Channel feedback method, apparatus and system

Abstract

A channel feedback method, apparatus and system. The method comprises: acquiring multipath component information of a communication channel between a first device and a second device; on the basis of a self-supervised neural network, performing processing on the multipath component information and second channel information, so as to obtain intermediate information; and sending the intermediate information, wherein the second channel information is determined on the basis of first channel information and multipath delay information, the second channel information comprises a variable channel impulse response (CIR), the first channel information comprises a channel frequency response (CFR), which is obtained by means of a measurement, and the intermediate information and the multipath component information are used for determining a reconstructed CIR. By means of the solution, channel feedback is performed in combination with multipath component information and a self-supervised neural network, thereby reducing measurement overheads and feedback overheads.

Description

A channel feedback method, apparatus and system

[0001] The present application claims priority to the Chinese patent application No. 202411171931.8, filed on August 23, 2024, and entitled "A channel feedback method, apparatus and system", the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0002] The present application relates to the field of communication, and more particularly, to a communication, apparatus and system for channel feedback. BACKGROUND

[0003] Different communication devices can communicate using multi-input multi-output (MIMO) technology, in which channel information can be used to meet the demand for high-speed transmission. For example, a communication device can use precoding information corresponding to channel information to perform high-speed data transmission; for another example, a communication device can use channel information to perform multi-user resource allocation, which can reduce the interference between different users and improve the overall system performance. Generally, channel information is obtained through measurement of a reference signal, and the overhead of the reference signal is related to the number of ports through which the communication device transmits the reference signal.

[0004] However, as the frequency band increases and the demand for high-speed communication increases, the number of ports through which the communication device transmits the reference signal is likely to increase, which will increase the overhead of the reference signal used to obtain channel information and occupy more transmission resources, thereby increasing the power consumption of the communication device. SUMMARY

[0005] The present application provides a channel feedback method, apparatus and system to reduce the measurement overhead and feedback overhead.

[0006] In a first aspect, a channel feedback method is provided. The method can be performed by a first device. Unless otherwise specified, the first device in the present application can refer to a communication device (e.g., a terminal device or a network device), a component (e.g., a communication module, a processor, a circuit, a chip, or a chip system) in the communication device, or a logic module or software that can realize all or part of the functions of the communication device.

[0007] The method comprises: obtaining composition information of a multipath of a communication channel between the first device and the second device; processing the composition information of the multipath and second channel information based on a self-supervised neural network to obtain intermediate information; and transmitting the intermediate information; wherein the second channel information is determined based on first channel information and time delay information of the multipath, the second channel information comprises a variable channel impulse response (CIR), the first channel information comprises channel frequency response (CFR) measured based on a reference signal transmitted on the communication channel, the time delay information of the multipath belongs to the composition information of the multipath, and the intermediate information and the composition information of the multipath are used to determine a reconstructed CIR.

[0008] Based on the above scheme, the first device obtains the composition information of the multipath of the communication channel between the first device and the second device, processes the composition information of the multipath and the second channel information based on a self-supervised neural network to obtain intermediate information, and feeds back the intermediate information to the second device, so as to facilitate the second device to subsequently reconstruct the CIR based on the intermediate information and the composition information of the multipath. The scheme combines the composition information of the multipath and the self-supervised neural network, can reduce the measurement overhead (for example, sparse measurement based on a reference signal) and the feedback overhead (for example, only intermediate information in channel information is fed back), and thus improves resource utilization and reduces device power consumption.

[0009] Exemplarily, the first device can determine the composition information of the multipath in a local manner through wireless signal sensing, reference signal measurement, ray-tracing, artificial intelligence (AI), or other manners; or the first device can receive the composition information of the multipath from the second device.

[0010] In this application, the self-supervised neural network comprises an encoder and a decoder, and can be a pre-trained network, that is, the data input to the self-supervised neural network can also be used as the output of the self-supervised neural network. The encoder is used to quantitatively compress the CIR, and the decoder is used to reconstruct the CIR. Specifically, the encoder is used to compress the second channel information (for example, the variable CIR) to obtain the intermediate information, and the decoder is used to output the reconstructed CIR. That is, the self-supervised neural network is used to compress the variable CIR into an intermediate representation, and then reconstruct the CIR using the intermediate representation, to realize the compression and reconstruction process.

[0011] In the present application, the self-supervised neural network can be deployed on the first device or the second device, i.e., the encoding module and the decoding module in the supervised neural network are deployed on the same device, such as the first device or the second device; or the encoding module in the self-supervised neural network is deployed on the first device, and the decoding module of the self-supervised neural network is deployed on the second device; or the encoding module in the supervised neural network is deployed on the second device, and the decoding module of the self-supervised neural network is deployed on the first device, without limitation. For ease of description, in the embodiments of the present application, the first device and the second device both deploy the self-supervised neural network.

[0012] The second channel information is determined based on the first channel information and the delay information of the multipath, which can be understood as: the first device determines the second channel information based on the first channel information and the delay information of the multipath; or the second device determines the second channel information based on the first channel information and the delay information of the multipath, and sends the second channel information to the first device, which is used for the first device to process the composition information of the multipath and the second channel information based on the self-supervised neural network to obtain the intermediate information.

[0013] In the present application, the first channel information is obtained by measuring a reference signal (such as a pilot reference signal for channel measurement) transmitted on the communication channel, for example, the first channel information can be the actually measured CFR.

[0014] As an example, for the downlink channel estimation scenario, the network device sends a pilot reference signal, such as a channel state information-reference signal (CSI-RS), to the terminal device, the terminal device receives and measures the CSI-RS to obtain a channel measurement result, which includes the actually measured CFR. Optionally, if the second channel information is determined by the network device based on the first channel information and the delay information of the multipath, the terminal device also sends the actually measured CFR to the network device; if the second channel information is determined by the terminal device based on the first channel information and the delay information of the multipath, the terminal device does not need to send the actually measured CFR to the network device, reducing signaling overhead.

[0015] As an example, for an uplink channel estimation scenario, the terminal device sends a pilot reference signal, such as a sounding reference signal (SRS), to the network device, the network device receives and measures the SRS to obtain a channel measurement result, which includes the actually measured CFR. Optionally, if the second channel information is determined by the network device based on the first channel information and the delay information of the multipath, the network device does not need to send the actually measured CFR to the terminal device; if the second channel information is determined by the terminal device based on the first channel information and the delay information of the multipath, the network device also sends the actually measured CFR to the terminal device.

[0016] It can be understood that using a self-supervised neural network to reconstruct the CIR can ensure the rationality of the reconstructed CIR under the composition information of the multipath.

[0017] Exemplarily, the composition information of the multipath can be a multipath component (MPC) when the first device is connected to the second device, also known as a deterministic part of the multipath, such as at least one of the number, strength, direction of departure (DoD), direction of arrival (DoA), power, and delay of the multipath.

[0018] Exemplarily, the intermediate information includes phase information of the multipath, and other information of the communication channel, such as one or more of frequency shift, time offset, frequency offset, and instantaneous channel error of the communication channel. Wherein, the phase information of each path can include the phase of one or more polarization directions, such as x polarizations at the transmitting end and y polarizations at the receiving end, then the phase information of each path can include the phase of x times y polarizations, for example, the transmitting end includes horizontal and vertical polarizations, and the receiving end also includes horizontal and vertical polarizations, then each path includes 4 polarizations, which correspond to: the combination of the horizontal polarization of the transmitting end and the horizontal polarization of the receiving end, the combination of the vertical polarization of the transmitting end and the horizontal polarization of the receiving end, the combination of the horizontal polarization of the transmitting end and the vertical polarization of the receiving end, and the combination of the vertical polarization of the transmitting end and the vertical polarization of the receiving end.

[0019] In the present application, after any one of the first communication device and the second communication device sends a signal on a communication channel between the first communication device and the second communication device, the signal received by the other communication device can be used to reflect channel characteristic information of the communication channel, which can be used to determine channel information of the communication channel. The channel characteristic information can include information that is likely to remain unchanged for a long time and information that is likely to change in a short time. The former can be referred to as long time (LT) information, and the latter can be referred to as instantaneous (IN) information. In addition, the information can be used to reflect time-angular domain channel property (TADCP) of the communication channel. Therefore, the LT information can also be referred to as TADCP LT information, and the IN information can also be referred to as TADCP IN information.

[0020] As an example, the LT information can be referred to as multipath composition information or MPC, which includes one or more of the number of multipaths, the strength of the multipaths, the angle of the multipaths, and the time delay of the multipaths.

[0021] As an example, the IN information can include phase information of the multipaths, and other information of the communication channel, including but not limited to one or more of frequency shift, time offset, frequency offset, and instantaneous channel error of the communication channel.

[0022] In a MIMO system, both the multipath composition information and the phase information of the multipaths are influencing factors of the channel information. Therefore, compared with a process of determining the channel information only by the multipath composition information, in the above scheme, the determination of the channel information is also based on the phase information of the multipaths, which can improve the accuracy of the channel information determined based on the multipath composition information and the phase information of the multipaths.

[0023] In a possible implementation, the variable CIR includes a first CIR, and the second channel information is determined based on the first channel information and the time delay information of the multipaths, including: the first CIR is determined based on the second CIR, the time delay information of the multipaths, the measured CFR, and the reconstructed CFR.

[0024] Optionally, the second CIR can be replaced by an initial CIR. For example, the first device or the second device can randomly set parameters of the CIR to determine the initial CIR, and update the parameters of the CIR through iterative loops until convergence, that is, the reconstruction of the CIR can be completed. Therefore, the second CIR can be predefined or preconfigured, or randomly set, without limitation.

[0025] Optionally, the second CIR can also be regarded as the previously determined CIR of the first CIR, at this time the second CIR is equivalent to the third CIR; or, the first CIR can be determined by updating the parameters of the second CIR; or, the first CIR can also be determined by updating the parameters of the third CIR which is determined by updating the parameters of the second CIR.

[0026] Exemplarily, the reconstructed CFR satisfies:

[0027] Wherein, CFR' represents the reconstructed CFR, CIR n represents the third CIR, n represents the number of multipaths, f represents the carrier frequency corresponding to the measured CFR, τ represents the time delay information of the multipaths, the first CIR is the CIR obtained by updating the third CIR, or the third CIR is the previously determined CIR of the first CIR.

[0028] Optionally, in the process of reconstructing the CIR and / or the CFR, if it is required to perform one loop iteration, the third CIR can be the initial CIR, at this time the initial CIR (i.e. the second CIR), the number of multipaths, the time delay information of the multipaths and the carrier frequency corresponding to the measured CFR can be collectively input to the right side of the above formula, i.e. the reconstructed CFR'1 can be obtained, the difference value loss1 between the reconstructed CFR'1 and the measured CFR is calculated, the parameters of the initial CIR are updated based on the loss1 by gradient descent, and the updated CIR (i.e. the first CIR) is obtained. It can be understood that with the updating of the parameters of the CIR, the loss will gradually decrease.

[0029] Optionally, in the process of reconstructing the CIR and / or the CFR, if multiple loop iterations are needed to be performed, the first CIR can be the CIR obtained by updating the third CIR, or the third CIR is the previously determined CIR of the first CIR. Assuming that there are two loop iteration processes, for the first loop iteration, the initial CIR (i.e., the second CIR), the number of multipaths, the time delay information of the multipaths, and the carrier frequency corresponding to the measured CFR are collectively input to the right side of the above formula, and the reconstructed CFR'1 is obtained. The difference value loss' between the reconstructed CFR'1 and the measured CFR is calculated, and the parameters of the initial CIR are updated based on the gradient descent of the loss' to obtain the updated CIR (i.e., the third CIR). Then, for the second loop iteration, the third CIR, the number of multipaths, the time delay information of the multipaths, and the carrier frequency corresponding to the measured CFR are collectively input to the right side of the above formula, and the reconstructed CFR'2 is obtained. The difference value loss'' between the reconstructed CFR'2 and the measured CFR is calculated, and the parameters of the second CIR (i.e., the first CIR) are updated based on the gradient descent of the loss''.

[0030] It should be noted that the number of loop iterations in the present application is not limited, and can be one or more times until convergence, i.e., the loop iteration process can be stopped, and the parameters of the current CIR are updated. In other words, the current CIR at this time is the reconstructed wireless channel, for example, the reconstructed CIR. Therefore, the loop iteration process between the above first CIR, second CIR and third CIR is only an example given for understanding, and does not limit the technical solutions of the present application, for example, the fourth CIR or the fifth CIR can also be included in the loop iteration process.

[0031] It can be understood that the above formula can be regarded as a channel expert knowledge (CEK) model, or a differentiable time-frequency domain conversion expert knowledge model. The variables (e.g., CIR n , n, f or τ) involved in the formula can be fine-tuned when matching the actual channel. By adjusting these variables, the reconstructed CFR can be determined, and the wireless channel can be reconstructed.

[0032] In a possible implementation, the self-supervised neural network includes an encoding module and a decoding module, and the self-supervised neural network is used to process the composition information of the multipath and the second channel information to obtain intermediate information, including: inputting the composition information of the multipath and the second channel information into the encoding module to output the intermediate information; wherein the intermediate information includes phase information of the multipath, and the decoding module is used to output the reconstructed CIR.

[0033] Based on the above scheme, the first device inputs the composition information of the multipath and the second channel information into the encoding module through the self-supervised neural network, outputs the intermediate information (or intermediate representation), and combines the decoding module to output the reconstructed CIR, realizes the self-supervised neural network-based channel feedback of the set MPC, and can obtain effective reconstructed channel information.

[0034] In a possible implementation, the reconstructed CIR is determined when the weighted sum is less than or equal to a preset threshold, and the weighted sum satisfies: Loss = W1 x loss1 + W2 x loss2.

[0035] Wherein, Loss represents the weighted sum, W1 represents the first weight, W2 represents the second weight, loss1 represents the first difference value between the reconstructed CFR and the measured CFR, and loss2 represents the second difference value between the reconstructed CIR and the variable CIR, the first difference value and the second difference value are both used to update the variable CIR, and the first weight and the second weight are both not 0.

[0036] Based on the above scheme, the channel information is reconstructed by comprehensively considering the sparse measurement and the self-supervised neural network-based channel feedback. By comparing the size of the weighted sum and the preset threshold, the accuracy of the predicted channel information can be improved.

[0037] In a possible implementation, the method further includes: receiving first configuration information, the first configuration information being used to indicate that the second device supports self-supervised channel feedback (SSF) corresponding to the self-supervised neural network.

[0038] Exemplarily, the first configuration information includes at least one of the following: information of at least one model (or model library, or model set, or model combination) supported by the second device; a first field, the first field being used to indicate whether the first device starts the SSF; a second field, the second field being used to indicate that the channel information is fed back by using the SSF, the channel information being measured based on a reference signal transmitted on a communication channel; or a third field, the third field being used to indicate an antenna configuration supported by the at least one model.

[0039] Exemplarily, the self-supervised neural network can be an implementation of the at least one model supported by the second device. Different models can have different input and output dimensions or neural network parameters. It can be understood that the neural network is a complex network composed of neurons. It is a mathematical model constructed based on artificial neurons or bionic neurons, and can simulate the learning and cognitive process of the human brain.

[0040] Based on the above scheme, by receiving the first configuration information, it can be determined that the second device supports the SSF corresponding to the self-supervised neural network, and then based on the composition information of the multipath, the self-supervised neural network is used for channel feedback to realize channel reconstruction.

[0041] In a possible implementation, the method further includes: sending capability information, the capability information indicating that the first device supports the SSF corresponding to the self-supervised neural network, and / or the capability information indicating a first model supported by the first device, the first model belonging to at least one model corresponding to the self-supervised neural network.

[0042] Based on the above scheme, by sending the capability information to indicate that the first device supports the SSF corresponding to the self-supervised neural network, in combination with the second device supporting the SSF corresponding to the self-supervised neural network, therefore, the first device and the second device can use the self-supervised neural network for channel feedback based on the composition information of the multipath to realize channel reconstruction. In addition, by sending the capability information, it can be indicated that the first device and the second device simultaneously support the first model, that is, the first device and the second device can use the first model for channel feedback and channel reconstruction subsequently.

[0043] In a possible implementation, the method further includes: receiving a monitoring reference signal, the monitoring reference signal being used to monitor the performance of the SSF; and in a case where the performance of the SSF does not meet a requirement, sending monitoring response information.

[0044] Exemplarily, the performance of the SSF can be represented by the correlation or similarity between the reconstructed CFR and the measured CFR. It can be understood that the higher the correlation or similarity, the smaller the difference between the reconstructed CFR and the measured CFR. If the correlation or similarity is greater than or equal to a preset threshold, it indicates that the reconstructed CFR meets the requirement; otherwise, if the correlation or similarity is less than or equal to the preset threshold, it indicates that the reconstructed CFR does not meet the requirement.

[0045] Exemplarily, the performance of the SSF can be represented by the normalized mean squared error (NMSE) between the reconstructed CFR and the measured CFR, or the NMSE of the two channels corresponding to the reconstructed CFR and the measured CFR. It can be understood that the greater the NMSE, the more dispersed the distribution of the reconstructed CFR and the measured CFR; otherwise, the smaller the NMSE, the more concentrated the distribution of the reconstructed CFR and the measured CFR data. If the NMSE is greater than or equal to a preset threshold, it indicates that the reconstructed CFR does not meet the requirement; otherwise, if the NMSE is less than or equal to the preset threshold, it indicates that the reconstructed CFR meets the requirement.

[0046] In a possible implementation, the method further includes: receiving second configuration information, the second configuration information including a preset threshold and / or a monitoring period of the performance of the SSF.

[0047] Based on the above scheme, by receiving the second configuration information, the preset threshold and the monitoring period of monitoring the performance of the SSF can be determined, the performance of the SSF is monitored in a targeted manner, by comparing the size relationship between the performance of the SSF and the preset threshold, it can be determined whether the performance of the SSF meets the requirement, and then measures can be taken to fine-tune to ensure the accuracy of channel reconstruction.

[0048] In a possible implementation, the method further includes: sending or receiving indication information, the indication information including at least one of the following: information of a second model, the second model belonging to at least one model supported by the second device; a fourth field, the fourth field being used to indicate switching the first model; a fifth field, the fifth field being used to indicate adjusting information of the first model; or a sixth field, the sixth field being used to indicate updating a transmission density and / or a transmission resource corresponding to a reference signal.

[0049] The second model is different from the first model, and it can be understood that the second model is also a model supported by the first device, that is, in the case that the performance of the SSF does not meet the requirement, the first device and the second device can continue to perform the loop iteration process using the updated second model to obtain the reconstructed CIR.

[0050] Based on the above scheme, by sending or receiving the indication information, in the case that the performance of the SSF does not meet the requirement, the subsequent improvement measures can be explicitly determined, for example, adjusting the neural network parameters of the currently used first model, or using a new second model to reconstruct the channel, or changing the transmission density of the reference signal, for example, increasing or decreasing the transmission number of the reference signal, effectively creating a wireless channel, and improving the transmission efficiency.

[0051] In a second aspect, a channel feedback method is provided. The method can be performed by a second device. Unless otherwise specified, the second device in the present application can refer to a communication device (for example, a network device or a terminal device), a component (for example, a communication module, a processor, a circuit, a chip, or a chip system) in the communication device, or a logic module or software capable of realizing all or part of the functions of the communication device.

[0052] The method comprises: obtaining composition information of a multipath of a communication channel between the first device and the second device; receiving intermediate information, the intermediate information being obtained by processing the composition information of the multipath and second channel information based on a self-supervised neural network; and processing the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed CIR, wherein the second channel information is determined based on first channel information and time delay information of the multipath, and the second channel information comprises a variable channel impulse response (CIR), the first channel information comprises a channel frequency response (CFR) measured based on a reference signal transmitted on the communication channel, and the time delay information of the multipath belongs to the composition information of the multipath.

[0053] Based on the above scheme, the second device obtains the composition information of the multipath of the communication channel between the first device and the second device, and processes the composition information of the multipath and the received intermediate information based on the self-supervised neural network to obtain the reconstructed CIR. The scheme combines the composition information of the multipath and the self-supervised neural network to provide a deployment process of channel feedback, which can reduce the measurement overhead (e.g., sparse measurement based on a reference signal) and the feedback overhead (e.g., feedback of the intermediate information).

[0054] Exemplarily, the second device can determine the composition information of the multipath in a local manner through wireless signal sensing, reference signal measurement, ray tracing, AI, or other manners; or the second device can receive the composition information of the multipath from the first device.

[0055] In a possible implementation, the variable CIR comprises a first CIR, and the second channel information is determined based on the first channel information and the time delay information of the multipath, comprising: the first CIR is determined based on the second CIR, the time delay information of the multipath, the measured CFR, and the reconstructed CFR.

[0056] In a possible implementation, the reconstructed CFR satisfies:

[0057] wherein CFR' represents the reconstructed CFR, CIR n represents a third CIR, n represents the number of the multipath, f represents a carrier frequency corresponding to the measured CFR, τ represents the time delay information of the multipath, the first CIR is a CIR obtained by updating the third CIR, or the third CIR is a CIR determined at a previous time of the first CIR.

[0058] In a possible implementation, the self-supervised neural network comprises an encoding module and a decoding module, the encoding module is configured to output intermediate information, the intermediate information comprises phase information of the multipath; the self-supervised neural network is configured to process the composition information of the multipath and the intermediate information to obtain the reconstructed CIR, comprising: inputting the composition information of the multipath and the intermediate information into the decoding module to output the reconstructed CIR.

[0059] In a possible implementation, the reconstructed CIR is determined in a case where a weighted sum is less than or equal to a preset threshold, and the weighted sum satisfies: Loss = W1 x loss1 + W2 x loss2.

[0060] Wherein, Loss represents the weighted sum, W1 represents the first weight, W2 represents the second weight, loss1 represents a first difference value between the reconstructed CFR and the measured CFR, and loss2 represents a second difference value between the reconstructed CIR and the variable CIR, the first difference value and the second difference value are both used to update the variable CIR, and the first weight and the second weight are both not 0.

[0061] In a possible implementation, the method further comprises: sending first configuration information, the first configuration information being used to indicate that the second device supports self-supervised channel feedback SSF corresponding to the self-supervised neural network.

[0062] In a possible implementation, the first configuration information comprises at least one of the following: information of at least one model supported by the second device; a first field, the first field being used to indicate whether the first device starts the SSF; a second field, the second field being used to indicate that channel information is fed back by using the SSF, the channel information being measured based on a reference signal transmitted on a communication channel; or a third field, the third field being used to indicate an antenna configuration supported by at least one model.

[0063] In a possible implementation, the method further comprises: receiving capability information, the capability information indicating that the first device supports the SSF corresponding to the self-supervised neural network, and / or the capability information indicating a first model supported by the first device, the first model belonging to at least one model corresponding to the self-supervised neural network.

[0064] In a possible implementation, the method further comprises: sending a monitoring reference signal, the monitoring reference signal being used to monitor performance of the SSF; and receiving monitoring response information in a case where the performance of the SSF does not meet a requirement.

[0065] In a possible implementation, the method further comprises: sending second configuration information, the second configuration information comprising a preset threshold and / or a monitoring period of the performance of the SSF.

[0066] In a possible implementation, the method further includes: receiving or sending indication information, the indication information including at least one of the following: information of the second model, the second model belonging to at least one model supported by the second device; a fourth field, the fourth field being used to indicate switching the first model; a fifth field, the fifth field being used to indicate adjusting the information of the first model; or a sixth field, the sixth field being used to indicate updating a transmission density and / or a transmission resource corresponding to the reference signal.

[0067] The second aspect and some implementation forms thereof and their beneficial effects can correspond to the description related to the first aspect, which will not be repeated here.

[0068] In a third aspect, a channel feedback method is provided. The method can be applied to a first device and a second device. For example, the first device is a terminal device, and the second device is a network device; or the first device is a network device, and the second device is a terminal device.

[0069] The method includes: the first device and the second device respectively obtaining composition information of a multipath; the first device processing the composition information of the multipath and second channel information based on a self-supervised neural network to obtain intermediate information; the first device sending the intermediate information to the second device; the second device receiving the intermediate information from the first device; and the second device processing the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed CIR; wherein the second channel information is determined based on first channel information and time delay information of the multipath, and the second channel information includes a variable channel impulse response (CIR); the first channel information includes a channel frequency response (CFR) measured based on a reference signal transmitted on a communication channel; and the time delay information of the multipath belongs to the composition information of the multipath.

[0070] In a possible implementation, the processing, based on the self-supervised neural network, of the composition information of the multipath and the intermediate information to obtain the reconstructed CIR includes: inputting the composition information of the multipath and the intermediate information to a decoding module of the self-supervised neural network to output the reconstructed CIR; and the intermediate information includes phase information of the multipath.

[0071] In a possible implementation, the processing, based on the self-supervised neural network, of the composition information of the multipath and the second channel information to obtain the intermediate information includes: inputting the composition information of the multipath and the second channel information to an encoding module of the self-supervised neural network to output the intermediate information; and the intermediate information includes phase information of the multipath.

[0072] The third aspect and some implementation forms thereof and their beneficial effects can correspond to the description related to the first aspect, which will not be repeated here.

[0073] In a fourth aspect, a channel feedback method is provided. The method can be performed by a first device or a second device. Unless specifically stated, the first device or the second device in the present application can refer to a communication device (e.g., a terminal device or a network device), a component (e.g., a communication module, a processor, a circuit, a chip, or a chip system) in the communication device, or a logic module or software capable of realizing all or part of the functions of the communication device.

[0074] The method comprises: obtaining, by the first device or the second device, composition information of a multipath of a communication channel between the first device and the second device; processing, by the first device or the second device, the composition information of the multipath and second channel information based on a self-supervised neural network to obtain intermediate information; and processing, by the first device or the second device, the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed CIR, wherein the second channel information is determined based on first channel information and time delay information of the multipath, and the second channel information comprises a variable CIR, the first channel information comprises a CFR measured based on a reference signal transmitted on the communication channel, and the time delay information of the multipath belongs to the composition information of the multipath.

[0075] In a possible implementation, the self-supervised neural network comprises an encoding module and a decoding module, and the processing of the composition information of the multipath and the second channel information based on the self-supervised neural network to obtain the intermediate information comprises: inputting, by the first device or the second device, the composition information of the multipath and the second channel information into the encoding module to output the intermediate information, wherein the intermediate information comprises phase information of the multipath.

[0076] In a possible implementation, the self-supervised neural network comprises an encoding module and a decoding module, and the processing of the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain the reconstructed CIR comprises: inputting, by the first device or the second device, the composition information of the multipath and the intermediate information into the decoding module to output the reconstructed CIR.

[0077] In a possible implementation, the variable CIR comprises a first CIR, and the second channel information is determined based on the first channel information and the time delay information of the multipath, and the determination comprises: determining the first CIR based on the second CIR, the time delay information of the multipath, a measured CFR, and a reconstructed CFR.

[0078] In a possible implementation, the reconstructed CFR satisfies:

[0079] wherein CFR' represents the reconstructed CFR, CIR nThe third CIR is denoted as CIR3, n denotes a number of multipaths, f denotes a carrier frequency corresponding to the measured CFR, τ denotes time delay information of the multipaths, the first CIR is a CIR obtained by updating the third CIR, or the first CIR is a CIR determined in a previous cycle of the third CIR.

[0080] In a possible implementation, the reconstructed CIR is determined in a case where a weighted sum is less than or equal to a preset threshold, and the weighted sum satisfies: Loss = W1 x loss1 + W2 x loss2.

[0081] wherein Loss denotes the weighted sum, W1 denotes a first weight, W2 denotes a second weight, loss1 denotes a first difference between the reconstructed CFR and the measured CFR, and loss2 denotes a second difference between the reconstructed CIR and the variable CIR, the first difference and the second difference are both used to update the variable CIR, and the first weight and the second weight are both not 0.

[0082] The fourth aspect and some implementation forms thereof and the beneficial effects thereof can correspond to the description related to the first aspect or the second aspect, and will not be described here.

[0083] In a fifth aspect, a communication apparatus is provided, which has the function of implementing the first aspect, for example, the communication apparatus includes a module or unit or means corresponding to the operations of the first aspect, which can be implemented by software, or by hardware, or by a combination of software and hardware.

[0084] Exemplarily, the communication apparatus can be the first device, for example, a module or unit (for example, a chip, or a chip system, or a circuit) corresponding to the method or operation or step or action described in the first aspect.

[0085] In a possible implementation, the communication apparatus includes a transceiver (or a communication module) and a processing unit (or a processing module) connected to the transceiver.

[0086] Exemplarily, the processing unit is configured to obtain composition information of multipaths of a communication channel between the first device and a second device, and the processing unit is further configured to process the composition information of the multipaths and second channel information based on a self-supervised neural network to obtain intermediate information, and the transceiver is configured to send the intermediate information. The second channel information is determined based on first channel information and time delay information of the multipaths, the second channel information includes a variable channel impulse response CIR, the first channel information includes a channel frequency response CFR measured based on a reference signal transmitted on the communication channel, the time delay information of the multipaths belongs to the composition information of the multipaths, and the intermediate information and the composition information of the multipaths are used to determine a reconstructed CIR.

[0087] In a sixth aspect, a communication apparatus is provided with the function of implementing the second aspect, for example, the communication apparatus includes a module or unit or means corresponding to the operation of the second aspect, which can be implemented by software, or by hardware, or by a combination of software and hardware.

[0088] For example, the communication apparatus can be the second device described above, for example, a module or unit (such as a chip, or a chip system, or a circuit) corresponding to the method or operation or step or action described in the second aspect.

[0089] In a possible implementation, the communication apparatus includes a transceiver unit (or a communication module) and a processing unit (or a processing module) connected to the transceiver unit.

[0090] For example, the processing unit is configured to obtain composition information of a multipath of a communication channel between the first device and the second device; the transceiver unit is configured to receive intermediate information, the intermediate information being obtained by processing the composition information of the multipath and second channel information based on a self-supervised neural network; and the processing unit is further configured to process the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed CIR, wherein the second channel information is determined based on first channel information and time delay information of the multipath, the second channel information includes a variable channel impulse response (CIR), the first channel information includes a channel frequency response (CFR) measured based on a reference signal transmitted on the communication channel, and the time delay information of the multipath belongs to the composition information of the multipath.

[0091] In a seventh aspect, a communication apparatus is provided with the function of implementing the fourth aspect, for example, the communication apparatus includes a module or unit or means corresponding to the operation of the fourth aspect, which can be implemented by software, or by hardware, or by a combination of software and hardware.

[0092] For example, the communication apparatus can be the first device or the second device described above, for example, a module or unit (such as a chip, or a chip system, or a circuit) corresponding to the method or operation or step or action described in the fourth aspect.

[0093] In a possible implementation, the communication apparatus includes a transceiver unit (or a communication module) and a processing unit (or a processing module) connected to the transceiver unit.

[0094] Exemplarily, the processing unit is configured to: obtain composition information of a multipath of a communication channel between the first device and the second device; process the composition information of the multipath and second channel information based on the self-supervised neural network to obtain intermediate information; and process the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed CIR, wherein the second channel information is determined based on first channel information and time delay information of the multipath, and the second channel information includes a variable channel impulse response (CIR), the first channel information includes a channel frequency response (CFR) measured based on a reference signal transmitted on the communication channel, and the time delay information of the multipath belongs to the composition information of the multipath.

[0095] In an eighth aspect, a communication apparatus is provided. The communication apparatus can be the first device or the second device. The communication apparatus includes a transceiver, a processor, and a memory. The processor is configured to control the transceiver to transmit and receive signals. The memory is configured to store a computer program. The processor is configured to invoke and run the computer program from the memory, so that the communication apparatus performs the method in any possible implementation manner of the first aspect or the second aspect.

[0096] Optionally, the processor is one or more, and the memory is one or more.

[0097] Optionally, the memory can be integrated with the processor, or the memory is disposed separately from the processor.

[0098] Optionally, the transceiver includes a transmitter (transmitter) and a receiver (receiver).

[0099] In a ninth aspect, a communication apparatus is provided. The communication apparatus includes one or more processors configured to execute a computer program or instructions, which when executed cause the communication apparatus to implement the method in any possible implementation manner of the first aspect or the second aspect. Optionally, the communication apparatus further includes a memory configured to store part or all of the computer program or instructions that implement the functions related to the first aspect or the second aspect.

[0100] In a possible implementation manner, the communication apparatus can further include an interface circuit, and the processor is configured to communicate with other devices or components through the interface circuit.

[0101] The communication apparatus can be a terminal device, a communication module in a terminal device, a chip responsible for communication functions such as a Modem chip (also known as a baseband chip), or a system on chip (SoC) chip or a system in a package (SIP) chip containing a modem module.

[0102] The communication device can be a network device, or a communication module in the network device, or a circuit or chip responsible for communication functions in the network device, or a functional module capable of invoking and executing a program in the network device.

[0103] In a tenth aspect, a communication system is provided. The communication system includes a first device and / or a second device, wherein the first device is configured to perform the method in any possible implementation of the first aspect, and the second device is configured to perform the method in any possible implementation of the second aspect.

[0104] For example, the first device or the second device can be a terminal device, or a chip or circuit in the terminal device, or a functional module capable of invoking and executing a program in the terminal device; or the second device or the first device can be a network device, or a chip or circuit in the network device, or a central unit (CU) or a distributed unit (DU) in the network device, or a functional module capable of invoking and executing a program in the network device.

[0105] In an eleventh aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores computer program codes or instructions to cause the method in any possible implementation of the first aspect or the second aspect to be implemented. For example, when the computer program codes or instructions are run, the method in any possible implementation of the first aspect or the second aspect is caused to be implemented.

[0106] In a twelfth aspect, a computer program product is provided. The computer program product includes computer program codes or instructions to cause the method in any possible implementation of the first aspect or the second aspect to be implemented. For example, when the computer program product is read and executed by a computer, the method in any possible implementation of the first aspect or the second aspect is caused to be implemented.

[0107] In a thirteenth aspect, a computer program is provided. When the computer program is run, the method in any possible implementation of the first aspect or the second aspect is caused to be implemented.

[0108] The beneficial effects of the fifth aspect to the thirteenth aspect can refer to the first aspect or the second aspect and any possible implementation thereof, which will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0109] FIG. 1 and FIG. 2 are schematic diagrams of a communication system suitable for the present application;

[0110] FIG. 3 shows a schematic diagram of a neuron structure;

[0111] FIG. 4 shows a structural diagram of a feedforward neural network (FNN);

[0112] FIG. 5 shows a structural diagram of a multi-layer perceptron;

[0113] FIG. 6 shows a diagram of iterative optimization of neural network parameters by a gradient descent method;

[0114] FIG. 7 shows a diagram of gradient recursion of multi-layer parameters;

[0115] FIG. 8 is an interactive flowchart of a channel feedback method provided by an embodiment of the present application;

[0116] FIG. 9 is a diagram of an implementation mode of reconstruction of a CIR in combination with sparse measurement and MPC provided by an embodiment of the present application;

[0117] FIG. 10 is a diagram of an implementation mode of a self-supervised neural network provided by an embodiment of the present application;

[0118] FIG. 11 is an interactive flowchart of another channel feedback method provided by an embodiment of the present application;

[0119] FIG. 12 is a schematic block diagram of a communication device provided by an embodiment of the present application;

[0120] FIG. 13 is a schematic block diagram of another communication device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0121] In order to facilitate understanding of the above embodiments provided by the present application, the following points are explained:

[0122] 1) In the present application, the terms and / or descriptions of different embodiments are consistent and can be mutually referenced if there is no special description and logical conflict. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0123] 2) In the present application, "at least one" means one or more, and "multiple" means two or more. The association relationship of "and / or" between the associated objects indicates that there can be three relationships, for example, A and / or B can represent the following three cases: A exists alone, A and B exist together, and B exists alone, where A and B can be singular or plural. In the textual description of the present application, the character " / " generally represents an "or" relationship between the associated objects. "At least one of the following" or similar expressions means any combination of these items, including any combination of single item or multiple items. For example, at least one of a, b and c can represent: a, or b, or c, or a and b, or a and c, or b and c, or a, b and c. Where a, b and c can be single or multiple.

[0124] 3) In the present application, "first", "second", and various numerical numbers (for example, #1, #2, etc.) indicate the differentiation for the convenience of description, and are not used to limit the scope of the embodiments of the present application. For example, different messages are distinguished, rather than used to describe a specific order or sequence. It should be understood that the objects thus described can be interchanged under appropriate circumstances, so as to be able to describe schemes other than the embodiments of the present application.

[0125] 4) In the present application, "when", "in the case of", "if" and the like all refer to the objective situation in which the device will make corresponding processing, and are not limited to time, and do not require the device to have a judgment action when implemented, nor does it mean that there are other limitations.

[0126] 5) In the present application, "indicate" or "for indicating" can include direct indication and indirect indication. When describing that certain indication information is used to indicate A, it can include that the indication information directly indicates A or indirectly indicates A, and it does not mean that A must be carried in the indication information.

[0127] The indication manner involved in the embodiments of the present application should be understood as covering various methods that can enable the to-be-indicated party to know the to-be-indicated information. The to-be-indicated information can be sent as a whole, or can be sent separately in multiple sub-information, and the sending period and / or sending opportunity of these sub-information can be the same or different, and the present application does not limit the sending method.

[0128] The "indication information" in the embodiments of the present application can be explicit indication, that is, directly indicated through signaling, or obtained according to the parameters indicated by the signaling, combined with other rules or combined with other parameters or through derivation. It can also be implicit indication, that is, obtained according to rules or relationships, or according to other parameters, or through derivation. The present application does not make specific limitations on this.

[0129] 6) In the present application, “protocol” can refer to a standard protocol in the field of communication, which can include, for example, a 5th generation (5G) protocol, a new radio (NR) protocol, and a related protocol applied in a future communication system, and the present application does not limit this. “Predefined” can include predefinition. For example, a protocol definition. “Preconfigured” can be implemented by pre-storing corresponding codes, tables, or other means that can be used to indicate related information in a device, and the present application does not limit the implementation thereof.

[0130] 7) In the present application, “communication” can also be described as “data transmission”, “information transmission”, “data processing”, etc. “Transmission” includes “sending” and “receiving”. “Transmission” can be described as “output”.

[0131] 8) In the present application, “sending information to XX (device)” can be understood as that the destination of the information is the device. It can include directly or indirectly sending information to the device. “Receiving information from XX (device), or receiving information from XX (device)” can be understood as that the source of the information is the device, and it can include directly or indirectly receiving information from the device. The information can be processed as necessary between the source and the destination of the information transmission, such as format change, etc., but the destination can understand the valid information from the source. Similar expressions in the present application can be understood similarly, and will not be described here.

[0132] 9) In the present application, words such as “exemplarily” and “for example” are used to represent examples, illustrations or descriptions. Any embodiment or design scheme described as “example” in the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the use of the word “example” is intended to present the concept in a specific way. In the embodiments of the present application, “of”, “corresponding” and “corresponding” can be used interchangeably at times, and it should be pointed out that when their differences are not emphasized, the meanings they express are consistent.

[0133] The technical solutions in the present application will be described below with reference to the accompanying drawings.

[0134] The technical solutions of the embodiments of the present application can be applied to various communication systems, for example, a long term evolution (LTE) system, an LTE frequency division duplex (FDD) system, an LTE time division duplex (TDD), a fifth generation 5G system or new radio NR, and a future communication system. The technical solutions provided in the present application can also be applied to device to device (D2D) communication, vehicle-to-everything (V2X) communication, machine to machine (M2M) communication, machine type communication (MTC), and internet of things (IoT) communication system.

[0135] In addition, the embodiments of the present application are applicable to homogeneous network and heterogeneous network scenarios, and are not limited to transmission points, and can be applied to multi-point cooperative transmission systems between macro base stations and macro base stations, micro base stations and micro base stations, and macro base stations and micro base stations. The embodiments of the present application are applicable to low frequency scenarios, terahertz, optical communication, etc.

[0136] A device in a communication system can send a signal to another device or receive a signal from another device. The signal can include a reference signal, information, signaling, or data, etc. In the present application, the device can be replaced by an entity, a network entity, a communication device, a communication module, a node, a communication node, etc.

[0137] FIG. 1 is a schematic diagram of a communication system applicable to the embodiments of the present application. As shown in FIG. 1, the communication system 10 includes a radio access network (RAN) 100 and a core network (CN) 200. The RAN 100 includes at least one RAN node (such as 110a and 110b in FIG. 1, collectively referred to as 110) and at least one terminal (such as 120a-120j in FIG. 1, collectively referred to as 120). The RAN 100 can also include other RAN nodes, such as wireless relay devices and / or wireless backhaul devices (not shown in FIG. 1), etc. The terminal 120 is connected to the RAN node 110 in a wireless manner. The RAN node 110 is connected to the core network 200 in a wireless or wired manner. The core network device in the core network 200 and the RAN node 110 in the RAN 100 can be different physical devices, or can be the same physical device integrated with the logical functions of the core network and the logical functions of the radio access network.

[0138] The RAN 100 can be a 3rd generation partnership project (3GPP) related cellular system, e.g., a 4th generation (4G) mobile communication system, a 5G mobile communication system, or a future evolution of the 5G mobile communication system. The RAN 100 can also be an open RAN (O-RAN or ORAN), a cloud radio access network (CRAN), or a wireless fidelity (WiFi) system. The RAN 100 can also be a communication system in which two or more of the above systems are integrated.

[0139] The RAN node 110, which can also be referred to as a network device, an access network device, a RAN entity, or an access node, etc., forms part of the communication system and is configured to facilitate wireless access by terminals. The RAN nodes 110 in the communication system 100 can be of the same type or of different types. In some scenarios, the roles of the RAN node 110 and the terminal 120 are relative, e.g., the network element 120i in Figure 1 can be a helicopter or a drone, which can be configured to be a mobile base station. For a terminal 120j that accesses the RAN 100 via the network element 120i, the network element 120i is a base station. But for the base station 110a, the network element 120i is a terminal. The RAN nodes 110 and the terminals 120 are sometimes referred to as communication apparatuses, e.g., the network elements 110a and 110b in Figure 1 can be understood as communication apparatuses with base station functionalities, and the network elements 120a-120j can be understood as communication apparatuses with terminal functionalities.

[0140] In a possible scenario, the RAN node can be a base station (BS), an evolved NodeB (eNodeB), an access point (AP), a transmission reception point (TRP), a base station in a future mobile communication system, or an access node in a WiFi system, etc. The RAN node can be a macro base station (e.g., 110a in Figure 1), a micro base station or an indoor station (e.g., 110b in Figure 1), a relay node or a donor node, or a wireless controller in a CRAN scenario. Optionally, the RAN node can also be a server, a wearable device, a vehicle or a vehicle-mounted device, etc. For example, the access network device in vehicle to everything (V2X) technology can be a road side unit (RSU).

[0141] In another possible scenario, multiple RAN nodes cooperate to assist a terminal to implement wireless access, and different RAN nodes respectively implement part of functions of a base station. For example, a RAN node can be a central unit (CU), a distributed unit (DU), a CU-control plane (CU-CP), a CU-user plane (CU-UP), a radio unit (RU), or a CU-radio unit (CU-RU), etc. The CU and the DU can be separately arranged, or can be included in the same network element, such as a baseband unit (BBU). The RU can be included in a radio frequency device or a radio frequency unit, such as a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH).

[0142] In different systems, the CU (including an open CU-CP (O-CU-CP) and an open CU-UP (O-CU-UP), a DU, or an RU can also have different names, but those skilled in the art can understand their meanings. For example, in an ORAN system, the CU can also be referred to as an open central unit (O-CU), the DU can also be referred to as an open distributed unit (O-DU), the CU-CP can also be referred to as an O-CU-CP, the CU-UP can also be referred to as an O-CU-UP, and the RU can also be referred to as an open radio unit (O-RU). For the convenience of description, the CU, the CU-CP, the CU-UP, the DU, and the RU are taken as examples for description in this application. Any one of the CU (or the CU-CP, the CU-UP), the DU, and the RU in this application can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0143] The terminal 120 can be a device or module with corresponding communication functions for accessing the above-mentioned communication system. The terminal can also be referred to as user equipment (UE), terminal, user device, access terminal, user unit, user station, mobile station, mobile station (MS), remote station, remote terminal, mobile device, user terminal, terminal unit, terminal station, terminal device, wireless communication device, user agent, or user apparatus. The terminal is usually provided with a communication module, circuit or chip for performing corresponding communication functions. The terminal is also configured with program instructions for performing corresponding communication functions.

[0144] For example, the terminal in the embodiments of the present application can be a mobile phone, a personal digital assistant (PDA) computer, a laptop computer, a tablet computer (Pad), a drone, a computer with wireless transceiver function, a machine type communication (MTC) terminal, a virtual reality (VR) terminal, an augmented reality (AR) terminal, an internet of things (IoT) terminal, a wireless terminal in industrial control, a wireless terminal in self driving, a wireless terminal in remote medical, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, a wireless terminal in smart home (such as game consoles, smart televisions, smart speakers, smart refrigerators and fitness equipment, etc.), a transport vehicle with wireless communication function, a communication module, a roadside unit (RSU) with terminal function.

[0145] The RAN 100 and the terminal 120 can be deployed on land, including indoor or outdoor, handheld or vehicle-mounted; can also be deployed on the water surface; can also be deployed on aircraft, balloons and satellites in the air. The scene where the RAN 100 and the terminal 120 are located is not limited in the embodiments of the present application.

[0146] The communication between the access network device and the terminal device follows a certain protocol layer structure. The protocol layer can include a control plane protocol layer and a user plane protocol layer. The control plane protocol layer can include at least one of a radio resource control (RRC) layer, a packet data convergence protocol (PDCP) layer, a radio link control (RLC) layer, a medium access control (MAC) layer, or a physical (PHY) layer, etc. The user plane protocol layer can include at least one of a service data adaptation protocol (SDAP) layer, a PDCP layer, an RLC layer, a MAC layer, or a physical layer, etc.

[0147] For the correspondence between the network elements in the ORAN system and the protocol layer functions that can be implemented by the network elements, refer to Table 1 below.

[0148] Table 1

[0149] The CN 200 can be a 5G core (5GC) or an evolved 5G core. Taking the 5G core as an example, the CN 200 includes an access and mobility management function (AMF) network element responsible for mobility management, access management, and other services, a session management function (SMF) network element responsible for session management, a user plane function (UPF) network element responsible for user plane packet routing and forwarding and quality of service (QoS) control, a policy control function (PCF) network element, and the like. The above core network elements can work independently or can be combined together to implement certain control functions, such as: the AMF, the SMF, and the PCF can be combined together as a core network device.

[0150] The communication system 10 provided in this application can also include an artificial intelligence (AI) network element for implementing some or all AI-related operations. The AI network element can also be referred to as an AI node, an AI device, an AI entity, an AI module, an AI model, or an AI unit, etc. The AI network element can be built into a network element of the communication system. For example, the AI network element can be an AI module built into an access network device, a core network device, a cloud server, or an operation, administration and maintenance (OAM) device for implementing AI-related functions. The OAM can be a core network device OAM and / or an access network device OAM. Alternatively, the AI network element can also be a network element independently arranged in the communication system. Optionally, an AI entity can also be included in a terminal or a chip built into a terminal for implementing AI-related functions.

[0151] FIG. 2 is a schematic diagram of a possible application framework in a communication system suitable for embodiments of the present application. As shown in FIG. 2, the network elements in the communication system are connected through interfaces (such as NG, Xn) or air interfaces. One or more AI modules (only one is shown in FIG. 5 for clarity) are arranged in one or more devices in the network element nodes, such as a core network device, an access network node (RAN node), a terminal, or an OAM. The access network node can be a separate RAN node or can include multiple RAN nodes, such as a CU and a DU. The CU and / or the DU can also be arranged with one or more AI modules. Optionally, the CU can also be split into a CU-CP and a CU-UP. One or more AI models are arranged in the CU-CP and / or the CU-UP.

[0152] The AI module is used to implement corresponding AI functions. The AI modules deployed in different network elements can be the same or different. The AI module can implement different functions according to different parameter configurations of the model of the AI module. The model of the AI module can be configured based on one or more of the following parameters: a structural parameter (such as at least one of 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 bias in the activation function), an input parameter (such as the type of input parameter and / or the dimension of input parameter), or an output parameter (such as the type of output parameter and / or the dimension of output parameter). The bias in the activation function can also be referred to as the bias of the neural network.

[0153] One AI module can have one or more models. One model can infer an output including one parameter or multiple parameters. The learning process, the training process, or the inference process of different models can be deployed in different nodes or devices, or can be deployed in the same node or device.

[0154] It should be understood that the above naming is only defined for the convenience of distinguishing different functions, and should not constitute any limitation to the present application. The present application does not exclude the possibility of using other names in 5G networks and other future networks. For example, part or all of the above-mentioned network elements can use the terms in 5G, or other names, etc.

[0155] It can be understood that FIG. 1 or FIG. 2 is only an example given for the convenience of understanding, and does not constitute a limitation to the protection scope of the present application. The channel feedback method provided by the embodiments of the present application can involve network elements not shown in FIG. 1 or FIG. 2, or can only include part of the network elements shown in FIG. 1 or FIG. 2.

[0156] In order to facilitate the understanding of the embodiments of the present application, the terms, concepts or related technologies involved in the present application are first explained.

[0157] 1, AI: AI can enable machines to have human intelligence, for example, machines can apply computer hardware and software to simulate some intelligent behavior of humans. In order to realize artificial intelligence, machine learning methods can be used. In machine learning methods, machines learn (or train) models using training data. The model represents the mapping between input and output. The learned model can be used for inference (or prediction), that is, the model can be used to predict the output corresponding to a given input. The output can also be referred to as inference result (or prediction result).

[0158] The model can also be referred to as an AI model, a rule, or other names, etc. 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. The AI function can include one or more of the following: data collection, model training (or model learning), model information publishing, model inference (or model inference, inference, or prediction, etc.), model monitoring or model verification, or inference result publishing, etc. The AI function can also be referred to as AI (related) operation, or AI related function.

[0159] Machine learning can include supervised learning, unsupervised learning and reinforcement learning. Unsupervised learning can also be referred to as non-supervised learning.

[0160] Supervised learning learns the mapping relationship from sample values to sample labels according to the collected sample values and sample labels, and uses an AI model to express the learned mapping relationship. The process of training a machine learning model is the process of learning this mapping relationship. In the training process, the sample values are input into the model to obtain the predicted values of the model, and the model parameters are optimized by calculating the error between the predicted values of the model and the sample labels (ideal values). After the mapping relationship is learned, the learned mapping can be used to predict new sample labels. The learned mapping relationship of supervised learning can include linear mapping or nonlinear mapping. According to the type of label, the learned task can be divided into classification task and regression task.

[0161] Unsupervised learning uses algorithms to discover the internal patterns of samples according to the collected sample values. In unsupervised learning, a class of algorithms uses the sample itself as a supervision signal, that is, the model learns the mapping relationship from the sample to the sample, which is called self-supervised learning. In training, the model parameters are optimized by calculating the error between the predicted values of the model and the sample itself. Self-supervised learning can be used for signal compression and decompression recovery applications. Common algorithms include autoencoders and generative adversarial networks.

[0162] Reinforcement learning is different from supervised learning, and is a class of algorithms that learn strategies to solve problems by interacting with the environment. Unlike supervised and unsupervised learning, reinforcement learning problems do not have clear "correct" action label data. The algorithm needs to interact with the environment to obtain the reward signal of the environmental feedback, and then adjust the decision action to obtain a larger reward signal value. In the following power control, the reinforcement learning model adjusts the downlink transmission power of each user according to the system total throughput rate feedback by the wireless network, and then expects to obtain a higher system throughput rate. The goal of reinforcement learning is also to learn the mapping relationship between the environment state and the optimal (for example, the optimal) decision action. However, because the "correct action" label cannot be obtained in advance, the network cannot be optimized by calculating the error between the action and the "correct action". Reinforcement learning training is achieved through iterative interaction with the environment.

[0163] Optionally, the terms AI model, neural network model, AI neural network model, machine learning model, AI processing model, etc. in this application can be replaced with each other.

[0164] 2. Neural Network (NN): A neural network is a specific model in machine learning technology. According to the general approximation theorem, a neural network can theoretically approximate any continuous function, thus enabling it to learn arbitrary mappings. Traditional communication systems require extensive expert knowledge to design communication modules, while deep learning communication systems based on neural networks can automatically discover hidden pattern structures from large datasets, establish mapping relationships between data, and achieve performance superior to traditional modeling methods.

[0165] The idea behind neural networks comes from the neuronal structure of the brain. For example, each neuron performs a weighted summation of its input values ​​and outputs the result through an activation function.

[0166] Figure 3 shows a schematic diagram of a neuron structure. As shown in Figure 3, assume the neuron's input is x = [x0, x1, ..., x2]. n The weights corresponding to each input are w = [w0, w1, ..., w]. n ], where n is a positive integer, w i and x i It can be any possible type, such as a decimal, an integer (e.g., 0, a positive integer, or a negative integer), or a complex number. i As x i The weights are used to assign weights to x. i Weighting is applied. The bias for the weighted summation of the input values ​​is, for example, b. Activation functions can take many forms. Assuming a neuron's activation function is y = f(z) = max(0, z), then the neuron's output is: For example, if the activation function of a neuron is y = f(z) = z, then the output of that neuron is: Here, b can be any possible type, such as a decimal, an integer (e.g., 0, a positive integer, or a negative integer), or a complex number. The activation functions of different neurons in a neural network can be the same or different.

[0167] In addition, a neural network generally includes multiple layers, and each layer can include one or more neurons. By increasing the depth and / or width of the neural network, the expressiveness of the neural network can be improved, providing a more powerful information extraction and abstract modeling capability for complex systems. The depth of the neural network can refer to the number of layers included in the neural network, and the number of neurons included in each layer can be referred to as the width of the layer. In an implementation, the neural network includes an input layer and an output layer. The input layer of the neural network processes the received input information through neurons, and transmits the processing result to the output layer to obtain the output result of the neural network. In another implementation, the neural network includes an input layer, a hidden layer, and an output layer. The input layer of the neural network processes the received input information through neurons, and transmits the processing result to the intermediate hidden layer. The hidden layer calculates the received processing result to obtain a calculation result, and transmits the calculation result to the output layer or the next adjacent hidden layer. Finally, the output layer obtains the output result of the neural network. The neural network can include one hidden layer, or include multiple sequentially connected hidden layers, without limitation.

[0168] The neural network is, for example, a deep neural network (DNN). According to the construction manner of the network, the DNN can include an FNN, a convolutional neural network (CNN), and a recurrent neural network (RNN).

[0169] FIG. 4 shows a structural schematic diagram of an FNN network. As shown in FIG. 4, the FNN network is characterized in that the neurons of adjacent layers are completely connected two by two. This feature makes the FNN usually require a large amount of storage space, resulting in a high calculation complexity.

[0170] The CNN is a neural network specially used to process data with a similar grid structure. For example, time series data (time axis discrete sampling) and image data (two-dimensional discrete sampling) can be considered as data with a similar grid structure. The CNN does not use all the input information for operation at one time, but uses a fixed-size window to extract part of the information for convolution operation, which greatly reduces the calculation amount of model parameters. In addition, according to the different types of information extracted by the window (such as people and objects in the same image), each window can use different convolution kernel operations, which enables the CNN to better extract the features of the input data.

[0171] RNN is a kind of DNN network using feedback time series information. Its input includes new input value at current time and its own output value at previous time. RNN is suitable for obtaining sequence features with correlation in time, and is particularly suitable for speech recognition, channel coding and decoding and other applications.

[0172] The implementation process of the full connection neural network will be described below with reference to the accompanying drawings. The full connection neural network is also called multilayer perceptron (MLP).

[0173] FIG. 5 shows a structural schematic diagram of a multilayer perceptron. As shown in FIG. 5, an MLP includes an input layer (left side), an output layer (right side), and multiple hidden layers (middle). Each layer of the MLP includes a plurality of nodes, which are called neurons. The neurons of adjacent two layers are connected to each other.

[0174] Optionally, considering the neurons of adjacent two layers, the output h of the neuron of the next layer is the weighted sum of all the neurons x of the previous layer connected to the neuron and is subjected to an activation function, which can be expressed as: h = f(wx + b).

[0175] where w is a weight matrix, b is a bias vector, and f is an activation function.

[0176] Further optionally, the output of the neural network can be recursively expressed as: y = f n (w n f n-1 (…)+b n ).

[0177] where n is the index of the layer of the neural network, 1 <= n <= N, where N is the total number of layers of the neural network.

[0178] In other words, the neural network can be understood as a mapping relationship from an input data set to an output data set. Usually, the neural network is randomly initialized, and the process of obtaining this mapping relationship from random w and b using existing data is called training of the neural network.

[0179] Optionally, the specific way of training is to evaluate the output result of the neural network by using a loss function.

[0180] In the above model training process of machine learning, 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 loss function can be embodied in various forms, and the specific form of the loss function is not limited. The model training process can be regarded as the following process: by adjusting part or all of the parameters of the model, the value of the loss function is less than a threshold value or meets the target requirement.

[0181] Figure 6 shows a schematic diagram of iterative optimization of neural network parameters by gradient descent method. As shown in Figure 6, error can be backpropagated, and neural network parameters (including w and b) can be iteratively optimized by gradient descent method until the loss function reaches a minimum, i.e. the“better point (e.g. optimal point)” in Figure 6. It can be appreciated that the neural network parameters corresponding to the“better point (e.g. optimal point)” in Figure 6 can be used as the neural network parameters in the trained AI model information.

[0182] Optionally, the process of gradient descent can be represented as:

[0183] where θ is the parameter to be optimized (including w and b), L is the loss function, η is the learning rate, which controls the step size of gradient descent, represents the derivation operation, represents the derivative of L with respect to θ.

[0184] Optionally, the process of backpropagation utilizes the chain rule of partial derivation.

[0185] Figure 7 shows a schematic diagram of gradient recursion process of multi-layer parameters. As shown in Figure 7, the gradient of the previous layer parameters can be calculated recursively from the gradient of the next layer parameters, which can be expressed as:

[0186] where w ij is the weight of node j connected to node i, s i is the input weighted sum on node i.

[0187] 3. Reference signal: The reference signal involved in the present application includes but is not limited to:

[0188] Pilot reference signal (e.g. channel state information reference signal CSI-RS and / or sounding reference signal SRS), demodulation reference signal (DMRS), tracking reference signal (TRS), phase tracking reference signal (PT-RS), positioning reference signal (PRS), or sensing reference signal (SeRS), etc.

[0189] Optionally, the pilot reference signal can be referred to as a pilot or a pilot signal, where the pilot signal is used for channel measurement. The reference signal in this application can also be a reference signal capable of being carried in an orthogonal frequency division multiplexing (OFDM) symbol in addition to the reference signals listed above, which will not be described here.

[0190] 4. Channel information: measurement information, which refers to information about a path and / or a measured channel obtained by a device through measurement on a channel.

[0191] The channel information indicates information related to a channel between the first device and the second device, for example, at least one of channel state information, channel precoding information, beam information, beam angle information, beam power information, beam indication information, channel eigenvectors, channel eigenvalues, amplitude information of the channel, or phase information of the channel. The channel involved in this application can be an uplink channel, a downlink channel, or a sidelink channel, etc., which is not limited.

[0192] The channel state information is used to indicate the state of the channel. The channel precoding information is used to indicate the precoding matrix of the channel, etc. The beam information is used to indicate a beam used for transmitting or receiving a signal, etc., for example, including an index of the beam. The beam angle information, for example, includes at least one of a beam pointing direction, a beam width, or a beam forming method. The beam pointing direction, for example, includes a main lobe direction formed by beamforming. The beam width refers to the degree of widening of the main lobe formed by beamforming in space. The beam forming method refers to the method of beamforming, for example, numerical method, etc. The beam power information is used to indicate the power of the beam. The beam indication information refers to parameters required for beamforming. The channel eigenvectors are vectors used to represent the channel transmission characteristics. The channel eigenvalues refer to the eigenvalues of the channel matrix. The amplitude information of the channel refers to the amplitude variation of the signal in the transmission process. The phase information of the channel refers to the phase variation of the signal in the transmission process.

[0193] The above description of the terms is only for the convenience of understanding and does not limit the protection scope of the embodiments of the application.

[0194] In a wireless communication system, a multi-input multi-output (MIMO) technology is usually used to increase the system capacity, i.e., multiple antennas are used at the transmitting end and the receiving end at the same time, in which case the transmission of signals between the transmitting end and the receiving end can be multipath propagation. Among them, the multipath propagation can be that the signals reach the receiving antenna after passing through two or more paths in the wireless propagation environment, and the reflection and diffraction of electromagnetic waves by objects in the environment cause multipath. The signals passing through different paths have different time delays and phases, and the receiving antenna receives the superposition of these multipath signals. The time delay spread of multipath causes inter-symbol interference, and the destructive interference caused by multipath causes signal fading. Although multipath causes these problems for the communication system, multipath also increases the number of spatial division multiplexing streams of the communication system. Therefore, it is crucial to predict the multipath in the wireless propagation environment to improve the service capability of the communication system, and predicting the multipath means predicting the possible multipath characteristics of the terminal communication device when communicating with a certain base station at a certain spatial position, such as the number of paths, the strength of the paths, the angle of the paths, the time delay spread of the multipath, the angle spread of the multipath, etc.

[0195] The multipath interference is manifested as time delay spread in the time domain and frequency selective fading in the frequency domain. The time delay spread of multipath causes inter-symbol interference, and the destructive interference caused by multipath causes signal fading. Although multipath causes these problems for the communication system, multipath also increases the number of spatial division multiplexing streams of the communication system. Therefore, it is crucial to predict the multipath in the wireless propagation environment to improve the service capability of the communication system, and predicting the multipath means predicting the possible multipath characteristics of the terminal communication device when communicating with a certain base station at a certain spatial position, such as the number of paths, the strength of the paths, the angle of the paths, the time delay spread of the multipath, the angle spread of the multipath, etc.

[0196] Although the deterministic part of the multipath (such as MPC, including the number, strength, angle, or time delay of the multipath) can be obtained, the phase of each path cannot be obtained by simulation, because the phase of the path changes with the wavelength level in the environment (the position of the communication device and the environmental information cannot be accurately determined to the wavelength level), so the phase of the multipath in the actual scene cannot be obtained by simulation. At the same time, the phase of the path has a great influence on the characteristic pattern between the communication devices, and then affects the precoding performance. Because the phase changes faster and is more unpredictable, the multipath can be regarded as two parts, one part is long-term, i.e., the MPC part of the multipath; the other part is instantaneous, i.e., the phase part of the multipath.

[0197] In order to obtain accurate precoding, the FDD system is to feed back the measured downlink channel information to the BS, and the BS obtains the precoding matrix by using the channel information. This process deals with the frequency domain channel, not the multipath (time domain). The problem of feeding back the frequency domain channel on the air interface is that when the number of antennas of the MIMO system increases, the pilot overhead required for measuring the channel increases, resulting in a higher proportion of communication resources occupied by the pilot, which affects the spectral efficiency of the communication system.

[0198] Currently, the way of using AI or preset codebook to compress and feed back the channel information is to process the channel in the frequency domain. After the frequency domain channel information is quantized and compressed by the encoder, it is transformed into bits to be fed back. These bits are input into the decoder at the receiving end and restored to the original frequency domain channel information. However, the dimension of the frequency domain channel increases with the increase of the number of antennas and the number of bands. Different communication parameters require different processing methods, and the generalization of the method is poor. In order to feed back the frequency domain channel, the frequency domain channel must be measured first, which results in large pilot overhead and feedback overhead.

[0199] In order to solve the above technical problems, the present application provides a channel feedback method, which can reduce the measurement overhead and feedback overhead by combining the composition information of the multipath and the self-supervised neural network for channel feedback.

[0200] The channel feedback method provided by the embodiments of the present application will be described in detail below with reference to the accompanying drawings. The method can be applied to the communication system shown in FIG. 1 or FIG. 2. It should be understood that the embodiments of the present application can be applied to the scenario of communication between the sending end and the receiving end, such as the uplink or downlink transmission scenario.

[0201] It should also be understood that the embodiments shown below do not particularly limit the specific structure of the subject performing the method provided by the embodiments of the present application, as long as the subject is capable of performing communication according to the method provided by the embodiments of the present application by running code or program recorded with the method provided by the embodiments of the present application. For example, the method provided by the embodiments of the present application can be performed by a first device and a second device. In the absence of special description, the "first device" in the present application can refer to a communication device (for example, a terminal device or a network device), a component in the communication device (for example, a communication module, a processor, a circuit, a chip (such as a modem chip, also known as a baseband chip, or a SoC chip or a SIP chip containing a modem core), or a chip system, etc.), or a logic module or software capable of realizing all or part of the functions of the communication device. The "second device" in the present application can refer to a communication device (for example, a network device or a terminal device), a component in the communication device (for example, a communication module, a processor, a circuit, a chip, or a chip system, etc.), or a logic module or software capable of realizing all or part of the functions of the communication device.

[0202] FIG. 8 is a flow diagram of a channel feedback method provided by an embodiment of the present application. As shown in FIG. 8, the method 800 includes the following steps.

[0203] S810, the first device and the second device respectively acquire the composition information of the multipath of the communication channel between the first device and the second device.

[0204] As an example, the first device can be a terminal device, or a chip or circuit in the terminal device, or a functional module capable of invoking and executing a program in the terminal device, and the second device can be a network device, or a chip or circuit in the network device, or a centralized unit CU or a distributed unit DU in the network device, or a functional module capable of invoking and executing a program in the network device.

[0205] As another example, the second device can be a terminal device, or a chip or circuit in the terminal device, or a functional module capable of invoking and executing a program in the terminal device, and the first device can be a network device, or a chip or circuit in the network device, or a centralized unit CU or a distributed unit DU in the network device, or a functional module capable of invoking and executing a program in the network device.

[0206] For ease of description, in the following embodiments, the first device is taken as a terminal device, the second device is taken as a network device, and a downlink transmission scenario is taken as an example for description. The terminal device in the embodiments of the present application can also be referred to as a "terminal side" (UE side) or a "terminal part" (UE part). The network device can also be referred to as a "network side" (Network side) or a "network part" (Network part).

[0207] In the present application, the composition information of the multipath can be a multipath component (MPC) at a position where the first device is connected with the second device, also referred to as a deterministic part of the multipath, including at least one of a direction of departure (DoD), a direction of arrival (DoA), a power, a pitch angle, an azimuth angle, and a time delay. The DoD refers to an angle at which a path departs from a transmitting end, including a departure angle in a horizontal direction (also referred to as an azimuth angle) and a departure angle in a vertical direction (also referred to as a pitch angle). The DoA refers to an angle at which a path arrives at a receiving end, including an arrival angle in a horizontal direction (also referred to as an azimuth angle) and an arrival angle in a vertical direction (also referred to as a pitch angle). The time delay refers to a time consumed by a path from being transmitted at a transmitting end to being received at a receiving end, also referred to as a time of flight.

[0208] In a possible implementation, the first device and the second device can obtain the MPC of the multipath based on a radio frequency map (RF Map), or the first device and the second device can obtain the MPC of the multipath according to a ray-tracing technique, or the first device and the second device can also process environmental information by using an AI neural network model to obtain the MPC of the multipath. Generally, the radio frequency map is input with information of a user and a base station (such as position coordinates, environmental information, etc.), and is output with a multipath component (MPC) at a position where the user is connected with the base station. The radio frequency map refers to a map used to display a coverage range and a signal strength distribution of a wireless signal, and can reflect parameter values of various position points in a wireless network. In short, the first device and the second device can determine the composition information of the multipath in a local manner by means of wireless signal sensing, reference signal measurement, ray tracing, artificial intelligence (AI), or other manners, without the need to transmit the composition information of the multipath, so that the overhead can be reduced.

[0209] In another possible implementation, the first device can obtain the MPC of the multipath from the second device. That is, the first device can send the MPC of the multipath to the second device in the case of obtaining the MPC of the multipath. In other words, the first device can generate / obtain / acquire the MPC of the multipath locally, and the first device can send the MPC of the multipath so that the receiver (for example, the second device) can determine the composition information of the multipath of the communication channel between the first device and the second device, which can reduce the complexity of the receiver.

[0210] In another possible implementation, the second device can obtain the MPC of the multipath from the first device. That is, the second device can also send the MPC of the multipath to the first device in the case of obtaining the MPC of the multipath. In other words, the second device can generate / obtain / acquire the MPC of the multipath locally, and the second device can send the MPC of the multipath so that the receiver (for example, the first device) can determine the composition information of the multipath of the communication channel between the first device and the second device, which can reduce the complexity of the receiver.

[0211] In another possible implementation, the composition information of the multipath can be predefined or preconfigured. The predefinition can include predefinition, such as protocol definition, and the preconfiguration can be implemented by pre-saving the corresponding code, table, function, text, string or other means that can be used to indicate relevant information (for example, the composition information of the multipath) in the first device and / or the second device, and the specific implementation of the present application is not limited.

[0212] It can be understood that the above is an example of the specific implementation of the first device or the second device obtaining the MPC of the multipath, and other schemes are not excluded.

[0213] S820, the first device obtains intermediate information by processing the composition information of the multipath and the second channel information based on the self-supervised neural network.

[0214] In an implementation, the first device obtains intermediate information (or intermediate representation, or intermediate vector, or compressed vector) by processing the composition information of the multipath and the second channel information based on the self-supervised neural network, including: the first device inputs the composition information of the multipath and the second channel information into an encoding module, and outputs the intermediate information, wherein the intermediate information includes phase information of the multipath; and a decoding module is used to output reconstructed CIR.

[0215] It can be understood that the composition information of the multipath and the phase information of the multipath are both influencing factors of the channel information. The phase information of the multipath can be referred to as initial phase information of the multipath. Wherein, the phase information of each path can include the phase of one or more polarization directions, such as x polarization at the transmitting end and y polarization at the receiving end, and the phase information of each path can include x times y polarization, for example, the transmitting end includes horizontal and vertical polarizations, and the receiving end also includes horizontal and vertical polarizations, and each path includes 4 polarizations, which correspond to: the combination of the transmitting end horizontal polarization and the receiving end horizontal polarization, the combination of the transmitting end vertical polarization and the receiving end horizontal polarization, the combination of the transmitting end horizontal polarization and the receiving end vertical polarization, and the combination of the transmitting end vertical polarization and the receiving end vertical polarization.

[0216] In the present application, the self-supervised neural network includes an encoding module (for example, the first part of the self-supervised neural network CIR-SS-Net part 1) and a decoding module (for example, the second part of the self-supervised neural network CIR-SS-Net part 2), which can be a pre-trained network, that is, the data input to the self-supervised neural network can also be used as the output of the self-supervised neural network. Wherein, the encoding module is used for quantization compression CIR, and the decoding module is used for reconstructing CIR. Specifically, the encoding module is used for compressing the second channel information (for example, the variable CIR) to obtain intermediate information, and the decoding module is used for outputting the reconstructed CIR. That is, the self-supervised neural network is used to compress the variable CIR into an intermediate representation, and then reconstruct the CIR using the intermediate representation, realizing the process of compression and reconstruction.

[0217] In the present application, the self-supervised neural network can be deployed in the first device or the second device, that is, the encoding module and the decoding module in the self-supervised neural network are deployed on the same device, for example, the first device or the second device; or the encoding module in the self-supervised neural network is deployed in the first device, and the decoding module in the self-supervised neural network is deployed in the second device; or the encoding module in the self-supervised neural network is deployed in the second device, and the decoding module in the self-supervised neural network is deployed in the first device, without limitation. For ease of description, in the embodiments of the present application, the first device and the second device both deploy the self-supervised neural network.

[0218] It can be understood that using the self-supervised neural network to reconstruct the CIR can ensure the rationality of the reconstructed CIR.

[0219] As an example, the second channel information is determined based on the first channel information and the delay information of the multipath, and the delay information of the multipath belongs to the composition information of the multipath.

[0220] In the present application, the variable CIR includes a first CIR, and the second channel information is determined based on the first channel information and the time delay information of the multipath, and the first CIR is determined based on the second CIR, the time delay information of the multipath, the measured CFR and the reconstructed CFR.

[0221] In the present application, the variable CIR includes a first CIR, and the second channel information is determined based on the first channel information and the time delay information of the multipath, and the first CIR is determined based on the second CIR, the time delay information of the multipath, the measured CFR and the reconstructed CFR.

[0222] Exemplarily, the reconstructed CFR satisfies:

[0223] In the present application, the variable CIR includes a first CIR, and the second channel information is determined based on the first channel information and the time delay information of the multipath, and the first CIR is determined based on the second CIR, the time delay information of the multipath, the measured CFR and the reconstructed CFR. n wherein, CFR' represents the reconstructed CFR, CIR

[0224] It can be understood that CIR represents a time domain channel, and CFR represents a frequency domain channel, and the above formula can be understood as a process of transforming the time domain channel to the frequency domain channel. In addition, the number of times of updating the variable CIR or the number of loop iterations can be one or more, which is not limited in the present application. Wherein, the number of loop iterations is how many times, and the updated CIR corresponds to how many, and after each loop iteration, the obtained updated CIR is usually different.

[0225] Optionally, if it is required to perform one loop iteration, the third CIR can be an initial CIR, at this time, the initial CIR (that is, the second CIR), the number of multipaths, the time delay information of the multipath and the carrier frequency corresponding to the measured CFR can be collectively input to the right side of the above formula, that is, the reconstructed CFR'1 can be obtained, the difference value loss1 between the reconstructed CFR'1 and the measured CFR is calculated, and the parameters of the initial CIR are updated based on the gradient descent of the loss1, to obtain the updated CIR (that is, the first CIR).

[0226] Optionally, if multiple loop iterations are needed to be performed, the first CIR can be a CIR obtained by updating the third CIR, or the third CIR is a previously determined CIR of the first CIR. Assuming there are two loop iteration processes, for the first loop iteration, the initial CIR (i.e., the second CIR), the number of multipaths, the delay information of the multipaths, and the carrier frequency corresponding to the measured CFR are collectively input to the right side of the above formula, to obtain a reconstructed CFR'1, calculate the difference loss' between the reconstructed CFR'1 and the measured CFR, and update the parameters of the initial CIR based on the gradient descent of the loss' to obtain an updated CIR (i.e., the third CIR). Then, for the second loop iteration, the third CIR, the number of multipaths, the delay information of the multipaths, and the carrier frequency corresponding to the measured CFR are collectively input to the right side of the above formula, to obtain a reconstructed CFR'2, calculate the difference loss'' between the reconstructed CFR'2 and the measured CFR, and update the parameters of the second CIR (i.e., the first CIR) based on the gradient descent of the loss''.

[0227] It should be noted that the number of loop iterations is not limited in the present application, and can be one or more times until convergence, i.e., the process of loop iteration can be stopped, i.e., the parameters of the current CIR are stopped being updated. In other words, the current CIR at this time is the reconstructed wireless channel, e.g., the reconstructed CIR.

[0228] In the present application, the first channel information is obtained by measuring a reference signal (e.g., a sparse pilot signal for channel measurement, which can be understood as a reference signal transmitted at certain specific time-frequency points (such as 4 time-frequency points determined by symbol 1, symbol 3, subcarrier 10, and subcarrier 5)) transmitted on the communication channel. For example, the first channel information can include the actually measured CFR. It can be understood that the CFR is a method for describing how the frequency components of a signal change in the process of being transmitted from the sending end to the receiving end on a specific transmission medium. The CFR reflects the amplification or attenuation ability of the communication channel to different frequency signals, and affects the quality of the signal and the reliability of the communication.

[0229] As an example, in a downlink channel estimation scenario, the network device transmits a pilot reference signal (e.g., CSI-RS) to the terminal device, the terminal device receives and measures the CSI-RS to obtain a channel measurement result, and the channel measurement result includes the actually measured CFR. Optionally, if the second channel information is determined by the network device based on the first channel information and the delay information of the multipaths, the terminal device also transmits the actually measured CFR to the network device; if the second channel information is determined by the terminal device based on the first channel information and the delay information of the multipaths, the terminal device does not need to transmit the actually measured CFR to the network device, thereby reducing the signaling overhead.

[0230] As an example, for an uplink channel estimation scenario, the terminal device sends a pilot reference signal, e.g., SRS, to the network device, the network device receives and measures the SRS to obtain a channel measurement result, which includes the actually measured CFR. Optionally, if the second channel information is determined by the network device based on the first channel information and the delay information of the multipath, the network device does not need to send the actually measured CFR to the terminal device; if the second channel information is determined by the terminal device based on the first channel information and the delay information of the multipath, the network device also sends the actually measured CFR to the terminal device.

[0231] In this application, the second channel information is determined based on the first channel information and the delay information of the multipath, which can be understood as: the first device determines the second channel information based on the first channel information and the delay information of the multipath; or the second device determines the second channel information based on the first channel information and the delay information of the multipath, and sends the second channel information to the first device, which is used by the first device to process the composition information of the multipath and the second channel information based on the self-supervised neural network to obtain intermediate information.

[0232] The second channel information can include a variable CIR. It can be understood that the CIR represents the response of a unit impulse signal after passing through a system. The variable CIR can be represented as CIR B*N*R*T The variable parameters include B (the number of samples), N (the number of multipaths), R (the number of receiving antennas), and T (the number of transmitting antennas). In the embodiments of the present application, updating the variable CIR is to update the above variable parameters to obtain an updated CIR, i.e., to obtain a reconstructed CIR.

[0233] In this application, the transmission of the reference signal (e.g., pilot reference signal or monitoring reference signal) can be beamformed, and the beam direction of the beamforming is determined by the composition information of the multipath. For example, in the case where the composition information of the multipath is known, the communication device can select the direction of the strongest path in the multipath as the beam direction to improve the communication quality.

[0234] In this application, the reconstructed CIR is determined when the weighted sum is less than or equal to a preset threshold, and the weighted sum satisfies: Loss = W1 x loss1 + W2 x loss2 (2)

[0235] wherein, Loss represents a weighted sum, W1 represents a first weight, W2 represents a second weight, loss1 represents a first difference between the reconstructed CFR and the measured CFR, loss2 represents a second difference between the reconstructed CIR and the variable CIR, and the first weight and the second weight are both not 0, and the first difference and the second difference are both used to update the variable CIR.

[0236] Optionally, the first weight and the second weight can be predefined or preconfigured values, for example, W1=1 and W2=5. It is found through research that, if the first weight or the second weight is 0, for example, W1=1 and W2=0, or W1=0 and W2=5, the corresponding first difference or second difference gradually increases, and the similarity or similarity between the reconstructed channel and the measured channel is low; if the first weight or the second weight is both not 0, for example, W1=1 and W2=0, the corresponding first difference and second difference are close and gradually decrease, and the similarity or similarity between the reconstructed channel and the measured channel is high, and the performance is better.

[0237] In the present application, the first difference can also be referred to as a first loss calculation result, that is, obtained by the first device or the second device through loss calculation on the reconstructed CIR (or referred to as predicted channel information) and the variable CIR. Similarly, the second difference can also be referred to as a second loss calculation result, that is, obtained by the first device or the second device through loss calculation on the reconstructed CFR (or referred to as predicted channel information) and the actually measured CFR (i.e., the first channel information).

[0238] Optionally, the first loss calculation result or the second loss calculation result can be represented as a deviation (or difference, difference, etc.) calculation method of a predicted value (reconstructed value) and a true value, such as a mean absolute error (MAE), a mean square error (MSE), a normalized mean square error (NMSE), a correlation calculation (wherein, the higher the correlation degree, the smaller the deviation, and vice versa, the lower the correlation degree, the greater the deviation), etc. It should be understood that the “predicted value” or “reconstructed value” can be understood as the predicted channel information, and the first channel information is measured based on the reference signal, and for this purpose, the first channel information can be understood as the “true value”.

[0239] Optionally, the loss calculation can be realized through a mathematical model, a simulation model, an AI model, etc.

[0240] It should be noted that the first device or the second device can determine whether to perform the sparse measurement and the channel feedback based on the self-supervised neural network again based on the first loss calculation result or the second loss calculation result.

[0241] For example, in a case where the first loss calculation result indicates that the loss is lower than a certain threshold, the first device or the second device can determine that the difference between the reconstructed CIR and the variable CIR is small, i.e., the correlation between the two is relatively high. For this purpose, the first device or the second device does not need to perform the channel feedback based on the self-supervised neural network again. In a case where the second loss calculation result indicates that the loss is lower than a certain threshold, the first device or the second device can determine that the difference between the reconstructed CFR and the actually measured CFR is small, i.e., the correlation between the two is relatively high. For this purpose, the first device or the second device does not need to perform the sparse channel measurement again.

[0242] For example, in a case where the first loss calculation result indicates that the loss is lower than a certain threshold, the first device or the second device can determine that the difference between the reconstructed CIR and the variable CIR is small, i.e., the correlation between the two is relatively high. For this purpose, the first device or the second device does not need to perform the channel feedback based on the self-supervised neural network again. In a case where the second loss calculation result indicates that the loss is lower than a certain threshold, the first device or the second device can determine that the difference between the reconstructed CFR and the actually measured CFR is small, i.e., the correlation between the two is relatively high. For this purpose, the first device or the second device does not need to perform the sparse channel measurement again.

[0243] Optionally, the implementation of repeatedly performing the sparse channel measurement and the channel feedback based on the self-supervised neural network can be implemented by a gradient descent method, a stochastic gradient descent method, etc.

[0244] S830, the first device sends the intermediate information to the second device;

[0245] Correspondingly, the second device receives the intermediate information from the first device.

[0246] S840, the second device processes the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain the reconstructed CIR.

[0247] In an implementation, the second device processes the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain the reconstructed CIR, including: the second device inputs the composition information of the multipath and the intermediate information into a decoding module to output the reconstructed CIR.

[0248] Considering that the initial phase of the multipath is unpredictable, the CIR cannot be obtained accurately by using MPC only. In addition, the CIR cannot be obtained accurately by using sparse measurement only, because there are many solutions for the CIR at this time. Therefore, the application proposes to use sparse measurement and MPC simultaneously, and to obtain the unique solution of the CIR by using the gradient descent method.

[0249] Next, in combination with FIG. 9 and FIG. 10, the specific implementation mode of reconstructing the CIR and the CFR is exemplarily illustrated.

[0250] FIG. 9 is a schematic diagram of an implementation mode of reconstructing the CIR by combining sparse measurement and MPC according to an embodiment of the application. As shown in FIG. 9, the predicted composition information of the multipath (for example, MPC B*N*X not containing the phase) includes the time delay information of the multipath. The left half corresponds to the sparse measurement, and the right half corresponds to the MPC-based self-supervised neural network feedback. The variable CIR B*N*R*T has trainable parameters (or variable parameters), which can be optimized by the gradient descent method.

[0251] For the left half, the time delay information of the multipath and the variable CIR B*N*R*T are used for the above formula (1), and the predicted CFR B*1*R*T is obtained after the calculation of the formula (1). The first difference between the predicted CFR B*1*R*T and the CFR obtained by the real measurement sparse pilot is loss1, and the gradient descent method is used to optimize or update the parameters of the variable CIR B*N*R*T based on the loss1 (wherein B, N, R, T represent four dimensions of the CIR multi-dimensional tensor, that is, the CIR is a [B, N, R, T] dimensional tensor, and each element in the tensor can be changed by the gradient descent, for example, B represents the batch size, that is, the number of samples used in each training iteration, that is, the sample number of the CIR, that is, B CIRs can be updated by the gradient descent method at the same time; N represents the number of multipaths; R represents the number of receiving antennas or receiving ports; and T represents the number of transmitting antennas or transmitting ports. For example, when B = 10, N = 8, R = 4, and T = 64, the CIR is a four-dimensional tensor with dimensions of [10, 8, 4, 64], and the total number of parameters of the CIR is 10*8*4*64 = 20480, which can be understood that these parameters can be updated), and with the update of the parameters of the CIR B*N*R*T , the loss1 gradually decreases.

[0252] For the right half, the variable CIR B*N*R*T and the MPC B*N*X(Wherein, B represents batchsize, N represents the number of multipath, X represents the number of information contained in each path for representing this path) is input into the CIR self-supervised neural network, and the CIR is output B*N*R*T The second difference value between the output CIR B*N*R*T And the input variable CIR B*N*R*T Is loss2, and the parameters of the variable CIR B*N*R*T Are optimized or updated based on the loss2 gradient descent method, and the loss2 gradually decreases with the update of the parameters of the CIR B*N*R*T It should be pointed out that the self-supervised neural network is pre-trained, that is, the parameters of the self-supervised neural network are frozen (unchanged), but the gradient backpropagation is supported.

[0253] It can be understood that the above two parts of the process of updating the parameters of the CIR B*N*R*T Can be understood as a loop iteration, for example, it can be executed once or more times until convergence, which is not limited by the present application. Therefore, the variable CIR B*N*R*T Can include the second CIR B*N*R*T And the updated CIR B*N*R*T It can be understood that the number of loop iterations is how many times the updated CIR B*N*R*T Corresponding to how many, and after each loop iteration, the updated CIR B*N*R*T Obtained is usually different. Assuming that 100 loop iterations can achieve convergence, that is, Iter=100, then the corresponding weighted sum Loss i Satisfies: Loss i =W 1i ×loss 1i +W 2i ×loss 2i

[0254] Next, based on the calculated loss Loss i Complete the parameter update of the CIR B*N*R*T . Among them, W 1i And W 2i May change at each iteration, the first weight and the second weight described above are both not zero, which means that in the whole Iter=100 iterations, W 1i And W 2i At least in some i iterations are not zero. For example, there are 100 loop iterations to update the parameters of the CIR B*N*R*T You can set the first weight corresponding to all odd loop iterations to be non-zero and the second weight to be zero, for example, you can set W 1i =1, W 2i= 0, all even order iteration corresponding to the first weight can be set to zero and the second weight is set to non-zero, for example, W 1i = 0, W 2i = 1.

[0255] In summary, the left half and the right half are iterated simultaneously to update the parameters of the variable CIR B*N*R*T until convergence, that is, the weighted sum Loss in the above formula (2) is less than or equal to the preset threshold, and the updating process of the parameters of the CIR B*N*R*T can be stopped, and the reconstructed CIR can be obtained, that is, the CIR obtained after the sparse measurement and MPC comprehensive optimization. It should be pointed out that during the iterative process, if the left half and the right half are executed by the second device and the first device respectively, the first device and the second device need to perform signaling interaction after each iteration, for aligning the current parameters of the variable CIR B*N*R*T , so that the first device and the second device use the latest CIR B*N*R*T to perform the next round of iteration.

[0256] It should be pointed out that the above-mentioned left half and right half of the iterative process in Figure 9 can be executed by the first device and / or the second device. That is, the iterative process of the left half and the right half is executed by the first device, or the iterative process of the left half and the right half is executed by the second device, or the iterative process of the left half is executed by the first device and the iterative process of the right half is executed by the second device, or the iterative process of the left half is executed by the second device and the iterative process of the right half is executed by the first device, which is not limited by the present application.

[0257] Figure 10 is a schematic diagram of an implementation of a self-supervised neural network (for example, CIR self-supervised Net) according to an embodiment of the present application, which can be understood as a specific implementation of the right half of Figure 9. As shown in Figure 10, the self-supervised neural network includes two parts (for example, CIR-SS-Net part), which are an encoding module and a decoding module. The execution subject of the encoding module can be a terminal device, and the execution subject of the decoding module can be a network device. Specifically, the input of the encoding module is the variable CIR B*N*R*T and the MPC B*N*X without phase, and the output is intermediate information (which can be understood as a compressed representation of CIR), the dimension of the intermediate information is less than the dimension of the CIR B*N*R*T , and the intermediate information can be understood as containing the CIR B*N*R*T but the MPC B*N*XThe instantaneous phase part not included, or intermediate information can replace the initial phase of the MPC. Then, the terminal device can only feed back the intermediate information to the network device without feeding back all the channel information. Further, the network device will combine the intermediate information and the MPC without phase to obtain the final channel information. B*N*X Splicing, input decoding module, output reconstructed CIR B*N*R*T .

[0258] Optionally, based on the deployment process of the self-supervised neural network-based channel feedback, the second device can broadcast the configuration, and the first device can report the capability, according to whether the first device and the second device support the SSF. That is, the method further includes the following step S801.

[0259] S801, the second device sends first configuration information to the first device;

[0260] Correspondingly, the first device receives the first configuration information from the second device.

[0261] The first configuration information is used to indicate that the second device supports self-supervised channel feedback (SSF) corresponding to the self-supervised neural network.

[0262] Optionally, the present application does not limit the implementation of the second device to obtain the first configuration information, for example, the second device can receive the first configuration information from the core network side, or the first configuration information can be predefined or preconfigured.

[0263] Optionally, the first configuration information can be broadcast or multicast by the second device through a system message.

[0264] Exemplarily, the first configuration information includes at least one of the following: information of at least one model supported by the second device (which can be equivalent to information of a CIR-SS-Net model library), a first field, a second field, a third field, or a fourth field. The first field is used to indicate whether the first device starts the SSF, the second field is used to indicate that the channel information is fed back by using the SSF, the channel information is measured based on a reference signal transmitted on a communication channel, and the third field is used to indicate an antenna configuration supported by the at least one model.

[0265] The model library includes at least one CIR-SS-Net model (for example, a self-supervised neural network) supported by the second device and available, and it can be understood that different CIR-SS-Net models can have different input / output dimensions or neural network parameters.

[0266] It can be understood that a neural network is a complex network composed of neurons. It is a mathematical model constructed based on artificial neurons or bionic neurons, which can simulate the learning and cognitive process of the human brain. That is, the self-supervised neural network is an implementation of at least one model supported by the second device. Different models can have different input / output dimensions or neural network parameters.

[0267] Next, the first configuration information is illustrated in the form of a table, as shown in Table 2.

[0268] Table 2

[0269] Optionally, the CIR-SS-Net model library can be maintained at the network side (for example, the core network (network, NW)). For example, the NW sends the configuration table of the model library to the network device, and the network device sends it to the terminal device through broadcast signaling. The terminal device feeds back and the current scene matching (for example, the input / output dimension of the CIR-SS-Net model and the current UE matching / supporting / available) model ID, and the network device can obtain it from the network side if it does not have the model. The terminal device can also obtain it from the network side through the network device if it also does not have the model. If the network device and the terminal device both have the model, it is not necessary to obtain it from the network side, that is, both of them can start and execute the channel feedback process based on the SSF.

[0270] Next, the information of the CIR-SS-Net model library is illustrated in the form of a table. As shown in Table 3, different models correspond to different serial numbers or indexes, and each model has different CIR-SS-Net model configuration parameters, wherein the CIR-SS-Net model configuration parameters can include antenna configuration, polarization case, dimension of intermediate information, and numerical precision. Optionally, the CIR-SS-Net model configuration parameters can also include at least one of the following: neural network structure, input / output dimension, applicable scene (such as urban area, rural area, non line of sight (NLoS), line of sight (LoS)), number of multipaths corresponding to MPC (such as more than 10 or less than 10), and the like.

[0271] Table 3

[0272] It should be noted that the above Table 3 is only an example given for understanding, and other schemes are not excluded.

[0273] Optionally, the number of correspondence relations (for example, a row in the table) in the above-mentioned Table 3 is not limited by the present application, for example, one or more rows can be added or reduced. Optionally, the above-mentioned Table 3 can be split into multiple independent tables, and the present application does not limit the splitting manner, for example, the CIR-SS-Net model configuration parameters with even model numbers can be independently formed into a new table, and the CIR-SS-Net model configuration parameters with odd model numbers can be independently formed into a new table.

[0274] Optionally, the deployment process of the channel feedback based on the self-supervised neural network includes the following step S802.

[0275] S802, the first device sends capability information to the second device;

[0276] Correspondingly, the second device receives the capability information from the first device.

[0277] The capability information indicates that the first device supports the SSF corresponding to the self-supervised neural network, and / or the capability information indicates a first model supported by the first device, and the first model belongs to at least one model corresponding to the self-supervised neural network. For example, the terminal device can report the model number 2, i.e. the first model, based on the self-capability information and the model library information in Table 3, the corresponding antenna configuration and polarization is (8, 4, 2)T(2, 2, 2)R, and the dimension and numerical accuracy of the corresponding intermediate information is 32 (bit number). Subsequently, the first device and the second device can perform channel feedback of the CIR self-supervised neural network based on the first model.

[0278] Optionally, the present application does not limit the execution order of the above-mentioned steps S801 and S802.

[0279] Optionally, in the process of performing the above-mentioned channel feedback based on the SSF, the first device and the second device can perform monitoring and fine-tuning of the performance of the SSF, that is, the method further includes the following steps S803-S806.

[0280] S803, the second device sends a monitoring reference signal to the first device;

[0281] Correspondingly, the first device receives the monitoring reference signal from the second device.

[0282] The monitoring reference signal is used to monitor the performance of the SSF, or in other words, the monitoring reference signal is used to monitor whether the reconstructed CFR is better than the CFR measured by the real measurement.

[0283] Optionally, compared with the sparse pilot for channel measurement sent by the second device, the density of the monitoring reference signal (or in other words, the monitoring pilot or the monitoring pilot signal) is larger, which can assist the first device to evaluate the performance of the SSF.

[0284] Optionally, the monitoring reference signal can be transmitted non-periodically or periodically, without limitation.

[0285] S804, in the case that the performance of the SSF does not meet the requirement, the first device sends monitoring response information to the second device;

[0286] Correspondingly, the second device receives the monitoring response information from the first device.

[0287] The monitoring response information indicates that the performance of the SSF does not meet the requirement.

[0288] As an example, the performance of the SSF can be characterized by the correlation or similarity between the reconstructed CFR and the measured CFR. It can be understood that the higher the correlation or similarity, the smaller the difference between the reconstructed CFR and the measured CFR. If the correlation or similarity is greater than or equal to a preset threshold, it means that the reconstructed CFR meets the requirement; otherwise, if the correlation or similarity is less than or equal to the preset threshold, it means that the reconstructed CFR does not meet the requirement.

[0289] As another example, the performance of the SSF can be characterized by the normalized mean squared error (NMSE) between the reconstructed CFR and the measured CFR, or the NMSE of the two channels corresponding to the reconstructed CFR and the measured CFR. It can be understood that the larger the NMSE, the more dispersed the distribution of the reconstructed CFR and the measured CFR; otherwise, the smaller the NMSE, the more concentrated the distribution of the reconstructed CFR and the measured CFR data. If the NMSE is greater than or equal to a preset threshold, it means that the reconstructed CFR does not meet the requirement; otherwise, if the NMSE is less than or equal to the preset threshold, it means that the reconstructed CFR meets the requirement.

[0290] It can be understood that the correlation less than or lower than the threshold means evaluating whether the correlation degree (or similarity degree) between the two channel matrices is lower than the threshold. For example, the correlation degree can be characterized by a mathematical calculation result of the two channel matrices (e.g., corresponding to the reconstructed channel and the measured channel), which can include mean square error (MSE), normalized mean square error (NMSE), or cosine similarity, etc. For another example, the correlation degree can also be characterized by a mathematical calculation result based on other processing results of the two channel matrices, such as calculating the covariance matrix of the channel matrix, or calculating the vector after singular value decomposition (SVD) decomposition of the channel matrix, etc.

[0291] S805, the second device sends second configuration information to the first device;

[0292] Correspondingly, the first device receives the second configuration information from the second device.

[0293] The second configuration information includes a preset threshold and / or a monitoring period of the performance of the SSF.

[0294] Optionally, the preset threshold can be predefined or preconfigured. The predefinition can include predefinition, such as protocol definition, and the preconfiguration can be implemented by pre-saving corresponding codes, tables, functions, texts, strings or other means that can be used to indicate relevant information (e.g., the preset threshold) in the first device and / or the second device. The specific implementation of the present application is not limited.

[0295] Optionally, the present application does not limit the execution order of step S805. For example, step S805 can be executed before step S803, or it can also be executed before step S804.

[0296] Next, the second configuration information is illustrated in the form of a table, as shown in Table 4.

[0297] Table 4

[0298] S806, the second device sends indication information to the first device; correspondingly, the first device receives the indication information from the second device.

[0299] Alternatively, the first device sends indication information to the second device; correspondingly, the second device receives the indication information from the first device.

[0300] That is, the indication information can be sent by the first device or the second device, which is not limited. Or if the performance of the SSF decreases or does not meet the requirements, some fine-tuning measures can be taken to improve the performance of the SSF. The specific fine-tuning measures can be determined by the first device or the second device. Based on the indication information, the first device or the second device can determine how to adjust the performance of the SSF.

[0301] Exemplarily, the indication information includes at least one of the following: second model information, fourth field, fifth field, or sixth field. The second model belongs to at least one model supported by the second device, the fourth field is used to indicate switching the first model, the fifth field is used to indicate adjusting the information of the first model, and the sixth field is used to indicate updating the transmission density and / or transmission resource corresponding to the reference signal.

[0302] The following illustrates the indication information for SSF performance fine-tuning in the form of a table. As shown in Table 5, when the SSF performance is found to be degraded, different fields can be used for different fine-tuning measures. For example, if the switching model is indicated, the updated model ID can be sent; or if the current model is fine-tuned, the signaling of starting model fine-tuning can be sent; or if the configuration of the sparse pilot reference signal (e.g., placement density or placement location) is changed, the pilot configuration signaling can be sent.

[0303] Table 5

[0304] Exemplarily, after receiving the monitoring reference signal, the terminal device assumes to perform sparse channel measurement to obtain the actually measured CFR, and obtains the reconstructed CIR and the reconstructed CFR according to the method shown in FIGS. 8-10, and compares the reconstructed CFR with the actually measured CFR (for example, the correlation between the two can be calculated). When the reconstructed performance does not meet the requirement (for example, the correlation between the reconstructed channel and the measured channel is lower than or equal to a preset threshold, or the NMSE of the reconstructed channel and the measured channel is higher than or equal to a preset threshold), the terminal device can send a monitoring response message for feeding back that the SSF performance does not meet the requirement. At this time, the terminal device can send the indication information shown in Table 5 for feeding back the subsequent SSF performance improvement measures, or the network device determines and sends the indication information to the terminal device for feeding back the subsequent SSF performance improvement measures, such as the ID of the updated second model.

[0305] It can be understood that the technical solutions of the present application are applicable to uplink transmission scenarios and downlink transmission scenarios. For ease of understanding, the specific process applicable to the embodiments of the present application will be introduced in the following with reference to FIG. 11 for the downlink transmission scenario. In the following examples, the first device is taken as the UE and the second device is taken as the BS. It can be understood that the process described below is only an example for illustration, and the embodiments of the present application are not limited thereto. The content not described in detail below can be referred to the description in the method 800, which will not be described here.

[0306] FIG. 11 is an interactive flowchart of a channel feedback method according to an embodiment of the present application. As shown in FIG. 8, the method 800 includes the following steps.

[0307] S1101, the BS determines the MPC of the multipath.

[0308] S1102, the UE determines the MPC of the multipath.

[0309] It should be noted that the MPCs of the multi-paths determined by the BS and the UE are the same, i.e., the MPC-Net IDs of the two are aligned. For specific implementation of the MPCs of the multi-paths determined by the BS and / or the UE, reference can be made to the related description of step S810 of method 800 described above, and for brevity, will not be repeated here.

[0310] S1103, the BS sends a pilot reference signal to the UE; correspondingly, the UE receives the pilot reference signal from the BS.

[0311] Exemplarily, the pilot reference signal can be a CSI-RS, used for channel measurement. That is, the UE performs measurement based on the received pilot reference signal to obtain CIR and / or CFR.

[0312] Optionally, the pilot reference signal can be sent aperiodically or periodically, without limitation.

[0313] Optionally, the pilot reference signal can also be replaced by a sparse pilot signal or a sparse pilot, which can be understood as sending the pilot reference signal at certain specific time-frequency points (such as 4 time-frequency points determined by symbol 1, symbol 3, subcarrier 10 and subcarrier 5) to reduce channel measurement overhead.

[0314] S1104, the BS and the UE align the information of the self-supervised neural network.

[0315] As an example, the BS and the UE can align the identification of the self-supervised neural network, such as CIR-SS-Net ID, for implementing MPC-based self-supervised channel feedback; and the BS and the UE can align the first model corresponding to the self-supervised neural network, such as input / output dimension, neural network parameter, etc.

[0316] S1105, the UE inputs the composition information of the multi-paths and the second channel information to the encoding module (e.g., CIR-SS-Net part1) of the self-supervised neural network, and outputs intermediate information.

[0317] The second channel information includes variable CIR, and the intermediate information includes the phase of the multi-paths.

[0318] For specific implementation of the structure, deployment and UE obtaining of the intermediate information of the self-supervised neural network, reference can be made to the related description of step S820 of method 800 described above, and for brevity, will not be repeated here.

[0319] S1106, the UE sends the intermediate information to the BS; correspondingly, the BS receives the intermediate information from the UE.

[0320] S1107, the BS inputs the composition information of the multipath and the intermediate information to a decoding module (e.g., CIR-SS-Net part2) of the self-supervised neural network, and outputs a reconstructed CIR.

[0321] The specific implementation of the BS obtaining the reconstructed CIR can refer to the related description of step S840 of method 800, and will not be described here for brevity.

[0322] S1108, the BS obtains a reconstructed CFR according to the reconstructed CIR and the composition information of the multipath.

[0323] For example, based on the obtained reconstructed CIR and the composition information of the multipath (e.g., the delay information of the multipath), and the above formula (1), the reconstructed CFR can be calculated, and the specific implementation can refer to the related description of formula (1) of method 800.

[0324] S1109, the BS sends downlink data to the UE; correspondingly, the UE receives the downlink data from the BS.

[0325] For example, the BS calculates the precoding based on the reconstructed CFR, and completes the downlink data transmission by using the precoding, so that the BS can perform high-rate signal transmission based on the reconstructed CFR.

[0326] It should be noted that in the method shown in FIG. 11, the downlink transmission scenario is mainly exemplarily illustrated, for example, the CSI-RS based channel measurement. For brevity, the uplink transmission scenario will not be exemplarily illustrated herein, for example, the SRS based channel measurement. The difference from FIG. 11 is that the first device is the BS, the second device is the UE, and the pilot reference signal is sent by the UE to the BS, and the information is obtained by the BS completing the channel measurement.

[0327] It should also be noted that in the method shown in FIG. 11, the BS and the UE side are mainly deployed with MPC-Net and CIR-SS-Net. Alternatively, MPC-Net and CIR-SS-Net can be deployed on one side, for example, the BS or the UE. Correspondingly, signaling interaction is needed between the BS and the UE on the air interface, that is, the corresponding input and output are transmitted, for example, including the MPC of the multipath and the intermediate information.

[0328] As an example, when the MPC-Net is only deployed on the BS side, in order to align the MPC of the multipath, the BS needs to send the output of the MPC-Net, that is, the MPC of the multipath to the UE.

[0329] As another example, when the MPC-Net is only deployed on the UE side, in order to align the MPC of the multipath, the UE needs to send the output of the MPC-Net, that is, the MPC of the multipath to the BS.

[0330] As yet another example, when both MPC-Net and CIR-SS-Net are deployed at the BS side, the BS does not need to send MPC of the multipath to the UE, because the UE side cannot solve the CIR, or in other words, the UE side cannot obtain the reconstructed CIR. At this time, the UE can only feed back the result of sparse measurement, i.e., the actually measured CFR. Then, the MPC of the multipath is obtained at the BS side, and the channel feedback of the self-supervised neural network of the CIR is performed to reconstruct the CIR. For example, a second device obtains composition information of a multipath of a communication channel between the first device and the second device; processes the composition information of the multipath and second channel information based on a self-supervised neural network to obtain intermediate information; processes the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed CIR; wherein the second channel information is determined based on first channel information and delay information of the multipath, and the second channel information includes a variable channel impulse response CIR, the first channel information includes a channel frequency response CFR measured based on a reference signal transmitted on the communication channel, and the delay information of the multipath belongs to the composition information of the multipath.

[0331] As yet another example, when both MPC-Net and CIR-SS-Net are deployed at the UE side, the UE does not need to send MPC of the multipath to the BS, because the BS side cannot solve the CIR, or in other words, the BS side cannot obtain the reconstructed CIR. At this time, the BS can only feed back the result of sparse measurement, i.e., the actually measured CFR. Then, the MPC of the multipath is obtained at the UE side, and the channel feedback of the self-supervised neural network of the CIR is performed to reconstruct the CIR. For example, a first device obtains composition information of a multipath of a communication channel between the first device and a second device; processes the composition information of the multipath and second channel information based on a self-supervised neural network to obtain intermediate information; processes the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed CIR; wherein the second channel information is determined based on first channel information and delay information of the multipath, and the second channel information includes a variable channel impulse response CIR, the first channel information includes a channel frequency response CFR measured based on a reference signal transmitted on the communication channel, and the delay information of the multipath belongs to the composition information of the multipath.

[0332] As yet another example, when MPC-Net is deployed at the UE side and CIR-SS-Net is deployed at the BS side, the UE side can obtain the predicted MPC of the multipath and feed back the MPC of the multipath to the BS, and at the same time, the UE feeds back the result of sparse measurement, i.e., the actually measured CFR, to the BS, and then the BS reconstructs the CIR.

[0333] As another example, when the MPC-Net is deployed at the BS side and the CIR-SS-Net is deployed at the UE side, the BS side can obtain the predicted MPC of the multipath and feed back the MPC of the multipath to the UE, while the BS feeds back the sparse measurement result, i.e., the CFR actually measured, to the UE, and then the UE reconstructs the CIR. Similarly.

[0334] That is, the present application does not limit the deployment network element of the MPC-Net and the CIR-SS-Net, which can be deployed in a network device, or can also be deployed in a terminal device. Among them, the MPC-Net is used to obtain the MPC of the multipath, and the CIR-SS-Net is used to perform self-supervised channel feedback based on the MPC.

[0335] Based on the above scheme, the present application proposes a self-supervised channel feedback method based on MPC, which can reduce the measurement overhead (for example, sparse measurement based on pilot reference signals) and feedback overhead (for example, feedback of intermediate information) by aligning the MPC-Net and the CIR-SS-Net between the first device and the second device, combining the composition information of the multipath and the self-supervised neural network, and reconstructing the CIR and the CFR.

[0336] The above describes the channel feedback method side embodiment of the present application in combination with FIGS. 1 to 11, and the communication device side embodiment of the present application will be described in detail below in combination with FIGS. 12 and 13. It should be understood that the description of the device embodiment corresponds to the description of the method embodiment, and therefore, the parts not described in detail can be referred to the foregoing method embodiment.

[0337] FIG. 12 is a schematic block diagram of a communication device 1000 provided by an embodiment of the present application. As shown in FIG. 12, the communication device 1000 includes a processing module 1010 and a communication module 1020. The communication device 1000 can be a sending end device, or can be a communication device applied to or matched with the sending end device and capable of realizing the method performed by the sending end device, such as a chip, a chip system or a circuit; or the communication device 1000 can be a receiving end device, or can be a communication device applied to or matched with the receiving end device and capable of realizing the method performed by the receiving end device, such as a chip, a chip system or a circuit.

[0338] Among them, the communication module can also be referred to as a transceiver module, a transceiver, a transceiver, a transceiver unit or a transceiver device, etc. The processing module can also be referred to as a processor, a processing board, a processing unit or a processing device, etc. Optionally, the communication module is used to perform the sending operation and the receiving operation of the sending end device and the receiving end device in the above method, and the device in the communication module for realizing the receiving function can be regarded as a receiving unit, and the device in the communication module for realizing the sending function can be regarded as a sending unit, that is, the communication module includes a receiving unit and a sending unit.

[0339] Optionally, the communication apparatus 1000 further comprises a storage module 1030 for storing device program codes and / or data.

[0340] In one example, when the communication apparatus 1000 is applied to a first device (e.g. a terminal device or a network device), the processing module 1010 can be configured to implement the processing functions of the first device in the above embodiments, and the communication module 1020 can be configured to implement the transceiving functions of the first device in the above embodiments.

[0341] In another example, when the communication apparatus 1000 is applied to a second device (e.g. a network device or a terminal device), the processing module 1010 can be configured to implement the processing functions of the second device in the above embodiments, and the communication module 1020 can be configured to implement the transceiving functions of the second device in the above embodiments.

[0342] It is further noted that the aforementioned communication module and / or processing module can be implemented by virtual modules, for example, the processing module can be implemented by a software function unit or a virtual apparatus, and the communication module can be implemented by a software function or a virtual apparatus. Alternatively, the processing module or the communication module can also be implemented by an entity apparatus, for example, if the apparatus is implemented by a chip / circuit (e.g. an integrated circuit or a logic circuit, etc.), the communication module can be an input / output circuit and / or a communication interface, performing input operation (corresponding to the aforementioned receiving operation) and output operation (corresponding to the aforementioned sending operation); and the processing module is an integrated processor or a microprocessor or a circuit (e.g. an integrated circuit or a logic circuit, etc.).

[0343] The division of the modules in the present application is illustrative, and is only a logical function division. In actual implementation, there can be another division manner, and each functional module in each example in the present application can be integrated in one processor, or can be a separate physical existence, or two or more modules can be integrated in one module. The integrated module can be implemented in the form of hardware or in the form of a software function module.

[0344] In one example, the functional units in any of the above apparatuses can be one or more integrated circuits configured to implement the above methods, e.g., one or more application specific integrated circuits (ASICs), or, one or more central processing units (CPUs), one or more microcontroller units (MCUs), one or more digital signal processors (DSPs), or, one or more field programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0345] In one example, the storage module 1030 can include random access memory, flash memory, read only memory, programmable read only memory, electrically programmable read only memory, and / or registers, etc.

[0346] Optionally, the communication apparatus 1000 can further include an AI module 1040 for implementing part or all of the AI-related operations. The AI modules deployed in different communication apparatuses (e.g., terminal devices or network devices) can be the same or different. One AI module can have one or more models. One model can infer an output including one parameter or multiple parameters. The model of an AI module can be configured based on one or more of the following parameters: a structure parameter, an input parameter, or an output parameter.

[0347] Illustratively, when the AI module 1040 is applied to a first device (e.g., a terminal device), the AI module 1040 can be used to implement the AI function of the first device in the above embodiments, e.g., to obtain intermediate information based on a self-supervised neural network processing the composition information of the multipath and the second channel information. When the AI module 1040 is applied to a second device (e.g., a network device), the AI module 1040 can be used to implement the AI function of the second device in the above embodiments, e.g., to obtain the reconstructed CIR based on the self-supervised neural network processing the composition information of the multipath and the intermediate information.

[0348] FIG. 13 is a schematic block diagram of a communication apparatus 2000 according to an embodiment of the present application. The communication apparatus 2000 can be a chip or a chip system. Optionally, the chip system can be composed of a chip or can include a chip and other discrete devices. It can be understood that the communication apparatus 2000 includes necessary means such as modules, units, elements, circuits, or interfaces, and the like, which are configured to work together to implement the present solution. The communication apparatus 2000 can be a RAN node, a terminal, a core network device, or another network device, or a component (e.g., a chip) of these devices, to implement the methods described in the above method embodiments.

[0349] As shown in FIG. 13, the communication apparatus 2000 can be used to implement the functions of any device (e.g., a terminal device or a network device) in the communication system described in the foregoing examples. The communication apparatus 2000 can include at least one processor 2010. The processor 2010 can be a general processor or a special-purpose processor, etc. For example, it can be a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, and the central processing unit can be used to control the communication apparatus (e.g., a RAN node, a terminal, or a chip, etc.), execute software programs, and process data of the software programs. The processor 2010 can also be referred to as a processing unit, which controls the communication apparatus (e.g., a RAN node or a terminal).

[0350] Optionally, in one design, the processor 2010 can include a program (which can also be referred to as code or instructions), which can be run on the processor 2010, so that the communication apparatus 2000 performs the methods described in the above embodiments. In another possible design, the communication apparatus 2000 includes a circuit (not shown in the figure) for implementing the functions described in the above embodiments, such as processing the composition information of the multipath and the second channel information to obtain intermediate information based on a self-supervised neural network, or processing the composition information of the multipath and the intermediate information to obtain a reconstructed CIR based on the self-supervised neural network.

[0351] Optionally, the communication apparatus 2000 can include one or more memories 2020, which have programs (which can also be referred to as code or instructions) stored thereon, and the programs can be run on the processor 2010, so that the communication apparatus 2000 performs the methods described in the above method embodiments.

[0352] Optionally, the processor 2010 is coupled to a memory 2020, which can be within the device, or the memory can be integral to the processor, or the memory can be external to the device. For example, the communication device 2000 can further include at least one memory 2020. The memory 2020 stores computer program, computer programs or instructions and / or data necessary for implementing the above-described examples. The processor 2010 can execute the computer program stored in the memory 2020 to complete the method in any of the above-described examples.

[0353] Optionally, the processor 2010 and / or the memory 2020 can include an AI module for implementing AI-related functions. The AI module can be implemented by software, hardware, or a combination of software and hardware. For example, the AI module can include a Radio Access Network Intelligent Controller (RIC) module. For example, the AI module can be a near-real-time RIC or a non-real-time RIC.

[0354] Optionally, the processor 2010 and / or the memory 2020 can further store data. The processor and the memory can be separately provided or integrated together.

[0355] The communication device 2000 can further include a communication interface 2030 and an antenna 2050. The communication device 2000 can exchange information with other devices through the communication interface 2030. For example, the communication interface 2030 can be a transceiver, a circuit, a bus, a module, a pin, or other types of communication interfaces. When the communication device 2000 is a chip-type device or a circuit, the communication interface 2030 in the device 2000 can also be an input / output circuit that can input (or receive) information and output (or transmit) information. The processor 2010 can be an integrated processor, a microprocessor, an integrated circuit, or a logic circuit, etc. The processor can determine output information according to input information. The communication interface 2030 can also be referred to as a transceiver unit, a transceiver, or a transceiver circuit, etc., and is used to realize the transceiving function of the communication device through the antenna 2050.

[0356] In an example, when the communication device 2000 is applied to a first device (e.g., a terminal device or a network device), the processor 2010 can be used to implement the processing function of the first device in the above-described embodiments, and the communication interface 2030 can be used to implement the transceiving function of the first device in the above-described embodiments.

[0357] In another example, the communication apparatus 2000 is applied to a second device (for example, a network device or a terminal device), the processor 2010 can be configured to implement the processing function of the second device in the above embodiments, and the communication interface 2030 can be configured to implement the transceiving function of the second device in the above embodiments.

[0358] The coupling in the present application is an indirect coupling or communication connection between devices, units or modules, which can be electrical, mechanical or other forms, for information interaction between devices, units or modules. The processor 2010 can operate in cooperation with the memory 2020 and the communication interface 2030. The specific connection medium between the processor 2010, the memory 2020 and the communication interface 2030 is not limited in the present application.

[0359] Optionally, the communication apparatus 1000 can further include an AI module, for example, the AI module can be separately deployed, or can be combined with the processor 2010. The AI module is configured to implement part or all of the AI-related operations. The AI modules deployed in different communication apparatuses can be the same or different. One AI module can have one or more models. One model can infer an output, which includes one parameter or multiple parameters. The model of the AI module can be configured based on one or more of the following parameters: a structure parameter, an input parameter, or an output parameter.

[0360] Exemplarily, when the AI module is applied to a first device (for example, a terminal device), it can be configured to implement the AI function of the first device in the above embodiments, for example, based on a self-supervised neural network to process the composition information of the multipath and the second channel information to obtain intermediate information. When the AI module 1040 is applied to a second device (for example, a network device), it can be configured to implement the AI function of the second device in the above embodiments, for example, based on a self-supervised neural network to process the composition information of the multipath and the intermediate information to obtain the reconstructed CIR.

[0361] Optionally, as shown in FIG. 13, the processor 2010, the memory 2020 and the communication interface 2030 are connected to each other through a bus 2040. Optionally, the bus can include address bus, data bus, control bus and other types of buses. In addition, for ease of representation, one bus 2040 is shown in FIG. 13, but it does not mean that there is only one bus or only one type of bus.

[0362] It should be understood that the processor mentioned in the embodiments of the present application can be a device or a part of circuit for processing function in the following devices: a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0363] It should also be understood that the memory mentioned in the embodiments of the present application can be a volatile memory and / or a non-volatile memory. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically EPROM (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM). For example, the RAM can be used as an external cache. As an example but not limitation, the RAM includes the following various forms: static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM) and direct memory bus random access memory (direct rambus RAM, DR RAM).

[0364] It should be noted that when the processor is a general-purpose processor, DSP, ASIC, FPGA or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, the memory (storage module) can be integrated in the processor.

[0365] It is also important to note that the storage described herein is intended to comprise, without being limited to, these and any other suitable types of storage.

[0366] The embodiments of the present application further provide a computer readable storage medium, having stored thereon computer instructions for implementing the method performed by the communication device (e.g., the first device and / or the second device) in each of the above method embodiments.

[0367] The embodiments of the present application further provide a computer program product, containing instructions, which, when executed by a computer, implement the method performed by the communication device (e.g., the first device and / or the second device) in each of the above method embodiments.

[0368] The embodiments of the present application further provide a communication system, comprising the first device and / or the second device in the above embodiments.

[0369] Optionally, the communication system further comprises the first device and / or the second device in the above embodiments.

[0370] The explanations and beneficial effects of the related contents in any of the above devices can refer to the corresponding method embodiments provided above, and will not be repeated here.

[0371] In various embodiments of the present application, the size of the serial number of each process does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0372] The present application will present various aspects, embodiments or features around a system that can include multiple devices, components, modules, etc. It should be understood and appreciated that each system can include additional devices, components, modules, etc., and / or can not include all of the devices, components, modules, etc. discussed in connection with the figures. Furthermore, combinations of these approaches can also be used.

[0373] In the present application, each example can be mutually quoted between each example without logical contradiction, for example, the method and / or terms between the method embodiments can be mutually quoted, for example, the functions and / or terms between the device embodiments can be mutually quoted, for example, the functions and / or terms between the device examples and the method examples can be mutually quoted.

[0374] It should be understood that in some of the above embodiments, the existing network architecture is mainly exemplified by devices, and the specific form of the devices is not limited by the embodiments of the present application. For example, devices that can realize the same function in the future are also applicable to the embodiments of the present application.

[0375] Those skilled in the art can clearly understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0376] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0377] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0378] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0379] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically independently, or two or more units can be integrated into one unit.

[0380] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes various media that can store program codes, such as a U disk, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disk, etc.

[0381] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A channel feedback method, applied to a first device, characterized in that, The method comprises: obtaining composition information of a plurality of paths of a communication channel between the first device and the second device; processing the composition information of the plurality of paths and second channel information based on a self-supervised neural network to obtain intermediate information; sending the intermediate information; wherein the second channel information is determined based on first channel information and time delay information of the plurality of paths, the second channel information comprises a variable channel impulse response (CIR), the first channel information comprises a channel frequency response (CFR) measured based on a reference signal transmitted on the communication channel, the time delay information of the plurality of paths belongs to the composition information of the plurality of paths, and the intermediate information and the composition information of the plurality of paths are used to determine a reconstructed CIR.

2. The method of claim 1, wherein, The variable CIR comprises a first CIR, and the second channel information is determined based on the first channel information and the time delay information of the plurality of paths. The first CIR is determined based on a second CIR, the time delay information of the plurality of paths, the measured CFR, and a reconstructed CFR.

3. The method of claim 2, wherein, The reconstructed CFR satisfies: wherein the CFR' represents the reconstructed CFR, the CIR n represents a third CIR, the n represents the number of the multipaths, the f represents the carrier frequency corresponding to the measured CFR, the τ represents the time delay information of the multipaths, the first CIR is a CIR updated from the third CIR, or the third CIR is a previously determined CIR of the first CIR.

4. The method according to any one of claims 1 to 3, characterized in that, The self-supervised neural network comprises an encoding module and a decoding module, the processing of the composition information of the plurality of paths and the second channel information based on the self-supervised neural network to obtain the intermediate information comprises: inputting the composition information of the plurality of paths and the second channel information into the encoding module to output the intermediate information; wherein the intermediate information comprises phase information of the plurality of paths, and the decoding module is used to output the reconstructed CIR.

5. The method according to any one of claims 2 to 4, characterized in that, The reconstructed CIR is determined in a case where a weighted sum is less than or equal to a preset threshold, and the weighted sum satisfies: Loss = W1 x loss1 + W2 x loss2. wherein the Loss represents the weighted sum, W1 represents a first weight, W2 represents a second weight, loss1 represents a first difference between the reconstructed CFR and the measured CFR, and loss2 represents a second difference between the reconstructed CIR and the variable CIR, the first difference and the second difference are both used to update the variable CIR, and the first weight and the second weight are both not 0.

6. The method according to any one of claims 1 to 5, characterized in that, The method further comprises: receiving first configuration information, the first configuration information being used to indicate that the second device supports self-supervised channel feedback (SSF) corresponding to the self-supervised neural network.

7. The method of claim 6, wherein, The first configuration information comprises at least one of the following: information of at least one model supported by the second device; a first field used to indicate whether the first device starts the SSF; a second field used to indicate that channel information is fed back by using SSF, the channel information being measured based on a reference signal transmitted on the communication channel; or a third field used to indicate an antenna configuration supported by the at least one model.

8. The method according to any one of claims 1 to 7, characterized in that, The method further comprises: sending capability information, the capability information indicating that the first device supports the SSF corresponding to the self-supervised neural network, and / or the capability information indicating a first model supported by the first device, the first model belonging to at least one model corresponding to the self-supervised neural network.

9. The method according to any one of claims 6 to 8, characterized in that, The method further comprises: receiving a monitoring reference signal, the monitoring reference signal being used for monitoring performance of the SSF; in a case where the performance of the SSF does not meet a requirement, sending monitoring response information.

10. The method of claim 9, wherein, The method further comprises: receiving second configuration information, the second configuration information comprising a monitoring period of the performance of the SSF.

11. The method according to claim 9 or 10, characterized in that, The method further comprises: sending or receiving indication information, the indication information comprising at least one of: information of a second model, the second model belonging to at least one model supported by the second device; a fourth field, the fourth field being used for indicating switching the first model; a fifth field, the fifth field being used for indicating adjusting information of the first model; or a sixth field, the sixth field being used for indicating updating a transmission density and / or a transmission resource corresponding to the reference signal. 12.A channel feedback method applied to a second device, the method comprising: comprising: obtaining composition information of a multipath of a communication channel between a first device and a second device; receiving intermediate information, the intermediate information being obtained based on processing, by a self-supervised neural network, of the composition information of the multipath and second channel information; obtaining a reconstructed CIR based on processing, by the self-supervised neural network, of the composition information of the multipath and the intermediate information; wherein the second channel information is determined based on first channel information and time delay information of the multipath, the second channel information comprising a variable CIR, the first channel information comprising a CFR measured based on a reference signal transmitted on the communication channel, and the time delay information of the multipath belonging to the composition information of the multipath.

13. The method of claim 12, wherein, The variable CIR comprises a first CIR, and the second channel information is determined based on first channel information and time delay information of the multipath, comprising: The first CIR is determined based on a second CIR, the time delay information of the multipath, the measured CFR, and a reconstructed CFR.

14. The method of claim 13, wherein, The reconstructed CFR satisfies: wherein the CFR' represents the reconstructed CFR, the CIR n represents a third CIR, the n represents the number of the multipaths, the f represents the carrier frequency corresponding to the measured CFR, the τ represents the time delay information of the multipaths, the first CIR is a CIR updated from the third CIR, or the third CIR is a previously determined CIR of the first CIR.

15. The method according to any one of claims 12 to 14, characterized in that, The self-supervised neural network comprises an encoding module and a decoding module, the encoding module being used for outputting the intermediate information, the intermediate information comprising phase information of the multipath; obtaining a reconstructed CIR based on processing, by the self-supervised neural network, of the composition information of the multipath and the intermediate information, comprising: inputting the composition information of the multipath and the intermediate information into the decoding module to output the reconstructed CIR.

16. The method according to any one of claims 13 to 15, characterized in that, The reconstructed CIR is determined in a case where a weighted sum is less than or equal to a preset threshold, the weighted sum satisfying: Loss=W1×loss1+W2×loss2; wherein the Loss represents the weighted sum, the W1 represents a first weight, the W2 represents a second weight, the loss1 represents a first difference between a reconstructed CFR and the measured CFR, the loss2 represents a second difference between the reconstructed CIR and the variable CIR, the first difference and the second difference are both used for updating the variable CIR, and the first weight and the second weight are both not 0.

17. A method of channel feedback, the method comprising: comprising: the first device and the second device respectively obtaining composition information of a multipath; The first device processes the composition information of the multipath and second channel information based on a self-supervised neural network to obtain intermediate information; The first device sends the intermediate information to the second device; The second device receives the intermediate information from the first device, processes the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed channel impulse response (CIR); The second channel information is determined based on first channel information and delay information of the multipath, the second channel information includes a variable CIR, the first channel information includes a channel frequency response (CFR) measured based on a reference signal transmitted on a communication channel, and the delay information of the multipath belongs to the composition information of the multipath.

18. The method of claim 17, wherein, The second device processes the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed CIR, including: The second device inputs the composition information of the multipath and the intermediate information into a decoding module of the self-supervised neural network to output the reconstructed CIR; The intermediate information includes phase information of the multipath.

19. The method of claim 17 or 18, wherein, The first device processes the composition information of the multipath and second channel information based on a self-supervised neural network to obtain intermediate information, including: The first device inputs the composition information of the multipath and the second channel information into an encoding module of the self-supervised neural network to output the intermediate information; The intermediate information includes phase information of the multipath.

20. A method of channel feedback, the method comprising: Including: Obtaining composition information of a multipath of a communication channel between the first device and the second device; Processing the composition information of the multipath and second channel information based on a self-supervised neural network to obtain intermediate information; Processing the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed channel impulse response (CIR); The second channel information is determined based on first channel information and delay information of the multipath, the second channel information includes a variable CIR, the first channel information includes a channel frequency response (CFR) measured based on a reference signal transmitted on a communication channel, and the delay information of the multipath belongs to the composition information of the multipath.

21. The method of claim 20, wherein, The self-supervised neural network includes an encoding module and a decoding module, and the first device or the second device processes the composition information of the multipath and second channel information based on a self-supervised neural network to obtain intermediate information, including: The first device or the second device inputs the composition information of the multipath and the second channel information into the encoding module to output the intermediate information; The intermediate information includes phase information of the multipath.

22. The method of claim 20 or 21, wherein, The self-supervised neural network includes an encoding module and a decoding module, and the first device or the second device processes the composition information of the multipath and the intermediate information based on the self-supervised neural network to obtain a reconstructed CIR, including: The first device or the second device inputs the composition information of the multipath and the intermediate information to the decoding module, and outputs the reconstructed CIR.

23. A communications device, characterized by comprising modules for implementing the method of any one of claims 1 to 11, or for implementing the method of any one of claims 12 to 16.

24. A communications device, characterized by comprising a processor for executing computer program instructions in a memory to cause the method of any one of claims 1 to 11 to be performed, or to cause the method of any one of claims 12 to 16 to be performed.

25. The communication apparatus of claim 24, wherein The communication apparatus further comprises a memory for storing the computer program or instructions; and / or, The communication apparatus further comprises a communication interface coupled to the at least one processor, the communication interface configured to input and / or output information.

26. A computer-readable storage medium, characterized in that, The computer readable storage medium is configured to store a computer program which, when executed on a computer, causes the method of any one of claims 1 to 16 to be performed.

27. A computer program product, characterised in that, comprising a computer program or instructions which, when executed by a processor, cause the method of any one of claims 1 to 16 to be performed.

Images