Method, system and electronic device for switching channel state information feedback mode

Abstract

This application discloses a method, system, and electronic device for switching channel state information feedback modes. The method includes: measuring initial channel state information between the device and a network device, and monitoring the accuracy of a first encoding result output by a neural network-based coding model for the input initial channel state information; if the accuracy does not meet preset conditions, switching the feedback mode from the currently used primary feedback mode to a backup feedback mode; wherein, in the backup feedback mode, the uplink control information fed back to the network device carries a second encoding result and an identifier of the backup feedback mode; the identifier of the backup feedback mode is used to instruct the network device to decode based on the second encoding result; the second encoding result is a precoding matrix index generated based on the codebook specification for encoding the initial channel state information. This technical solution can significantly shorten the reconfiguration time without relying on RRC signaling during the switching process, while minimizing communication interruptions.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a method, system and electronic device for switching channel state information feedback modes. Background Technology

[0002] In the 3GPP Rel-18 Artificial Intelligence (AI) research, User Equipment (UE) uses a neural network model to compress and feed back Channel State Information (CSI), while network equipment (such as base stations) uses a corresponding decoder to recover the fed-back CSI. Compared to the traditional method of compressing and feeding back CSI according to the Type I / II codebook, the neural network model compression and feedback method can significantly reduce CSI feedback overhead.

[0003] Current neural network models suffer from a "generalization boundary" problem, meaning that when a user device enters a scenario where the neural network model has not been trained (such as high-speed movement or extreme multipath scenarios), the error in model inference, i.e. the error in the encoding and compression results of CSI, will increase instantaneously.

[0004] Therefore, related technologies rely on Radio Resource Control (RRC) signaling for model management. When a user equipment (UE) detects excessive inference errors in its neural network model, it needs to report an error measurement report to the network device via RRC signaling. The network device then issues an RRC reconfiguration command to the UE based on the error measurement report. Only then can the UE perform a reconfiguration operation based on the received RRC reconfiguration command, thereby stopping the use of the neural network model for CSI compression. However, the reconfiguration process via RRC signaling is time-consuming, and during this period, the network device continues to use the neural network model to compress the CSI with large errors for subsequent processing, leading to communication interruption. Summary of the Invention

[0005] This application provides a method, system, electronic device, and readable storage medium for switching channel state information feedback modes. The user equipment (UE) monitors the accuracy of the first encoding result output by a neural network-based coding model, autonomously decides which feedback mode to adopt, and efficiently instructs the network device to adopt the corresponding decoding mode by configuring an identifier of the currently decided feedback mode in the fed-out uplink control information. When the accuracy does not meet preset conditions, the UE autonomously and efficiently switches to a backup feedback mode, feeding back to the network device using a traditional precoding matrix index instead of the output result of the neural network-based coding model. The switching process does not rely on RRC signaling, significantly shortening the reconfiguration time. Simultaneously, during reconfiguration, the network device avoids using the feedback result with a large error output by the coding model, instead using the relatively more reliable traditional precoding matrix index for subsequent decoding processing, minimizing communication interruptions.

[0006] In a first aspect, embodiments of this application provide a method for switching channel state information feedback modes, applied to a user equipment, wherein the user equipment is communicatively connected to a network device, and the method includes: Measure the initial channel state information between the network device and the network device, and monitor the accuracy of the first encoding result output by the neural network-based encoding model in response to the input initial channel state information; If the accuracy does not meet the preset conditions, the feedback mode will be switched from the currently used main feedback mode to the backup feedback mode. In the backup feedback mode, the uplink control information fed back to the network device carries a second encoding result and an identifier of the backup feedback mode; the identifier of the backup feedback mode is used to instruct the network device to decode based on the second encoding result; the second encoding result is a precoding matrix index generated based on the codebook specification for encoding the initial channel state information; In the primary feedback mode, the uplink control information fed back to the network device carries the first encoding result and the identifier of the primary feedback mode; the identifier of the primary feedback mode is used to instruct the network device to decode based on the first encoding result.

[0007] Optionally, the network device is deployed with a first decoding model based on a neural network, and the user equipment is deployed with a second decoding model as a copy of the first decoding model; the second decoding model and the first decoding model are capable of outputting decoding results based on the encoding results of the input channel state information; Accordingly, the accuracy is determined in the following ways: The second decoding model is invoked, and the output result of the second decoding model is obtained using the first encoding result. The accuracy is determined based on the deviation between the output result and the initial channel state information.

[0008] Optionally, the uplink control information fed back to the network device includes first control information and second control information; the first control information is a high-priority transmission portion, and the second control information is a low-priority transmission portion; the first control information has an identifier bit for the feedback mode; when the feedback mode is the backup feedback mode, the identifier bit is configured with an identifier for the backup feedback mode; when the feedback mode is the primary feedback mode, the identifier bit is configured with an identifier for the primary feedback mode; the second control information, when the identifier bit is configured with an identifier for the backup feedback mode, configures the second encoding result; when the identifier bit is configured with an identifier for the primary feedback mode, configures the first encoding result.

[0009] Optionally, the method further includes: determining a target number of bits, and defining the identifier bit in the first control information according to the target number of bits; wherein the target number of bits is less than or equal to a preset threshold.

[0010] Optionally, after monitoring the accuracy of the first coding result output by the neural network-based coding model for the input initial channel state information, the method further includes: If the accuracy meets the preset conditions, the feedback mode remains the currently used main feedback mode.

[0011] Secondly, embodiments of this application provide a method for switching channel state information feedback modes, applied to a network device, wherein the network device is communicatively connected to a user equipment, and the method includes: Receive uplink control information fed back by the user equipment, and parse out the identifier of the feedback mode carried in the uplink control information; If the identifier of the feedback mode is determined to be the identifier of the main feedback mode, then the decoding mode of the channel state information being fed back will remain the currently used main decoding mode. If the identifier of the feedback mode is determined to be the identifier of the backup feedback mode, then the decoding mode is switched from the main decoding mode to the backup decoding mode; The main decoding mode is a mode in which a neural network-based decoding model is invoked to decode the first encoding result; the first encoding result is the encoding result of the initial channel state information measured by the user equipment by the neural network-based encoding model. The backup decoding mode is a mode that calls a decoder based on the codebook specification to decode the second encoding result; the second encoding result is a precoding matrix index generated by encoding the initial channel state information based on the codebook specification.

[0012] Optionally, the uplink control information includes first control information and second control information; the first control information is a high-priority transmission portion, and the second control information is a low-priority transmission portion; the first control information is provided with an identifier bit for the feedback mode; the second control information, when the identifier bit is configured to identify the backup feedback mode, configures the second encoding result; when the identifier bit is configured to identify the primary feedback mode, configures the first encoding result. Accordingly, parsing the identifier of the feedback mode carried in the uplink control information includes: The identifier of the feedback mode is parsed from the identifier bit; Accordingly, the method further includes: With the decoding mode remaining in the main decoding mode, the first encoding result is parsed from the second control information; When the decoding mode is switched to the backup decoding mode, the second encoding result is parsed from the second control information.

[0013] Thirdly, embodiments of this application provide a switching system for channel state information feedback mode, the system including user equipment and network equipment communicatively connected to the user equipment; The user equipment is configured to measure initial channel state information (ACSI) with the network device and monitor the accuracy of a first encoding result output by a neural network-based encoding model for the input ACSI. If the accuracy does not meet a preset condition, the feedback mode is switched from the currently used primary feedback mode to a backup feedback mode. In the backup feedback mode, the uplink control information fed back to the network device carries a second encoding result and an identifier of the backup feedback mode. The identifier of the backup feedback mode is used to instruct the network device to decode based on the second encoding result. The second encoding result is a precoding matrix index generated based on the codebook specification for encoding the ACSI. In the primary feedback mode, the uplink control information fed back to the network device carries the first encoding result and an identifier of the primary feedback mode. The identifier of the primary feedback mode is used to instruct the network device to decode based on the first encoding result. The network device is configured to receive uplink control information fed back by the user equipment and parse out the identifier of the feedback mode carried in the uplink control information; if the identifier of the feedback mode is determined to be the identifier of the primary feedback mode, the decoding mode for the feedback channel state information is maintained as the currently used primary decoding mode; if the identifier of the feedback mode is determined to be the identifier of the backup feedback mode, the decoding mode is switched from the primary decoding mode to the backup decoding mode; wherein, the primary decoding mode is a mode that calls a decoding model based on a neural network to decode the first encoding result; the backup decoding mode is a mode that calls a decoder based on a codebook specification to decode the second encoding result.

[0014] Fourthly, embodiments of this application provide an electronic device, the electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the method as described in the first or second aspect.

[0015] Fifthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the method described in the first or second aspect.

[0016] In this embodiment of the application, the mode of user equipment feedback channel status information is divided into two modes: main feedback mode and backup feedback mode.

[0017] The user equipment measures the initial channel state information between itself and the network equipment, and monitors the accuracy of the first coding result output by the neural network-based coding model in response to the input initial channel state information. In main feedback mode, the uplink control information fed back by the user equipment to the network equipment carries the first coding result and the main feedback mode identifier, which instructs the network equipment to decode based on the first coding result.

[0018] When the accuracy does not meet the preset conditions, the user equipment switches the feedback mode from the currently used primary feedback mode to the backup feedback mode. In the backup feedback mode, the uplink control information fed back by the user equipment to the network equipment carries the second encoding result and the identifier of the backup feedback mode. The identifier of the backup feedback mode is used to instruct the network equipment to decode based on the second encoding result, which is the precoding matrix index generated based on the codebook specification for encoding the initial channel state information.

[0019] As can be seen, the user equipment (UE) autonomously decides the current feedback mode by monitoring the accuracy of the first encoding result output by the neural network-based encoding model. It efficiently instructs the network device to adopt the corresponding decoding mode by configuring an identifier for the currently decided feedback mode in the feedback uplink control information. When the accuracy does not meet preset conditions, the UE autonomously and efficiently switches to the backup feedback mode, using the traditional precoding matrix index to provide feedback to the network device instead of the output of the neural network-based encoding model. This switching process does not rely on RRC signaling, significantly shortening the reconfiguration time. Furthermore, during reconfiguration, the network device avoids using the feedback result with a large error from the encoding model, instead using the relatively more reliable traditional precoding matrix index for subsequent decoding, minimizing communication interruptions. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the architecture of a channel state information feedback system in related technologies; Figure 2 This is a flowchart illustrating a method for switching channel state information feedback modes according to an embodiment of this application; Figure 3 This is a flowchart illustrating a method for switching channel state information feedback modes according to another embodiment of this application; Figure 4 This is a schematic diagram of the structure of a channel state information feedback mode switching system according to an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, specific embodiments of this application will be described in further detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely for explaining this application and not for limiting it. It should also be noted that, for ease of description, only the parts relevant to this application are shown in the drawings, not all of them. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe operations (or steps) as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0022] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0023] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0024] The following description, in conjunction with the accompanying drawings, details the channel state information feedback mode switching method, system, electronic device, and readable storage medium provided in this application through specific embodiments and application scenarios.

[0025] Please see Figure 1 First, we will provide an exemplary introduction to the architecture of the channel state information feedback system in related technologies: like Figure 1 As shown, Figure 1 This is a schematic diagram of the architecture of a channel state information feedback system in related technologies. Current massive multiple input multiple output (Massive MIMO) systems use frequency division duplex (FDD) in some scenarios. FDD is the mainstream duplex technology for achieving bidirectional transmission in communication systems. It can achieve full-duplex communication by allocating two symmetrical independent frequency bands to carry uplink signals (uplink channel) and downlink signals (downlink channel) respectively, and setting up guard bands to isolate the transmit and receive channels. FDD does not require strict time synchronization, and the transmit and receive links can work simultaneously.

[0026] In FDD mode, the uplink and downlink channels use different frequencies and operate in a simultaneous duplex mode. To achieve downlink beamforming in FDD mode, the base station needs to obtain the downlink channel state information (CSI). However, the biggest difference between FDD and Time-division Duplex (TDD) modes is the lack of channel reciprocity. This means that network equipment (such as base stations) cannot directly obtain the downlink channel CSI from the uplink signal; instead, the downlink channel CSI must rely on user equipment (UE), such as mobile phones, for measurement and feedback.

[0027] Before a user equipment measures the CSI between itself and a network device and sends it back to the network device via the uplink channel, it often needs to compress and encode the CSI before sending it to reduce the overhead of CSI feedback. The network device then decodes the compressed and encoded result of the received CSI to restore the original CSI.

[0028] In the 3GPP Rel-18 Artificial Intelligence (AI) research, user equipment uses a neural network model to compress and feed back Channel State Information (CSI), while network equipment (such as base stations) uses a corresponding decoder to recover the fed-back CSI. Compared to the traditional method of compressing and feeding back CSI according to the Type I / II codebook, the neural network model compression and feedback method can significantly reduce CSI feedback overhead.

[0029] Current neural network models suffer from a "generalization boundary" problem, meaning that when a user device enters a scenario where the neural network model has not been trained (such as high-speed movement or extreme multipath scenarios), the error in model inference, i.e. the error in the encoding and compression results of CSI, will increase instantaneously.

[0030] Therefore, related technologies rely on Radio Resource Control (RRC) signaling for model management. When a user equipment (UE) detects excessive inference errors in its neural network model, it needs to report an error measurement report to the network device via RRC signaling. The network device then issues an RRC reconfiguration command to the UE based on the error measurement report. Only then can the UE perform a reconfiguration operation based on the received RRC reconfiguration command, thereby stopping the use of the neural network model for CSI compression. However, the reconfiguration process via RRC signaling is time-consuming, and during this period, the network device continues to use the neural network model to compress the CSI with large errors for subsequent processing, leading to communication interruption.

[0031] It is worth noting that the above description of CSI feedback scenarios in related technologies is only an illustrative example. In actual applications, other types of CSI feedback scenarios may exist, and no specific limitations are made here.

[0032] To address the aforementioned issues, this application provides a method for switching channel state information feedback modes. This method eliminates the need for RRC signaling during the switching process, significantly reducing reconfiguration time. Furthermore, during reconfiguration, it avoids the network device from using feedback results with large errors output by the coding model. Instead, it uses the more reliable traditional precoding matrix index for subsequent decoding, minimizing communication interruptions.

[0033] The user equipment (UE) communicates with the network equipment. Under different circumstances, the UE feeds back the measured CSI to the network equipment using different feedback modes. Several embodiments are provided; one embodiment is described below. For details on the execution process of this embodiment, please refer to [link to relevant documentation]. Figure 2 As shown, Figure 2 This is a flowchart illustrating a method for switching channel state information feedback modes according to an embodiment of this application. An embodiment of the method for switching channel state information feedback modes provided in this application may include: 201. The user equipment measures the initial channel state information between itself and the network equipment, and monitors the accuracy of the first coding result output by the neural network-based coding model in response to the input initial channel state information.

[0034] User equipment can refer to terminal devices with communication functions, including mobile phones, tablets, laptops, smartwatches, etc., which enable users to communicate through wireless communication networks.

[0035] Network devices can be hardware devices used to build, maintain and manage communication networks. Their core function is to realize the transmission, exchange, routing and security control of information between different nodes, such as base stations, specifically 5G base stations or 6G base stations.

[0036] Initial channel state information can be the frequency domain CSI of the downlink channel between the network device and the user equipment, as measured by the user equipment. CSI is a channel attribute of the communication link, which describes the signal attenuation factor on each transmission path, i.e., the value of each element in the channel gain matrix H, such as information on signal scattering, fading, and power decay of distance.

[0037] A neural network (NN) can be a computational model that mimics the structure and function of biological neural networks. It consists of a large number of interconnected artificial neurons (nodes) and is used to simulate the information transmission and processing mechanisms between biological neurons. It can extract patterns from data through learning and training to complete tasks such as classification, prediction, and recognition.

[0038] It is understandable that the above-mentioned neural network-based encoding model belongs to the category of neural network models. A typical neural network model can include the following layers: Input Layer: receives raw data; Hidden Layer: located between input and output, responsible for extracting features; Output Layer: generates the final prediction result; where each neuron receives the input from the previous layer, calculates the weighted sum, and then generates the output through an activation function.

[0039] The first encoding result can be the output of the aforementioned neural network-based encoding model, which is the result of inputting the initial channel state information into the aforementioned neural network-based encoding model, which internally compresses and encodes the initial channel state information.

[0040] The accuracy of the first encoding result output by the above encoding model can refer to the degree of similarity between the CSI decoded and restored based on the first encoding result and the initial CSI measured by the user equipment.

[0041] Optionally, the user equipment (UE) can measure the initial channel state information (CSI) between itself and the network equipment by relying primarily on a known reference signal transmitted by the network equipment (such as a base station). The network equipment transmits a pilot symbol called a "channel state information reference signal" on specific time-frequency resources. This is equivalent to a standard signal sample known to both the UE and the network equipment. After receiving the channel state information reference signal, the UE compares the actual received signal with the difference between the signal and the reference signal to infer the attenuation, phase rotation, and delay changes the signal underwent in the wireless environment. In essence, the UE perceives the current wireless channel state information, i.e., the initial channel state information, by analyzing the changes in the known signal transmitted by the network equipment after transmission.

[0042] Optionally, the method by which the user equipment monitors the accuracy of the first encoding result output by the neural network-based encoding model for the input initial channel state information may include: the network device deploying a first decoding model based on the neural network, and the user equipment deploying a second decoding model as a copy of the first decoding model; the second decoding model and the first decoding model being able to output a decoding result based on the encoding result of the input channel state information; the user equipment calling the second decoding model, using the first encoding result output by the neural network-based encoding model for the input initial channel state information to obtain the output result of the second decoding model, and determining the accuracy based on the deviation between the output result of the second decoding model and the initial channel state information.

[0043] The output of the second decoding model can be the channel state information recovered by decoding and restoring the encoded result of the input channel state information.

[0044] As an example, given that the network device has deployed a first decoding model, the user equipment can simulate the decoding operation of the network device, that is, the user equipment also deploys a second decoding model as a copy of the first decoding model, that is, the second decoding model is exactly the same as the first decoding model.

[0045] 202. If the accuracy does not meet the preset conditions, the user equipment will switch the feedback mode from the currently used primary feedback mode to the backup feedback mode.

[0046] In the backup feedback mode, the uplink control information fed back to the network device carries the second encoding result and the identifier of the backup feedback mode; the identifier of the backup feedback mode is used to instruct the network device to decode based on the second encoding result; the second encoding result is the precoding matrix index generated by encoding the initial channel state information based on the codebook specification; In the main feedback mode, the uplink control information fed back to the network device carries the first encoding result and the main feedback mode identifier; the main feedback mode identifier is used to instruct the network device to decode based on the first encoding result.

[0047] The preset condition may refer to the similarity between the CSI decoded and restored based on the first encoding result and the initial CSI measured by the user equipment being greater than or equal to the similarity threshold.

[0048] Feedback mode can refer to the way the user equipment compresses and encodes the initial CSI measurement, and the way it generates and feeds back uplink control information accordingly.

[0049] The main feedback mode refers to the normal feedback mode that the user device maintains under normal circumstances. Normal circumstances mean that the inference performance of the aforementioned neural network-based encoding model is normal; when the inference performance is normal, the user device consistently uses this encoding model for compression encoding.

[0050] Backup feedback mode refers to a feedback mode that user equipment abandons in abnormal situations (i.e., the main feedback mode) and adopts as an emergency backup. Abnormal situations can refer to situations where the inference performance of the above-mentioned neural network-based encoding model is abnormal (the degree of deviation is too large). When the inference performance is abnormal, the user equipment abandons the encoding model and switches to a traditional encoder based on codebook specifications for traditional compression encoding.

[0051] Optionally, if the accuracy meets preset conditions, the user device will maintain the currently used primary feedback mode as the feedback mode.

[0052] 203. User equipment feeds back uplink control information to network equipment.

[0053] Uplink Control Information (UCI) refers to the real-time status report sent by the UE to the network device via an uplink channel (such as the Physical Uplink Control Channel (PUCCH) or the Physical Uplink Shared Channel (PUSCH)). It is one of the most important uplink signaling protocols at the physical layer. RRC signaling is the "superior" information to UCI, and network devices (such as base stations) can pre-configure the UCI format via RRC signaling—for example, how often CSI should be reported, how many bits to use for quantization, and where the PUCCH resources are located. UCI, on the other hand, is a real-time status report executed strictly according to the guidance of RRC signaling.

[0054] Optionally, the user equipment (UE) may feed back uplink control information to the network device in the following ways: the uplink control information fed back by the UE to the network device may include first control information and second control information; the first control information is a high-priority transmission portion, and the second control information is a low-priority transmission portion; the UE may set a feedback mode flag bit in the first control information; when the feedback mode is a backup feedback mode, the UE configures a backup feedback mode flag for the flag bit; when the feedback mode is a primary feedback mode, the UE configures a primary feedback mode flag for the flag bit; for the second control information, if the flag bit is configured with a backup feedback mode flag, the UE configures a second encoding result for the second control information; if the flag bit is configured with a primary feedback mode flag, the UE configures a first encoding result for the second control information.

[0055] Optionally, the user equipment may set the flag bit of the feedback mode for the first control information included in the uplink control information in the following ways: the user equipment determines the target number of bits and sets the flag bit of the feedback mode in the first control information included in the uplink control information according to the target number of bits; the target number of bits is less than or equal to a preset threshold.

[0056] 204. The network device resolves the feedback mode identifier carried in the uplink control information.

[0057] The feedback mode identifier can refer to a flag used to distinguish the type of feedback mode. For example, one flag can be used to declare that the current feedback mode is the primary feedback mode, and another flag can be used to declare that the current feedback mode is the backup feedback mode.

[0058] Optionally, the network device may parse the feedback mode identifier carried in the uplink control information by parsing the feedback mode identifier from the feedback mode identifier bit set in the first control information included in the uplink control information.

[0059] 205. If the feedback mode identifier is determined to be the primary feedback mode identifier, then the decoding mode for the feedback channel state information will remain the currently used primary decoding mode.

[0060] The identifier for the primary feedback mode can be a flag used to declare that the user device is currently using the primary feedback mode.

[0061] Decoding mode can refer to the way network devices decode and restore the compressed encoding result of the initial CSI measured in the UCI sent by user equipment.

[0062] 206. If the identifier of the feedback mode is determined to be the identifier of the backup feedback mode, then the decoding mode will be switched from the main decoding mode to the backup decoding mode.

[0063] The main decoding mode is a mode in which a neural network-based decoding model is invoked to decode the first encoding result; the first encoding result is the encoding result of the initial channel state information measured by the user equipment by the neural network-based encoding model.

[0064] The backup decoding mode is the mode of calling a decoder based on the codebook specification to decode the second encoding result; the second encoding result is the precoding matrix index generated by encoding the initial channel state information based on the codebook specification.

[0065] The identifier for the backup feedback mode can be a flag used to declare that the user equipment is currently using a backup feedback mode.

[0066] The primary decoding mode refers to the conventional decoding mode that network devices maintain under normal circumstances. Normal circumstances mean that the inference performance of the neural network-based encoding model on the user equipment side is normal. When the inference performance is normal, the user equipment always uses this encoding model for compression encoding, and correspondingly, the network device uses the neural network-based decoding model corresponding to this encoding model for decoding and restoration.

[0067] Backup decoding mode refers to a backup decoding mode used by network devices in abnormal situations, where the main decoding mode is abandoned. Abnormal situations can refer to situations where the inference performance of the neural network-based encoding model on the user equipment side is abnormal (excessive deviation). In such cases, the user equipment abandons the encoding model and switches to a traditional encoder based on the codebook specification for conventional compression encoding. Correspondingly, the network device uses a traditional decoder based on the codebook specification, corresponding to the encoder, for decoding and restoration.

[0068] It is understandable that the primary feedback mode on the user equipment side corresponds to the primary decoding mode on the network equipment side, and the backup feedback mode on the user equipment side corresponds to the backup decoding mode on the network equipment side. That is, when the user equipment is currently using the primary feedback mode, the network equipment will use the primary decoding mode accordingly, and when the user equipment is using the backup feedback mode, the network equipment will use the backup decoding mode accordingly.

[0069] Optionally, the network device may decode the first encoded result in the main decoding mode by parsing the first encoded result from the second control information included in the uplink control information while maintaining the main decoding mode, and calling a neural network-based decoding model to decode the parsed first encoded result.

[0070] Optionally, the network device may decode the second encoding result in the standby decoding mode in the following ways: when the decoding mode is switched to the standby decoding mode, the network device parses the second encoding result from the second control information included in the uplink control information and calls a decoder based on the codebook specification to decode the parsed second encoding result.

[0071] It should be noted that the order of steps 205 and 206 is not limited.

[0072] In this embodiment of the application, the mode of user equipment feedback channel status information is divided into two modes: main feedback mode and backup feedback mode.

[0073] The user equipment measures the initial channel state information between itself and the network equipment, and monitors the accuracy of the first coding result output by the neural network-based coding model in response to the input initial channel state information. In main feedback mode, the uplink control information fed back by the user equipment to the network equipment carries the first coding result and the main feedback mode identifier, which instructs the network equipment to decode based on the first coding result.

[0074] When the accuracy does not meet the preset conditions, the user equipment switches the feedback mode from the currently used primary feedback mode to the backup feedback mode. In the backup feedback mode, the uplink control information fed back by the user equipment to the network equipment carries the second encoding result and the identifier of the backup feedback mode. The identifier of the backup feedback mode is used to instruct the network equipment to decode based on the second encoding result, which is the precoding matrix index generated based on the codebook specification for encoding the initial channel state information.

[0075] As can be seen, the user equipment (UE) autonomously decides the current feedback mode by monitoring the accuracy of the first encoding result output by the neural network-based encoding model. It efficiently instructs the network device to adopt the corresponding decoding mode by configuring an identifier for the currently decided feedback mode in the feedback uplink control information. When the accuracy does not meet preset conditions, the UE autonomously and efficiently switches to the backup feedback mode, using the traditional precoding matrix index to provide feedback to the network device instead of the output of the neural network-based encoding model. This switching process does not rely on RRC signaling, significantly shortening the reconfiguration time. Furthermore, during reconfiguration, the network device avoids using the feedback result with a large error from the encoding model, instead using the relatively more reliable traditional precoding matrix index for subsequent decoding, minimizing communication interruptions.

[0076] The following describes another embodiment of the communication connection between the user equipment and the network equipment, in which the user equipment feeds back the measured CSI to the network equipment in different feedback modes under different circumstances. For the specific implementation process of this embodiment, please refer to [link to relevant documentation]. Figure 3 As shown, Figure 3 This is a flowchart illustrating a method for switching channel state information feedback modes according to another embodiment of this application. Another embodiment of the method for switching channel state information feedback modes provided in this application may include: 301. User equipment measures the initial channel state information between the user equipment and the network equipment.

[0077] Optionally, the method by which the user equipment measures the initial channel state information between itself and the network equipment may include: the user equipment measuring downlink reference signals (such as CSI-RS) to obtain the true channel estimation matrix at the current moment. (Initial channel state information), which serves as the benchmark for evaluating the performance of the coding model (Ground Truth).

[0078] 302. The user equipment obtains the first coding result output by the neural network-based coding model in response to the input initial channel state information.

[0079] Optionally, the user equipment may obtain the first encoding result in the following ways: The user equipment will use the above-mentioned true channel estimation matrix (Initial channel state information) is input into the AI ​​encoder (a coding model based on a neural network), which can produce the output of the AI ​​encoder, i.e., the first coding result.

[0080] 303. The user equipment calls the second decoding model, uses the first encoding result to obtain the output result of the second decoding model, and determines the accuracy of the first encoding result based on the deviation between the output result and the initial channel state information.

[0081] The network equipment is equipped with a first decoding model based on a neural network, and the user equipment is equipped with a second decoding model that is a copy of the first decoding model; the second decoding model and the first decoding model are able to output decoding results based on the encoding results of the input channel state information.

[0082] Optionally, the user equipment can invoke the second decoding model by deploying the second decoding model locally on the user equipment. When the user equipment obtains the first encoding result, it can invoke the second decoding model locally for decoding verification.

[0083] Optionally, the user equipment may obtain the output of the second decoding model using the first encoding result in the following manner: the user equipment runs a lightweight AI decoder (second decoding model) locally to verify the accuracy of the encoding result, inputs the first encoding result into the AI ​​decoder, and obtains the output of the AI ​​decoder, i.e., the reconstructed channel matrix. .

[0084] Optionally, the method by which the user equipment determines the accuracy of the first coding result based on the deviation between the output result and the initial channel state information may include: the user equipment calculating... and Similarity index between This similarity index This reflects the current AI model's (the aforementioned AI encoder) ability to reconstruct channel characteristics.

[0085] in, That is, the current-time true channel estimation matrix (initial channel state information) obtained by the user equipment measuring the downlink reference signal (such as CSI-RS). That is, the user equipment inputs the first encoding result into the AI ​​decoder, and the output result of the AI ​​decoder is the reconstructed channel matrix.

[0086] Optional, similarity metrics It can refer to cosine similarity, or normalized mean square error (NMSE).

[0087] Optionally, when the similarity index When using cosine similarity, the similarity index is... This can be expressed by the following formula: (1) in, That is, the above-mentioned true channel estimation matrix (initial channel state information). That is, the reconstructed channel matrix mentioned above.

[0088] 304. The user equipment determines the target number of bits and assigns an identifier bit in the first control information included in the uplink control information according to the target number of bits.

[0089] The target number of bits is less than or equal to a preset threshold.

[0090] In computer science and digital communication, a bit is the most basic unit of data, which takes the value of 0 or 1. The number of bits refers to how many 0s or 1s are used to encode a specific piece of information.

[0091] Correspondingly, the target number of bits can be the number of bits that the designated identifier bits are expected to occupy in the first control information included in the uplink control information.

[0092] A flag bit can be one or more bits in a communication protocol or data structure used to indicate the existence of specific information, its status, or how to parse subsequent content. Essentially, it is a label or switch that tells network devices "what to pay attention to next" or "what this field means."

[0093] The preset threshold can be a preset number of bits. For example, if the preset number of bits is 1, then the target number of bits is less than or equal to 1.

[0094] As an example, the uplink control information fed back by the user equipment to the network equipment includes first control information and second control information; the first control information is the high-priority transmission part, and the second control information is the low-priority transmission part; the user equipment sets a feedback mode flag bit for the first control information; when the feedback mode is the backup feedback mode, the user equipment configures the backup feedback mode flag bit; when the feedback mode is the primary feedback mode, the user equipment configures the primary feedback mode flag bit; for the second control information, if the flag bit is configured with the backup feedback mode flag, the user equipment configures the second encoding result for the second control information; if the flag bit is configured with the primary feedback mode flag, the user equipment configures the first encoding result for the second control information.

[0095] 305. If the accuracy meets the preset conditions, the user equipment will maintain the feedback mode as the currently used main feedback mode.

[0096] Optionally, the method by which the user equipment determines whether the accuracy meets the preset conditions may include: the user equipment calculating the above-mentioned accuracy... With the above Similarity index between Compared with a preset performance threshold, if If the accuracy meets the preset conditions, then the AI ​​encoder (a neural network-based encoding model) is considered to be working normally; if If the above AI encoder fails (for example, due to environmental changes causing a decrease in the model's generalization ability, resulting in inference performance failure), then it is determined that the above AI encoder has failed.

[0097] Optionally, the way the user device maintains the feedback mode as the currently used primary feedback mode may include: in the aforementioned similarity metrics If the accuracy meets the preset conditions, the AI ​​encoder is determined to be working normally. At this time, the user equipment will keep the feedback mode as the currently used main feedback mode and set the flag position of the feedback mode of the first control information (Part 1) of the uplink control information to be fed back to the network device to 0. At the same time, the user equipment will directly use the output result of the AI ​​encoder, i.e. the first encoding result, as the payload of the second control information (Part 2) of the uplink control information.

[0098] 306. If the accuracy does not meet the preset conditions, the user equipment will switch the feedback mode from the currently used primary feedback mode to the backup feedback mode.

[0099] In the backup feedback mode, the uplink control information fed back to the network device carries the second encoding result and the identifier of the backup feedback mode; the identifier of the backup feedback mode is used to instruct the network device to decode based on the second encoding result; the second encoding result is the precoding matrix index generated by encoding the initial channel state information based on the codebook specification; In the main feedback mode, the uplink control information fed back to the network device carries the first encoding result and the main feedback mode identifier; the main feedback mode identifier is used to instruct the network device to decode based on the first encoding result.

[0100] Optionally, the user device may switch the feedback mode from the currently used primary feedback mode to the backup feedback mode in the following ways: if the above similarity indicators If the user equipment determines that the AI ​​encoder has failed, it will switch the feedback mode from the currently used primary feedback mode to the backup feedback mode and set the flag of the feedback mode of the first control information (Part 1) of the uplink control information to be fed back to the network device to 1. At the same time, the user equipment calls the traditional Type I codebook generation algorithm (codebook specification) to encode the initial channel state information to generate a precoding matrix index (PMI) and uses it as the payload (second encoding result) of the second control information (Part 2) of the uplink control information.

[0101] As an example, user devices can use the aforementioned similarity metrics. After N time slots, the feedback mode is switched to the backup feedback mode.

[0102] Optionally, the codebook specification mentioned above can be a codebook specification that conforms to 3GPP Type I or a codebook specification that conforms to 3GPP Type II.

[0103] 307. User equipment feeds back uplink control information to network equipment.

[0104] Optionally, the user equipment (UE) may feed back uplink control information to the network device in the following ways: the UE assembles Part 1, which contains the aforementioned flag bit, and Part 2, which contains the payload corresponding to the flag bit, into a complete uplink control information data packet, and sends it to the network device via PUCCH or PUSCH. This transmission process belongs to the "L1 handshake" process, which only involves the transmission of physical layer data and does not involve the interaction of higher-layer protocol stacks (no higher-layer signaling intervention is required). This process is executed within each CSI reporting cycle or a preset monitoring cycle. With this design, the UE does not need to wait for the network device to issue complex RRC reconfiguration signaling; it only needs to flip one bit corresponding to the flag bit to complete the protocol switch from AI mode (primary feedback mode) to traditional mode (backup feedback mode) in milliseconds (within a single slot), ensuring service continuity.

[0105] 308. The network device parses the feedback mode identifier from the identifier bit set in the first control information included in the uplink control information.

[0106] Optionally, the first control information (Part 1) has a fixed payload size and high coding protection. It can be of fixed length and, in addition to the aforementioned identifier bits, can also carry information that is most critical to the link. For example, it can also include a Rank Indicator (RI) and a Channel Quality Indicator (CQI).

[0107] 309. If the identifier of the feedback mode is determined to be the identifier of the primary feedback mode, the network device will maintain the decoding mode of the feedback channel state information as the currently used primary decoding mode, and while maintaining the decoding mode as the primary decoding mode, it will parse the first encoding result from the second control information included in the uplink control information.

[0108] Optionally, after parsing the first encoding result from the second control information included in the uplink control information, the network device can decode the first encoding result and perform downlink beamforming based on the decoding result.

[0109] 310. If the identifier of the feedback mode is determined to be the identifier of the backup feedback mode, the network device will switch the decoding mode from the primary decoding mode to the backup decoding mode, and when the decoding mode is switched to the backup decoding mode, the second encoding result will be parsed from the second control information included in the uplink control information.

[0110] The main decoding mode is a mode in which a neural network-based decoding model is invoked to decode the first encoding result; the first encoding result is the encoding result of the initial channel state information measured by the user equipment by the neural network-based encoding model. The backup decoding mode is the mode of calling a decoder based on the codebook specification to decode the second encoding result; the second encoding result is the precoding matrix index generated by encoding the initial channel state information based on the codebook specification.

[0111] Optionally, the second control information (Part 2) carries detailed channel characteristics, and the payload size is variable.

[0112] Optionally, after parsing the second encoding result from the second control information included in the uplink control information, the network device can decode the second encoding result and perform downlink beamforming based on the decoding result.

[0113] It should be noted that the order of step 304 with steps 301 to 303 is not limited. The order of step 304 with steps 305 to 306 is not limited. The order of step 305 with step 306 is not limited. The order of step 309 with step 310 is not limited.

[0114] The following example illustrates the feedback mode switching process: Taking a network device as a base station (gNB) as an example, consider the following scenario: The user device suddenly enters the underground parking garage (a scenario where the AI ​​encoder has not been specifically trained) from the outdoors (a scenario where the AI ​​encoder has not been specifically trained).

[0115] At time T0: The user device detected that the AI ​​encoder compression coding error surged from 5% to 40%.

[0116] At time T1 (the next CSI reporting window): the user equipment decides not to send the erroneous AI encoder output data, but instead generates a Type I single codebook (low precision but robust), and sends a UCI to the base station based on this: [Part 1: RI=2, CQI=5, ASF=1] | [Part 2: i1=12, i2=3].

[0117] At time T2: The base station receives the UCI signal and analyzes Part 1 to find that the feedback mode flag (AI_Status_Flag, ASF) of the flag bit is 1. At this time, although the base station originally expected to receive 100 bits of AI encoded features (first encoding result), the base station switched to analyzing the Type I codebook (which may only have 20 bits) according to the protocol transformation. The base station used the parsed Type I PMI to perform downlink beamforming. At this time, the service was not interrupted, but only downgraded from "high-precision beam" to "normal beam", thus avoiding dropped calls.

[0118] Corresponding to the above method embodiments, this application also provides a channel state information feedback mode switching system, such as... Figure 4 As shown, Figure 4 This is a schematic diagram of a switching system for channel state information feedback mode according to an embodiment of this application. The system includes: user equipment 401 and network device 402 communicatively connected to user equipment 401. User equipment 401 is used to measure the initial channel state information between itself and network equipment 402, and to monitor the accuracy of the first encoding result output by the neural network-based coding model for the input initial channel state information. If the accuracy does not meet the preset conditions, the feedback mode is switched from the currently used primary feedback mode to the backup feedback mode. In the backup feedback mode, the uplink control information fed back to network equipment 402 carries the second encoding result and the identifier of the backup feedback mode. The identifier of the backup feedback mode is used to instruct network equipment 402 to decode based on the second encoding result. The second encoding result is a precoding matrix index generated based on the codebook specification for encoding the initial channel state information. In the primary feedback mode, the uplink control information fed back to network equipment 402 carries the first encoding result and the identifier of the primary feedback mode. The identifier of the primary feedback mode is used to instruct network equipment 402 to decode based on the first encoding result. Network device 402 is used to receive uplink control information fed back by user equipment 401 and parse out the identifier of the feedback mode carried in the uplink control information; if the identifier of the feedback mode is determined to be the identifier of the primary feedback mode, the decoding mode of the feedback channel state information is kept as the currently used primary decoding mode; if the identifier of the feedback mode is determined to be the identifier of the backup feedback mode, the decoding mode is switched from the primary decoding mode to the backup decoding mode; wherein, the primary decoding mode is the mode of calling a decoding model based on a neural network to decode the first encoding result; the backup decoding mode is the mode of calling a decoder based on a codebook specification to decode the second encoding result.

[0119] Optionally, network device 402 is deployed with a first decoding model based on a neural network, and user equipment 401 is deployed with a second decoding model as a copy of the first decoding model; the second decoding model and the first decoding model are capable of outputting decoding results based on the encoding results of the input channel state information; Accordingly, user equipment 401 is specifically used to call the second decoding model, use the first encoding result to obtain the output result of the second decoding model, and determine the accuracy based on the deviation between the output result and the initial channel state information.

[0120] Optionally, the uplink control information fed back by user equipment 401 to network equipment 402 includes first control information and second control information; the first control information is a high-priority transmission part, and the second control information is a low-priority transmission part; the first control information has an identifier bit for the feedback mode; when the feedback mode is a standby feedback mode, the identifier bit is configured with the identifier of the standby feedback mode; when the feedback mode is the primary feedback mode, the identifier bit is configured with the identifier of the primary feedback mode; the second control information, when the identifier bit is configured with the identifier of the standby feedback mode, configures the second encoding result; when the identifier bit is configured with the identifier of the primary feedback mode, configures the first encoding result.

[0121] Optionally, the user equipment 401 is also used to determine the target number of bits and to define an identifier bit in the first control information according to the target number of bits; the target number of bits is less than or equal to a preset threshold.

[0122] Optionally, user equipment 401 is also configured to maintain the currently used main feedback mode in the feedback mode if the accuracy meets preset conditions.

[0123] Optionally, network device 402 is specifically used to parse the identifier of the feedback mode from the identifier bit; network device 402 is also used to parse the first encoding result from the second control information when the decoding mode remains the primary decoding mode; and to parse the second encoding result from the second control information when the decoding mode is switched to the backup decoding mode.

[0124] like Figure 5 As shown, this application embodiment also provides an electronic device, including a processor 501, a memory 502, and a program or instructions stored in the memory 502 and executable on the processor 501. When the program or instructions are executed by the processor 501, they implement the various processes of the above-described channel state information feedback mode switching method embodiment and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0125] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described channel state information feedback mode switching method embodiment and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0126] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0127] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0128] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0129] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

[0130] The above description is merely a preferred embodiment and the technical principles employed in this application. This application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions that can be made by those skilled in the art will not depart from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of this application, the scope of which is determined by the scope of the claims.

Claims

1. A method for switching channel state information feedback modes, applied to a user equipment, wherein the user equipment is communicatively connected to a network device, characterized in that, The method includes: Measure the initial channel state information between the network device and the network device, and monitor the accuracy of the first encoding result output by the neural network-based encoding model in response to the input initial channel state information; If the accuracy does not meet the preset conditions, the feedback mode will be switched from the currently used main feedback mode to the backup feedback mode. In the backup feedback mode, the uplink control information fed back to the network device carries a second encoding result and an identifier of the backup feedback mode; the identifier of the backup feedback mode is used to instruct the network device to decode based on the second encoding result; the second encoding result is a precoding matrix index generated based on the codebook specification for encoding the initial channel state information; In the primary feedback mode, the uplink control information fed back to the network device carries the first encoding result and the identifier of the primary feedback mode; the identifier of the primary feedback mode is used to instruct the network device to decode based on the first encoding result.

2. The method according to claim 1, characterized in that, The network device is equipped with a first decoding model based on a neural network, and the user equipment is equipped with a second decoding model that is a copy of the first decoding model; the second decoding model and the first decoding model are capable of outputting decoding results based on the encoding results of the input channel state information; Accordingly, the accuracy is determined in the following ways: The second decoding model is invoked, and the output result of the second decoding model is obtained using the first encoding result. The accuracy is determined based on the deviation between the output result and the initial channel state information.

3. The method according to claim 1, characterized in that, The uplink control information fed back to the network device includes first control information and second control information; the first control information is a high-priority transmission part, and the second control information is a low-priority transmission part; The first control information includes an identifier bit for the feedback mode; when the feedback mode is the backup feedback mode, the identifier bit is configured with an identifier for the backup feedback mode; when the feedback mode is the primary feedback mode, the identifier bit is configured with an identifier for the primary feedback mode. The second control information configures the second encoding result when the identifier bit is configured to indicate the backup feedback mode; When the identifier of the main feedback mode is configured in the identifier bit, the first encoding result is configured.

4. The method according to claim 3, characterized in that, The method further includes: Determine the target number of bits and define the identifier bit in the first control information according to the target number of bits; the target number of bits is less than or equal to a preset threshold.

5. The method according to claim 1, characterized in that, After monitoring the accuracy of the first coding result output by the neural network-based coding model for the input initial channel state information, the method further includes: If the accuracy meets the preset conditions, the feedback mode remains the currently used main feedback mode.

6. A method for switching channel state information feedback modes, applied to a network device, wherein the network device is communicatively connected to a user equipment, characterized in that, The method includes: Receive uplink control information fed back by the user equipment, and parse out the identifier of the feedback mode carried in the uplink control information; If the identifier of the feedback mode is determined to be the identifier of the main feedback mode, then the decoding mode of the channel state information being fed back will remain the currently used main decoding mode. If the identifier of the feedback mode is determined to be the identifier of the backup feedback mode, then the decoding mode is switched from the main decoding mode to the backup decoding mode; The main decoding mode is a mode in which a neural network-based decoding model is invoked to decode the first encoding result; the first encoding result is the encoding result of the initial channel state information measured by the user equipment by the neural network-based encoding model. The backup decoding mode is a mode that calls a decoder based on the codebook specification to decode the second encoding result; the second encoding result is a precoding matrix index generated by encoding the initial channel state information based on the codebook specification.

7. The method according to claim 6, characterized in that, The uplink control information includes first control information and second control information; the first control information is a high-priority transmission part, and the second control information is a low-priority transmission part; the first control information is equipped with an identifier bit for the feedback mode; The second control information configures the second encoding result when the identifier bit is configured to indicate the backup feedback mode; When the identifier of the main feedback mode is configured in the identifier bit, the first encoding result is configured; Accordingly, parsing the identifier of the feedback mode carried in the uplink control information includes: The identifier of the feedback mode is parsed from the identifier bit; Accordingly, the method further includes: With the decoding mode remaining in the main decoding mode, the first encoding result is parsed from the second control information; When the decoding mode is switched to the backup decoding mode, the second encoding result is parsed from the second control information.

8. A switching system for channel state information feedback mode, characterized in that, The system includes user equipment and network devices that are communicatively connected to the user equipment. The user equipment is configured to measure initial channel state information (ACSI) with the network device and monitor the accuracy of a first encoding result output by a neural network-based encoding model for the input ACSI. If the accuracy does not meet a preset condition, the feedback mode is switched from the currently used primary feedback mode to a backup feedback mode. In the backup feedback mode, the uplink control information fed back to the network device carries a second encoding result and an identifier of the backup feedback mode. The identifier of the backup feedback mode is used to instruct the network device to decode based on the second encoding result. The second encoding result is a precoding matrix index generated based on the codebook specification for encoding the ACSI. In the primary feedback mode, the uplink control information fed back to the network device carries the first encoding result and an identifier of the primary feedback mode. The identifier of the primary feedback mode is used to instruct the network device to decode based on the first encoding result. The network device is configured to receive uplink control information fed back by the user equipment and parse out the identifier of the feedback mode carried in the uplink control information; if the identifier of the feedback mode is determined to be the identifier of the primary feedback mode, the decoding mode for the feedback channel state information is maintained as the currently used primary decoding mode; if the identifier of the feedback mode is determined to be the identifier of the backup feedback mode, the decoding mode is switched from the primary decoding mode to the backup decoding mode; wherein, the primary decoding mode is a mode that calls a decoding model based on a neural network to decode the first encoding result; the backup decoding mode is a mode that calls a decoder based on a codebook specification to decode the second encoding result.

9. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein when the program or instructions are executed by the processor, they implement the method for switching the channel state information feedback mode as described in any one of claims 1-7.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions, which, when executed by a processor, implement the method for switching the channel state information feedback mode as described in any one of claims 1-7.

Images