the dimension of the first channel information, so The method also includes: Sending information indicating that the first compression ratio is one of multiple candidate compression ratios to the terminal device. The method according to any one of claims 1-6, wherein the method further comprises: Sending information indicating that K is one of multiple candidate values to the terminal device. A channel information transmission method, characterized in that, comprising: Determine the sparse representation information of the first channel information according to the first channel information and the channel reconstruction model, wherein the sparse representation information includes M elements, and the M elements include K non-zero elements and M-K zero elements, wherein , M and K are positive integers; Send channel feedback information to the access network device, where the channel feedback information is used to indicate the sparse representation information. The method according to claim 9, wherein determining the sparse representation information of the first channel information according to the first channel information and the channel reconstruction model comprises: The sparse representation information of the first channel information is determined according to the following objective function:

Among them, ‖x‖ ₀ ≤ K, x represents the sparse representation information of the first channel information, H _w represents the first channel information, f _de ( ) represents the channel reconstruction model, ‖ ‖ ₂ represents the L2 norm, ‖ ‖ ₀ represents L0 norm; or, The sparse representation information of the first channel information is determined according to the following objective function:

Among them, ‖x‖ ₀ ≤ K, x represents the sparse representation information of the first channel information, H _w represents the first channel information, f _de ( ) represents the channel reconstruction model, f _W ( ) represents the precoding generation model, f _C (,) represents the channel capacity calculation model. The method according to claim 9 or 10, wherein the channel feedback information is used to indicate values of the K non-zero elements and positions of the K non-zero elements. The method according to claim 11, wherein the channel feedback information is used to indicate a first pattern, and the first pattern indicates the positions of the K non-zero elements, wherein the first pattern is One of multiple candidate patterns. The method according to any one of claims 9-12, wherein the channel feedback information is further used to indicate a scaling factor of the first channel information relative to the second channel information, wherein the first channel The information is normalized channel information of the second channel information. The method according to any one of claims 9-13, wherein the ratio of K and N is a first compression ratio, and the N represents the dimension of the first channel information, and the method further comprises: Information indicating that the first compression ratio is one of multiple candidate compression ratios is received from the access network device. The method according to any one of claims 9-13, wherein the method further comprises: Receive information indicating that K is one of multiple candidate values from the access network device. A model training method, characterized in that, comprising: Operation 1, determine a set of training data in the training data set; operation 2, for each training data in the set of training data, determine the sparse representation information of the training data, and reconstruct the model according to the sparse representation information and the current channel Determine the model output corresponding to the training data; operation 3, for the set of training data, if the loss function meets the performance requirements, the training ends, otherwise, update the channel reconstruction model, and perform operation 1 again. The method according to claim 16, wherein said determining the sparse representation information of the training data comprises: determining the sparse representation information of the training data according to the sparse representation algorithm and the current channel reconstruction model, or, According to the current sparse representation model, the sparse representation information of the training data is determined, and if the loss function does not meet the performance requirement, the method further includes: updating the sparse representation model. . The method according to claim 16 or 17, wherein, for the group of training data, the loss function meets performance requirements, comprising: the average value of the loss function of all training data in the group of training data (or using all The values of each loss function of the training data calculated by other methods) meet the threshold requirement, or, the loss functions of all the training data in the set of training data meet the threshold requirement. A communication device, characterized in that it is used to implement the method described in any one of claims 1-8. A communication device, characterized in that it includes a processor and a memory, the processor and the memory are coupled, and the processor is used to execute programs or instructions stored in the memory, so that the communication device implements any of claims 1-8. one of the methods described. A communication device, characterized in that it is used to implement the method described in any one of claims 9-15. A communication device, characterized in that it includes a processor and a memory, the processor and the memory are coupled, and the processor is used to execute programs or instructions stored in the memory, so that the communication device implements claims 9-15 any one of the methods described. A communication device, characterized in that it is used to implement the method described in any one of claims 16-18. A communication system, characterized in that it includes one or more of the following: The communication device according to claim 19 or 20, the communication device according to claim 21 or 22, or the communication device according to claim 23. A computer-readable storage medium, characterized in that instructions are stored on the computer-readable storage medium, and when the instructions are run on a computer, the computer is made to perform the method described in any one of claims 1-8, Or the method described in any one of claims 9-15, or the method described in any one of claims 16-18. A computer program product, characterized in that it includes instructions, and when the instructions are run on a computer, the computer is made to perform the method described in any one of claims 1-8, or the method described in any one of claims 9-15 The method, or the method described in any one of claims 16-18.

Images