Communication method and related equipment

Abstract

The embodiment of the invention is applied to the field of communication, and provides a communication method and related equipment, the method is applied to network equipment, and the method comprises the following steps: sending a first reference signal, the first reference signal being associated with M ports of the network equipment; and a first CSI report sent by the first user equipment is received, the first CSI report is determined by the first user equipment according to the first reference signal, the first CSI report is associated with N ports in the network equipment, N is greater than M, and M and N are both integers greater than 1. Based on the technical method provided by the invention, the accuracy of the CSI report can be improved under the condition of occupying less communication resources.

Description

Technical Field

[0001] This application relates to the field of communications, and more specifically, to a communication method and related equipment. Background Technology

[0002] Network devices can be equipped with multiple-input multiple-output (MIMO) antennas. MIMO antennas enable beamforming. To transmit information to user equipment (UE), the network device needs to accurately obtain the UE's downlink channel state information (CSI).

[0003] Network devices can transmit reference signals through multiple ports. User equipment (UE) can measure the received reference signals and send a CSI report to the network device. The network device then sends information back to the UE based on the CSI report. Obtaining highly accurate CSI reports with low power consumption and minimal communication resource usage is a pressing issue that needs to be addressed. Summary of the Invention

[0004] This application provides a communication method and related equipment that can improve the accuracy of CSI reports.

[0005] Firstly, a communication method is provided, applied to a network device. This method can be executed by the network device itself, or by a component configured within the network device (such as a circuit, chip, or chip system), or by a logic module or software capable of implementing all or part of the functions of the network device. This application does not limit the scope of this method. The following description uses a network device as an example.

[0006] The method includes: sending a first reference signal, the first reference signal being associated with network devices to obtain M ports; receiving a first CSI report sent by a first UE, the first CSI report being determined by the first UE based on the first reference signal, the first CSI report being associated with N ports in the network devices, N being greater than M, and M and N being integers greater than 1.

[0007] The communication method provided in this application involves a smaller number of ports associated with the network device transmitting the first reference signal, resulting in lower power consumption and less occupation of communication resources. The network device receives a first CSI report transmitted by the first UE associated with a larger number of ports. The number of ports associated with the CSI report is positively correlated with the accuracy of the first CSI report, thus the first CSI report has higher accuracy. In other words, the first CSI report can more accurately reflect the downlink CSI of the first UE. The network device can communicate with the first UE based on the first CSI report transmitted by the first UE, thereby enabling the first UE to receive the information transmitted by the network device more accurately.

[0008] In some possible implementations, reference signals are transmitted at multiple times. The reference signal transmitted at time t is associated with a first port group, and the reference signal transmitted at time t-1 is associated with a second port group. The ports in the first port group are different from the ports in the second port group. All M ports are ports in the first port group. The first CSI report is determined by the first UE based on the reference signals at multiple times, where t is an integer greater than 1.

[0009] At multiple times, network devices transmit reference signals through port groups including different ports. The first UE determines its first CSI report based on these multiple reference signals, thus enabling the first CSI report to include channel information from more ports and making it more accurate. In some possible implementations, the port groups associated with the reference signals sent by the network device at multiple times are port groups that the network device selects cyclically from Q port groups, where Q is an integer greater than 1.

[0010] By setting up multiple port groups, the network device can cycle through multiple port groups at each time period and send reference signals through the ports in the selected port group, making the method of determining the port for sending reference signals more flexible.

[0011] In some possible implementations, the method includes: sending indication information that instructs a first UE to report a CSI report based on reference signals at multiple times.

[0012] The indication information can instruct the UE that receives the indication information to send a CSI report based on reference signals at multiple times. The first UE can be the UE that received the indication information.

[0013] In some possible implementations, the length of Q of the multiple times is less than a preset duration, and the reference signals sent by the network device in the Q times are associated with the port groups of Q port groups respectively.

[0014] The length of Q time intervals can also be understood as the time span of Q time intervals. The first UE may move, causing a change in its location, which may lead to channel lag. Completing the transmission of reference signals associated with the Q port groups within a shorter preset time interval reduces the possibility of channel lag and improves the accuracy of the first CSI report.

[0015] In some possible implementations, the Q ports include multiple candidate ports, and the importance of the candidate ports is higher than that of the other ports in the N ports besides the multiple candidate ports.

[0016] Network devices use ports of higher importance as candidate ports, and by associating the transmitted reference signals with the ports of higher importance, the first CSI report becomes more accurate.

[0017] In some possible implementations, the method further includes determining a second order of the N ports in terms of importance based on CSI reports from multiple user equipment.

[0018] The order of port importance is determined based on CSI reports from multiple UEs, resulting in high environmental adaptability and a more accurate reflection of the current environment. The order of port importance also maintains high accuracy when the real-time nature of CSI reports from multiple UEs is high.

[0019] In some possible implementations, determining the second order of the N ports according to their importance may include: determining the importance of each of the N ports based on the CSI reports of the multiple UEs using a port model; in the case that there are a second port and a third port among the multiple candidate ports, the number of port groups including the second port is greater than the number of port groups including the third port in the Q port groups, the first importance of the second port is greater than the second importance of the third port, and the difference between the first importance and the second importance is greater than or equal to a preset importance difference threshold.

[0020] By using a port model and analyzing the CSI reports from multiple UEs, the importance of each port can be determined. In some cases, among the N ports of a network device, one or more ports are significantly more important than the others. For these ports, the network device transmits reference signals at a higher frequency, which improves the accuracy of the determined downlink CSI for the first UE.

[0021] In some possible implementations, the CSI report of any one of the multiple user equipment (UEs) represents the downlink channel state information of that UE. This downlink channel state information includes the complex channel gain parameters of multiple ports out of N ports of the network device. The port model is a model trained based on second training data. The second training data includes channel information samples and importance order labels. The channel information samples include the training downlink CSIs of multiple training UEs. The training downlink CSI of any one of the training UEs includes the complex channel gain parameters of N1 third training ports out of N third training ports of the second training network device determined by that training UE. The importance order labels represent the importance ranking result of the N1 third training ports, where the N third training ports are the same as the N ports in the network device, N1 is less than or equal to N, and N1 is an integer greater than 1.

[0022] During training, the port model learns the relationship between the downlink CSI of multiple training UEs and the port importance ranking of the second training network device. Therefore, by processing the downlink CSI of multiple UEs, the port model can accurately determine the importance of each port of the network device.

[0023] In some possible implementations, the port model is obtained by adjusting the parameters of the initial port model based on the loss value. The initial port model is used to process channel information samples to obtain the training importance value of each third training port. The loss value is positively correlated with the difference parameter of each of the N1 third training ports except for the third training port with the lowest importance. The difference parameter of any third training port except for the third training port with the lowest importance is the sum of at least one sub-difference of that third training port. The i-th sub-difference of any third training port is determined based on the training importance value of that third training port and the i-th training importance value of the i-th target training port corresponding to that third training port. In the importance order label, the importance of the i-th target training port corresponding to that third training port is lower than the importance of that third training port. If the first difference obtained by subtracting a preset difference value from the training importance value of any third training port is greater than or equal to the i-th training importance value, then the i-th sub-difference of any third training port is 0; if the first difference is less than the i-th training importance value, then the i-th sub-difference of any third training port is the second difference obtained by subtracting the first difference from the i-th training importance value, where the preset difference value is greater than 0. Here, i is a positive integer.

[0024] By setting a preset difference value greater than 0 during the training process, the trained port model can more accurately distinguish ports of different importance, thereby improving the data processing accuracy of the port model.

[0025] In some possible implementations, the loss value is obtained by weighted summation of the difference parameters of each of the N1 third training ports except for the third training port with the lowest importance. The weight of the difference parameter of any third training port is positively correlated with the importance of that third training port as represented by the importance order label.

[0026] During training, the loss value is calculated by assigning higher weights to ports with higher importance in the importance ranking labels. This ensures that the initial port model prioritizes the accuracy of its processing of these ports, resulting in a more accurate model for handling them. During inference, the port model's processing results are used by the network device to determine candidate ports. These candidate ports are the highest-importance ports among the network device's N ports. The higher the accuracy of the port model's processing of these more important ports, the more accurate the candidate ports determined by the network device.

[0027] In some possible implementations, the number of first ports is greater than the number of second ports. The number of first ports represents the number of ports in each of the Q port groups when a preset condition is met, and the number of second ports represents the number of ports in each of the Q port groups when the preset condition is not met. The preset condition is that the data transmission volume of the network device is less than a preset transmission volume threshold, and / or the movement speed of the user equipment within the coverage area of ​​the network device is greater than or equal to a preset speed threshold.

[0028] Based on preset conditions, the network device can adjust the number of ports in the port group, thus providing greater flexibility in setting the number of ports that send reference signals in a measurement cycle, and making the number of ports that send reference signals in each measurement cycle more adaptable to the current situation.

[0029] In some possible implementations, the first CSI report includes channel response features, which are obtained by feature extraction from a set of channel information, the set of channel information including complex channel gain parameters of K ports out of the M ports; the method further includes: processing the channel response features through a first prediction model to obtain downlink channel state information of the first user equipment, the downlink channel state information of the first user equipment including complex channel gain parameters of N ports in the network device.

[0030] The network device transmits a first reference signal through each of M ports, allowing the UE to obtain channel response characteristics based on the complex channel gain parameters of K ports. The network device processes the received channel response characteristics from the first UE using a first prediction model to obtain the first UE's downlink CSI. The first UE's CSI includes the complex channel gain parameters of N ports. Since N is greater than M, the network device transmits the reference signal through fewer ports, resulting in lower power consumption. Furthermore, compared to the number M ports used to transmit the first reference signal, the network device determines the first UE's CSI by including more ports corresponding to the complex channel gain parameters, making the first UE's downlink CSI more accurate. The network device can then communicate with the first UE based on the first UE's downlink CSI, enabling the first UE to receive the information transmitted by the network device more accurately.

[0031] In some possible implementations, the channel response features are obtained by the first UE through feature extraction on a sorted set of channel information. The complex channel gain parameters of multiple target ports in the sorted set of channel information are ordered by amplitude, and the channel response features include multiple port channel features. The method further includes: receiving port information sent by the first UE, where the port information represents the target port corresponding to each port channel feature in the multiple port channel features of the channel response features; according to the correspondence between ports and positions, and the port information, filling the port channel feature of any target port in the channel response features into the position corresponding to any target port in the initial features to obtain the expanded channel response features, where the N positions in the initial features are arranged according to the beam direction of the corresponding port. The processing of the channel response features through a first prediction model includes: processing the expanded channel response features through the first prediction model to obtain the downlink CSI of the first UE.

[0032] To extract more accurate channel response features, the first UE can sort the complex channel gains in the channel information set according to their amplitude before performing feature extraction. Channel response features include multiple port channel features. The port information that the first UE can send to the network device represents the target port corresponding to each port channel feature in the multiple port channel features of the channel response features. The signal transmitted through a port exhibits power leakage, meaning that the signal energy radiates beyond the expected main beam direction. The degree of power leakage in each direction of the port-transmitted signal is correlated with the directional difference between the beam direction corresponding to that port and the other directions. Before extracting features from the received channel response features, the network device can expand the complex channel gain parameters in the channel response features according to the beam direction corresponding to the port. Therefore, the downlink CSI of the first UE obtained by processing the expanded channel response features has higher accuracy.

[0033] Secondly, a communication method is provided, applied to a first UE. This method can be executed by the first UE, or by a component configured in the first UE (such as a circuit, chip, or chip system), or by a logic module or software capable of implementing all or part of the UE's functions. This application does not limit this. The following description uses a first UE (such as a satellite) as an example.

[0034] The method includes: receiving a first reference signal sent by a network device, the first reference signal being associated with M ports of the network device, where M is an integer greater than 1; and sending a first CSI report to the network device, the first CSI report being associated with N ports of the network device, where N is an integer greater than M.

[0035] In some possible implementations, the method further includes: receiving indication information sent by a network device, the indication information being used to instruct a first user equipment to report a CSI report associated with N ports based on reference signals at multiple times, the indication information also being used to indicate the port groups associated with the reference signals sent by the network device at the multiple times, wherein the reference signal sent by the network device at time t in the multiple times is associated with a first port group, the reference signal sent by the network device at time t-1 in the multiple times is associated with a second port group, the ports in the first port group are different from the ports in the second port group, and the M ports are ports in the first port group; the reception of the first reference signal may be to receive the reference signals sent by the network device at multiple times.

[0036] In some possible implementations, receiving the first reference signal can be achieved by receiving reference signals transmitted by the network device at multiple times. The method further includes: determining channel response information for each time period based on the reference signals at multiple times, wherein the channel response information for any time period includes the complex channel gain parameters of each port in the port group associated with the reference signal transmitted by the network device at that time, and the time t for transmitting the first reference signal is the latest time among the multiple times; determining a channel information set based on the channel response information for each time period, wherein the channel information set includes the complex channel gain parameters of M ports determined based on the first reference signal, and the complex channel gain parameters of the first port among N ports, wherein the port group associated with the reference signals at multiple times includes the first port, and the first port does not belong to any of the M ports; and determining a first CSI report based on the channel information set.

[0037] In some possible implementations, the network device transmits a first reference signal through each of its M ports for measurement. This can be achieved by periodically measuring the reference signal transmitted by the network device to obtain channel response information for each measurement period. The first channel response information is the channel response information for the t-th measurement period, where t is an integer greater than 1. The channel information set is determined based on the channel response information from multiple measurement periods. The channel response information for any one measurement period includes the complex channel gain parameters of the multiple ports that transmit the reference signal in that measurement period. The channel response information for these multiple measurement periods also includes the complex channel gain parameters of the first port (excluding the M ports) among the N ports of the network device.

[0038] Other ports besides the current M ports transmitted reference signals in previous measurement periods. Based on the channel response information from multiple measurement periods, a channel information set is determined, taking into account the influence of the complex channel gain parameters of the other ports determined in those other measurement periods. This makes the CSI of the first UE subsequently determined by the network device based on the channel information set more accurate.

[0039] In other words, the first channel response information includes the channel response information for the t-th measurement period. In the t-1 periods preceding the t-th measurement period, the network device transmitted a reference signal through a first port (outside of the M ports). The channel information set used to determine the first channel response information also includes the complex channel gain parameter of that first port. Therefore, the downlink CSI of the first UE determined by the network device based on the channel response characteristics is more accurate.

[0040] In some possible implementations, different times within multiple time periods belong to different measurement periods. For determining the channel information set, a second prediction model can be used to determine the channel response prediction information for the t-th measurement period based on the channel response information of the (t-1)-th measurement period. The channel response prediction information for the t-th measurement period includes the complex channel gain parameters of P ports out of N ports of the network device. These P ports include the first port. The channel information set includes the complex channel gain parameters of the first port in the channel response prediction information of the t-th measurement period. The ports associated with the reference signals transmitted by the network device in the (t-1)-th measurement periods prior to the t-th measurement period include these P ports.

[0041] Sending reference signals from a large number of ports requires significant communication resources. The location of the first UE may change in different measurement periods, causing the channel response information from the previous measurement period to inaccurately reflect the current channel response status of the first UE. By processing the channel response information from the (t-1)th measurement period using a second prediction model, the predicted channel response information for the tth measurement period is obtained. This predicted channel response information for the tth measurement period can more accurately reflect the channel status of P ports in the tth measurement period. In other words, the channel information set determined based on the predicted channel response information for the tth measurement period can more accurately reflect the channel response status of the first UE. Therefore, the CSI of the first UE subsequently determined by the network device based on the channel information set has higher accuracy.

[0042] In some possible implementations, determining the channel response prediction information for the t-th measurement period can include: processing the channel response prediction information for the (t-1)-th measurement period and the channel response information for the (t-2)-th measurement period using a second prediction model to obtain the channel response prediction information for the t-th measurement period. The channel response prediction information for the (t-1)-th measurement period is determined by the second prediction model based on the channel response information for the (t-2)-th measurement period, where t is greater than 2.

[0043] The first UE processes the channel response information and the predicted channel response information for one measurement period using the second prediction model to obtain the predicted channel response information for the next measurement period. Therefore, the determination of the channel information set considers the channel response information from multiple consecutive measurement periods, enabling the channel information set to more accurately reflect the downlink channel state from the network device to the first UE. Consequently, the downlink CSI of the first UE determined based on the channel information set is more accurate.

[0044] In some possible implementations, the method further includes: extracting features from the channel information set using a feature extraction model to obtain channel response features, wherein the channel information set includes complex channel gain parameters of K ports out of the M ports, the first CSI report includes the channel response features, and the channel response features are used by the network device to process the data through a first prediction model to obtain downlink channel state information of the first user equipment, wherein the downlink channel state information of the first user equipment includes complex channel gain parameters of N ports in the network device.

[0045] In some possible implementations, the feature extraction model and the first prediction model are obtained by adjusting the parameters of the initial feature extraction model and the first initial prediction model respectively based on the difference between the training channel state and the channel state label. The training channel state is obtained by processing the training channel response features through the first initial prediction model. The training channel response features are obtained by processing the training channel information set through the initial feature extraction model. The training channel information set includes first training channel response information, which is obtained by measuring the first training reference signal transmitted by the first training user equipment through each of the M' first training ports of the first training network device. The first training channel response information includes the complex channel gain parameters of K' of the M' first training ports. The channel state label includes the complex channel gain parameters of N first training ports of the first training network device. The N first training ports include the M' first training ports, where M' and K' are both positive integers greater than 1, and the N first training ports of the first training network device are the same as the N ports of the network device.

[0046] In some possible implementations, the second prediction model is obtained by adjusting the parameters of the second initial prediction model based on the difference. The first training channel response information is the training channel response information of the m-th training measurement period. The training channel information set also includes the complex channel gain parameter of the target first training port in the training channel response prediction information of the m-th training measurement period. The training channel response prediction information of the m-th training measurement period is determined by the second initial prediction model based on the training channel response information of the (m-1)-th training measurement period. The channel response prediction information of the m-th measurement period includes the complex channel gain parameter of P' first training ports among the N first training ports. The target first training port belongs to the P' first training ports and does not belong to the M' first training ports, where m is an integer greater than 1.

[0047] The channel information processing system comprises a feature extraction model, a first prediction model, and a second initial prediction model. The system is trained end-to-end. End-to-end training simplifies the training process and reduces the amount of data required for training. Furthermore, end-to-end training improves the data processing accuracy of the channel information processing system.

[0048] In some possible implementations, the method further includes: sorting the complex channel gain parameters of multiple target ports in the channel information set according to their magnitude to obtain a sorted channel information set, wherein the multiple target ports include K ports corresponding to the complex channel gain parameters in the first channel response information; determining the channel response features includes: extracting features from the sorted channel information set using a feature extraction model to obtain channel response features, wherein the channel response features include the port channel features of each target port; the method further includes: the first UE sending port information to the network device, wherein the port information is used to indicate the target port corresponding to each port channel feature in the channel response features.

[0049] The complex channel gain parameters of the target ports exhibit sparsity, meaning that the amplitude attenuation of the reference signals transmitted from multiple target ports to the first UE varies significantly. Feature extraction from sparse data results in low accuracy and stability. By sorting the complex channel gains in the channel information set according to their amplitude and performing feature extraction on the sorted channel information set, the downlink CSI accuracy of the first UE determined by the channel response characteristics is significantly improved.

[0050] In some possible implementations, the K ports corresponding to the complex channel gain parameters in the first channel response information are the M ports through which the network device transmits the first reference signal.

[0051] The signal transmitted through the port exhibits power leakage, meaning that the signal energy radiates beyond the beam direction corresponding to that port. The first channel response information includes the complex channel gain parameters of all M ports through which the network device transmits the first reference signal, while the channel information set includes the first channel response information. Therefore, the downlink CSI of the first UE determined based on the channel information set has higher accuracy.

[0052] The second aspect is the implementation on the UE side, which corresponds to the first aspect. The explanations, supplements, and descriptions of the beneficial effects of the first aspect also apply to the second aspect, and will not be repeated here.

[0053] Thirdly, a communication method is provided, applied to a network device. This method can be executed by the network device itself, or by components configured within the network device (such as circuits, chips, or chip systems), or by logic modules or software capable of implementing all or part of the network device's functions. This application does not limit the scope of this method. The following description uses a network device as an example.

[0054] The method includes: acquiring downlink CSIs of multiple UEs, where each UE's downlink CSI includes complex channel gain parameters of multiple ports among N ports of the network device, where different ports in the network device correspond to different beam directions, and N is a positive integer greater than 1; processing the downlink CSIs of the multiple UEs using a port model to obtain the importance of each of the N ports; determining multiple candidate ports among the N ports based on the importance of each port, where the importance of each candidate port is higher than the importance of other ports among the N ports besides the candidate ports; transmitting reference signals through the multiple candidate ports, where the reference signals transmitted by any two candidate ports are orthogonal; receiving channel information transmitted by a first UE, where the channel information is determined by the first UE based on the complex channel gain parameters of the port in the network device that transmits the reference signal; and determining the downlink CSI of the first UE based on the channel information.

[0055] The communication method provided in this application determines the importance of a port based on the downlink CSI of multiple UEs. Each UE's downlink CSI is influenced by its location and surrounding environment, thus the port's importance reflects the network device's environment's demand for each port. The network device transmits reference signals through candidate ports with higher importance, adapting the transmission of reference signals to its environment, thereby making the downlink CSI of the first UE more accurate.

[0056] Fourthly, a communication method is provided, applied to a network device, the method comprising: transmitting reference signals at Q1 time intervals, wherein the reference signals transmitted at different time intervals during the Q1 time intervals are associated with different port groups, the port groups associated with the reference signals transmitted at the Q1 time intervals include multiple candidate ports, and Q1 is an integer greater than 1; the network device receiving a first CSI report transmitted by a first user equipment, the first CSI report being associated with a port set, the port set including the multiple candidate ports.

[0057] For example, a first reference signal sent by the network device at a first time is associated with a third port group, and a second reference signal sent by the network device at a second time is associated with a fourth port group. The ports in the third port group and the fourth port group are different.

[0058] The reference signals transmitted by the network device in Q1 time periods are associated with different port groups. Therefore, at each time period, the reference signals transmitted by the network device can be associated with a subset of candidate ports, reducing the resource consumption of transmitting reference signals at each time period.

[0059] In some possible implementations, the first CSI report includes at least one of the following: precoding matrix information, codebook information, feature information, and downlink channel response information, wherein the precoding matrix information is associated with the port set, the codebook information is associated with the port set, the feature information is associated with the port set, and the downlink channel response information is associated with the port set.

[0060] The first CSI report can be used by network devices to determine the precoding result, and then send information to the first UE based on the precoding result.

[0061] In some possible implementations, the method further includes: receiving Q1-1 second CSI reports sent by the first user equipment, wherein the port group associated with any second CSI report is the port group associated with the reference signal sent by the network device at the time corresponding to any second CSI report in the preceding N-1 times of the Q1 times.

[0062] In the different Q1-1 second CSI reports, the different second CSI reports correspond to different times in the first N-1 times of that Q1 time period.

[0063] If the network device fails to transmit the reference signal at time Q1, for any reference signal transmitted by the network device at any time, the first UE may send a second CSI report corresponding to that time to the network device based on the reference signal transmitted by the network device at that time.

[0064] In some possible implementations, the method further includes: receiving Q1 second CSI reports sent by the first user equipment, wherein the port group associated with any second CSI report is the port group associated with the reference signal sent by the network device at the time corresponding to any second CSI report in the Q1 time period.

[0065] In some possible implementations, the method further includes sending an indication message to the first user equipment, the indication message instructing the first user equipment to report a CSI report associated with the port set.

[0066] As instructed by the network device, the first UE may send a first CSI report to the network device after the network device has completed the transmission of the reference signal for the Q1 time period.

[0067] In some possible implementations, the method further includes sending indication information to the first user equipment, the indication information indicating the port group associated with the reference signal transmitted in the Q1 time intervals.

[0068] In some possible implementations, the port set includes N ports in the network device, the N ports include the plurality of candidate ports, and N is greater than the number of the plurality of candidate ports.

[0069] The port associated with the reference signal transmitted by the network device is the candidate port. The first UE reports the first CSI report for N ports in the associated network device. The number of candidate ports is less than the number of ports N in the network device. The network device can send information to the first UE based on the channel information of more ports in the first CSI report, thus making information transmission more efficient.

[0070] In some possible implementations, the candidate port is considered more important than the other ports among the N ports.

[0071] Candidate ports can be determined by the network device. By associating the transmitted reference signals with ports of higher importance, the first CSI report becomes more accurate.

[0072] In some possible implementations, the method further includes configuring the following information for the first user equipment: the reference signal resources corresponding to the port group associated with the reference signals transmitted in the Q1 time intervals.

[0073] In some possible implementations, the port group associated with the reference signal transmitted in the Q1 time period includes a fifth port group and a sixth port group, and the reference signal resources corresponding to the fifth port group and the reference signal resources corresponding to the sixth port group include different frequency domain resource locations.

[0074] Wireless channels exhibit frequency-selective fading, meaning that channel quality can vary significantly across different frequencies. By distributing the reference signal across multiple frequency domain locations, a more comprehensive and accurate sampling of the channel state across the entire system bandwidth can be achieved.

[0075] In some possible implementations, the method further includes: configuring the first user equipment with a reference signal sequence corresponding to the port group associated with the reference signals transmitted in the Q1 time intervals.

[0076] In some possible implementations, the port groups associated with the reference signals transmitted in the Q1 time intervals include a seventh port group and an eighth port group, and the reference signal sequence corresponding to the seventh port group is different from the reference signal sequence corresponding to the eighth port group.

[0077] In some possible implementations, the Q1 times belong to time periods with a duration less than or equal to a preset duration.

[0078] In some possible implementations, the method further includes: The network device groups the multiple candidate ports to obtain the port groups associated with the reference signals transmitted at the Q1 time intervals.

[0079] In some possible implementations, the different times in the Q1 times belong to different measurement periods.

[0080] Fifthly, a communication method is provided, applied to a first user equipment, the method comprising: receiving reference signals transmitted by a network device at Q1 time intervals, wherein the reference signals transmitted at different time intervals in the Q1 time intervals are associated with different port groups, and the port groups associated with the reference signals transmitted at the Q1 time intervals include multiple candidate ports; and sending a first CSI report to the network device, the first CSI report being associated with the port set, the port set including the multiple candidate ports.

[0081] In some possible implementations, the first CSI report includes at least one of the following: precoding matrix information, codebook information, feature information, and downlink channel response information, wherein the precoding matrix information is associated with the port set, the codebook information is associated with the port set, the feature information is associated with the port set, and the downlink channel response information is associated with the port set.

[0082] In some possible implementations, the method further includes: sending Q1-1 second CSI reports to the network device, wherein the port group associated with any second CSI report is the port group associated with the reference signal sent by the network device at the time corresponding to any second CSI report in the preceding N-1 times of the Q1 times.

[0083] In some possible implementations, the method further includes: sending Q1 second CSI reports to the network device, wherein the port group associated with any second CSI report is the port group associated with the reference signal sent by the network device at the time corresponding to any second CSI report in the Q1 time period.

[0084] In some possible implementations, the system receives an instruction message from a network device that instructs the first user device to report a CSI report associated with the port set.

[0085] In some possible implementations, the method further includes receiving indication information sent by a network device, the indication information indicating the port group associated with the reference signal transmitted in the Q1 time period.

[0086] In some possible implementations, the port set includes N ports in the network device, which include the plurality of candidate ports, and N is greater than the number of the plurality of candidate ports.

[0087] In some possible implementations, the method further includes: receiving configuration information sent by a network device, the configuration information including reference signal resources corresponding to the port group associated with the reference signal transmitted in the Q1 time period.

[0088] In some possible implementations, the port groups associated with the reference signals transmitted in the Q1 time intervals include a fifth port group and a sixth port group, and the reference signal resources corresponding to the fifth port group and the reference signal resources corresponding to the sixth port group include different frequency domain resource locations.

[0089] In some possible implementations, the method further includes: receiving configuration information sent by a network device, the configuration information indicating the reference signal sequence corresponding to the port group associated with the reference signal transmitted at Q1 time intervals.

[0090] In some possible implementations, the port groups associated with the reference signals transmitted in the Q1 time period include a seventh port group and an eighth port group, and the reference signal sequence corresponding to the seventh port group is different from the reference signal sequence corresponding to the eighth port group.

[0091] In some possible implementations, the Q1 times belong to a time period whose duration is less than or equal to a preset duration.

[0092] In some possible implementations, the different times in the Q1 times belong to different measurement periods.

[0093] The fifth aspect is the implementation on the UE side, which corresponds to the fourth aspect. The explanations, supplements, and descriptions of the beneficial effects of the fourth aspect also apply to the fifth aspect, and will not be repeated here.

[0094] Sixthly, a communication apparatus is provided, comprising units for performing the methods in any possible implementation of any of the preceding aspects. The apparatus may be an electronic device or a chip within an electronic device. The electronic device may be a network device or a user equipment.

[0095] A seventh aspect provides a communication device including a processor. The processor is coupled to a memory and can be used to execute instructions or data in the memory to implement the methods in any possible implementation of any of the preceding aspects. Optionally, the communication device further includes a memory. Optionally, the communication device further includes a communication interface, and the processor is coupled to the communication interface.

[0096] In one implementation, the communication interface may be a transceiver, or an input / output interface.

[0097] In another implementation, the communication device is a chip configured in a user equipment. When the communication device is a chip configured in a user equipment, the communication interface can be an input / output interface.

[0098] Eighthly, a processor is provided, comprising: an input circuit, an output circuit, and a processing circuit. The processing circuit is configured to receive signals through the input circuit and transmit signals through the output circuit, causing the processor to execute a method in any possible implementation of any of the preceding aspects.

[0099] In specific implementation, the processor can be one or more chips, the input circuit can be input pins, the output circuit can be output pins, and the processing circuit can be transistors, gate circuits, flip-flops, and various logic circuits. The input signal received by the input circuit can be received and input by, for example, but not limited to, a receiver, and the signal output by the output circuit can be, for example, but not limited to, output to and transmitted by a transmitter. Furthermore, the input circuit and the output circuit can be the same circuit, which is used as both the input circuit and the output circuit at different times. This application does not limit the specific implementation of the processor and various circuits.

[0100] A ninth aspect provides a communication device including a processor and a memory. The processor is configured to read instructions stored in the memory, receive signals via a receiver, and transmit signals via a transmitter to execute the method in any possible implementation of any of the preceding aspects.

[0101] Optionally, the processor may be one or more, and the memory may be one or more.

[0102] In a tenth aspect, a computer program product is provided, the computer program product comprising: a computer program (also referred to as code or instructions) that, when the computer program is run, causes a computer to perform a method in any possible implementation of any of the above aspects.

[0103] Eleventhly, a computer-readable storage medium is provided that stores a computer program (also referred to as code or instructions) that, when run on a computer, causes the computer to perform the method in any possible implementation of any of the preceding aspects.

[0104] In a twelfth aspect, embodiments of this application provide a chip system including one or more processors for calling and executing instructions stored in memory, causing the methods in any of the above aspects or possible implementations to be performed. The chip system may be composed of chips or may include chips and other discrete devices.

[0105] The chip system may include input circuits or interfaces for transmitting information or data, and output circuits or interfaces for receiving information or data.

[0106] In a thirteenth aspect, a communication system is provided, including the aforementioned first UE and network device. Optionally, the communication system may further include other devices that communicate with the first UE and / or the network device. Attached Figure Description

[0107] Figure 1 This is a schematic diagram of a communication system applied in an embodiment of this application; Figure 2 This is a schematic flowchart of a communication method in a TDD system; Figure 3 This is a schematic flowchart of a communication method in an FDD system; Figure 4 This is a schematic diagram of the beam direction; Figure 5 This is a schematic diagram showing how the total power of the signals at the ports changes with the number of ports. Figure 6 This is a schematic diagram illustrating the relationship between the UE's location and the network device's ports; Figure 7 This is a diagram illustrating the accuracy of the downlink CSI. Figure 8 This is a schematic flowchart illustrating a communication method provided in an embodiment of this application; Figure 9 This is a schematic diagram illustrating the relationship between the candidate port update period and the measurement period provided in an embodiment of this application; Figure 10 This is a schematic flowchart illustrating another communication method provided in an embodiment of this application; Figure 11 This is a schematic diagram of the candidate ports and port groups provided in the embodiments of this application; Figure 12 This is a schematic structural diagram of a port model provided in an embodiment of this application; Figure 13 This is a schematic flowchart illustrating a training method provided in an embodiment of this application; Figure 14 This is a schematic structural diagram of a channel information processing system provided in an embodiment of this application; Figure 15 This is a schematic diagram of data processing for the second prediction model provided in an embodiment of this application; Figure 16 This is a schematic diagram of data processing for the second prediction model provided in an embodiment of this application; Figure 17This is a schematic flowchart illustrating another training method provided in an embodiment of this application; Figure 18 This is a schematic diagram of power leakage; Figure 19 This is a schematic diagram of the quantization layer and the smoothed differentiable function provided in the embodiments of this application; Figure 20 This is a schematic flowchart illustrating another communication method provided in the embodiments of this application; Figure 21 This is a schematic diagram of the beam direction of uniformly distributed ports and candidate ports; Figures 22 to 26 This is a schematic diagram of the UE downlink CSI error of the communication method provided in the embodiments of this application; Figure 27 This is a schematic diagram illustrating the spectral efficiency of the communication method provided in the embodiments of this application; Figure 28 This is a schematic flowchart illustrating another communication method provided in the embodiments of this application; Figure 29 This is a schematic block diagram of a communication device provided in an embodiment of this application; Figure 30 This is another schematic block diagram of the communication device provided in the embodiments of this application. Detailed Implementation

[0108] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0109] It should be understood that "at least one" in this application refers to one or more, and "more than one" refers to two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B; "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist, for example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, to facilitate a clear description of the technical solutions of this application, the terms "first," "second," etc., are used to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first," "second," etc., do not limit the quantity or execution order, and that "first," "second," etc., do not necessarily imply differences.

[0110] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0111] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties (e.g., the user has given explicit consent, the user has been notified, etc.), and the collection, use and processing of the relevant data must comply with the relevant regulations and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0112] Since the embodiments of this application involve the application of a large number of neural network models, for ease of understanding, the relevant terms and concepts such as neural network models involved in the embodiments of this application will be introduced below.

[0113] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0114] Machine learning is an important branch of artificial intelligence, and deep learning is an important branch of machine learning. Deep learning refers to using multi-layer neural network structures to learn representations of various things in the real world that can be directly used for computer calculations from big data.

[0115] A neural network model is a mathematical method for simulating the actual neural network in humans. A neural network model is a complex network system formed by extensively interconnecting a large number of simple processing units (called neurons or neural units). A neural network model is a network formed by linking many of these individual neural units together; that is, the output of one neural unit can be the input of another neural unit. The input of each neural unit can be connected to the local receptive field of the previous layer to extract features from the local receptive field, which can be a region composed of several neural units.

[0116] Deep neural networks (DNNs), also known as multi-layer neural networks, can be understood as neural networks with multiple hidden layers. Based on the position of the layers, DNNs can be divided into three categories: input layers, hidden layers, and output layers. Generally, the first layer is the input layer, the last layer is the output layer, and the layers in between are hidden layers. Layers are fully connected, meaning that any neuron in the i-th layer is connected to any neuron in the (i+1)-th layer. In deep neural networks, more hidden layers allow the network to better represent complex real-world situations. Theoretically, the more parameters a model has, the higher its complexity and "capacity," meaning it can perform more complex learning tasks. Training a deep neural network is essentially the process of learning the weight matrix, with the ultimate goal of obtaining the weight matrix (a weight matrix formed by vectors from many layers) of the trained deep neural network.

[0117] In training a deep neural network model, to ensure the model's output closely approximates the desired predicted value, we compare the network's prediction with the target value. Based on the difference, we update the weight vector of each layer (usually pre-configuring parameters before the initial update). For example, if the prediction is too high, the weight vector is adjusted to predict a lower value. This adjustment continues until the model predicts the target value accurately or very closely. Therefore, we need to predefine "how to compare the difference between the predicted and target values," which is the loss function or objective function. These are crucial equations for measuring the difference between the predicted and target values. Taking the loss function as an example, a higher loss value indicates a greater difference, and training the deep neural network model becomes a process of minimizing this loss.

[0118] Backpropagation (BP) is an algorithm used during training to correct the parameters of an initial neural network model, thereby minimizing the reconstruction error loss. Specifically, forward propagation of the input signal to the output generates an error loss; this error loss information is then propagated back to update the parameters of the initial neural network model, leading to convergence of the error loss. The backpropagation algorithm is an error-loss-driven backpropagation process aimed at obtaining the optimal parameters of the neural network model, such as the weight matrix.

[0119] A gated recurrent unit (GRU) is a type of deep neural network that includes a reset gate and an update gate. The reset gate controls the degree to which information from the previous time step influences the current information. The update gate controls the weights between the information from the previous time step and the current information. GRU uses a gating mechanism to balance long-term and short-term memory, achieving efficient modeling of sequential data with a more concise structure.

[0120] A multilayer perceptron (MLP) is a deep neural network consisting of an input layer, one or more hidden layers, and an output layer. Each hidden layer contains multiple neurons, and each neuron is connected to all neurons in the layer preceding it. In other words, every hidden layer in an MLP is a fully connected layer. The output layer can use either a linear or non-linear activation function.

[0121] A convolutional neural network (CNN) is a deep neural network with a convolutional structure. A CNN typically contains a feature extractor consisting of convolutional layers and pooling layers (also called subsampling layers), which automatically learns hierarchical feature representations of the input data. The convolution process can be viewed as using a set of trainable filters (i.e., convolutional kernels) to locally scan and compute the input data or the feature map of the previous layer. In a convolutional layer, each neuron is connected to only a portion of the input data; this local connectivity significantly reduces the number of parameters. A convolutional layer typically contains multiple feature maps, each composed of neurons arranged in two or higher dimensions. Neurons within the same feature map share the same convolutional kernel weights. This weight-sharing mechanism makes feature extraction translation-invariant, meaning that patterns learned from one region of data can be used to identify similar patterns in other regions. The underlying assumption is that the local statistical properties of data are repeatable across different locations, thus the model can learn globally effective feature representations through local perception.

[0122] The technical solutions of this application embodiment can be applied to various communication systems, such as: Global System for Mobile Communications (GSM) system, Code Division Multiple Access (CDMA) system, Wideband Code Division Multiple Access (WCDMA) system, General Packet Radio Service (GPRS), Long Term Evolution (LTE) system, LTE Frequency Division Duplex (FDD) system, LTE Time Division Duplex (TDD) system, Universal Mobile Telecommunication System (UMTS), Worldwide Interoperability for Microwave Access (WiMAX) system, 5th Generation (5G) mobile communication system or New Radio Access Technology (NR), Wireless Local Area Network (WLAN), Sidelink communication system, Non-Terrestrial Network (NTN) communication system, Satellite communication system, etc. The 5G mobile communication system may include non-standalone (NSA) and / or standalone (SA) networking. The technical solutions provided in this application can also be applied to future communication systems. This application does not limit the scope of the application.

[0123] Figure 1 This is a schematic diagram of a communication system 100 used in an embodiment of this application. The communication system 100 may include network devices, such as... Figure 1 The network device 110 is shown. The communication system 100 may also include user equipment (UE), such as... Figure 1 The UEs 121 to 123 are shown. Network device 110 can communicate with UEs 121 to 123 via a wireless link.

[0124] The network device in this application can be an access network device, a core network device, or other network-side device. The network device may have access network functionality. That is, network device 110 can be used to connect UEs 121 to 123 within its coverage area to the radio access network (RAN). Access network devices are sometimes also called access nodes. Access network devices have radio transceiver capabilities for communicating with UEs.

[0125] Access networks are sometimes also referred to as base stations (BS), access devices, or access network nodes. It is understood that the names of devices with base station functions may differ in systems employing different wireless access technologies. For ease of description, this application embodiment collectively refers to devices providing wireless communication access functions to UEs as access network devices. Access network devices include, but are not limited to, base stations, evolved NodeBs (eNodeBs), transmission reception points (TRPs), evolved node Bs (eNodeBs, eNBs) in long term evolution (LTE), next-generation node base stations (gNBs) in fifth-generation (5G) mobile communication systems, access network devices or modules of access network devices in open RAN (ORAN) systems, satellites in NTN communication systems, base stations in future mobile communication systems, or access nodes in WiFi systems. Access network devices can also be modules or units capable of implementing some of the functions of a base station. Access network equipment can be a macro base station, a micro base station, an indoor station, a relay node, a donor node, or a wireless controller in a cloud radio access network (CRAN) scenario. Optionally, access network equipment can also be a server, a wearable device, or an in-vehicle device. For example, in vehicle-to-everything (V2X) technology, the access network equipment can be a roadside unit (RSU). Multiple access network equipment in a communication system can be base stations of the same type or different types. Base stations can communicate with the UE or via relay stations. The UE can communicate with multiple base stations using different access technologies. The embodiments of this application do not limit the specific technology or device form used in the access network equipment. Access network equipment can also be a roadside device with wireless access functionality or a terminal. In the embodiments of this application, equipment capable of implementing the functions involved in the access network equipment in the embodiments of this application is collectively referred to as network equipment.

[0126] In this application, the means for implementing the functions of a network device can be a network device itself, or a means capable of supporting the network device in implementing those functions, such as a processor, circuit, chip, or chip system. This means can be installed in or connected to the network device. In the technical solutions provided in this application, the example of a network device being used to implement the functions of a network device is used to describe the technical solutions provided in this application.

[0127] In this application embodiment, the user equipment (UE) can be a wireless terminal device capable of receiving network device scheduling and instruction information. The wireless terminal device can be a device that provides voice and / or data connectivity to the user, a handheld device with wireless connectivity, or other processing devices connected to a wireless modem. For example, the UE can communicate with one or more core networks or the Internet via a radio access network (RAN). The terminal device can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), Internet of Things (IoT), ultra-reliable low-latency communication (URLLC), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grids, smart furniture, smart offices, smart wearables, smart transportation, smart cities, or satellite communication, etc. UE can refer to terminal equipment, terminal, access terminal, user unit, user station, mobile station, mobile station, mobile terminal, remote station, remote terminal, mobile device, user terminal, wireless communication equipment, user agent, or user equipment. The UE can also be a cellular phone, cordless phone, session initiation protocol (SIP) phone, wireless local loop (WLL) station, personal digital assistant (PDA), handheld device with wireless communication capabilities, computing device or other processing device connected to a wireless modem, in-vehicle device, wearable device, terminal device in future 5G networks or terminal device in future evolved public land mobile network (PLMN), etc. It can also be an end device, logical entity, smart device, such as mobile phone, tablet computer, computer with wireless transceiver capabilities, wearable device, vehicle, aircraft (e.g., drone, helicopter, airplane), hot air balloon, ship, robot, robotic arm, or smart home device, or communication equipment such as server, gateway, base station, controller, etc., or Internet of Things (IoT) device, such as sensor, electricity meter, water meter, etc. The embodiments of this application do not limit the form of the terminal device.

[0128] In this application, the apparatus for implementing the functions of a terminal device can be the terminal device itself, or any apparatus capable of supporting the terminal device in implementing those functions, such as a processor, circuit, chip, or chip system. This apparatus can be installed in or connected to the terminal device. In the technical solutions provided in this application, the example of a terminal device being used to implement the functions of a terminal device is used to describe the technical solutions provided in this application.

[0129] Access network devices and / or terminals can be fixed or mobile. They can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; on water; or in the air on aircraft, balloons, and satellites. This application does not limit the application scenarios of the access network devices and terminals. Access network devices and terminal devices can be deployed in the same or different scenarios; for example, both can be deployed on land; or the access network device can be deployed on land, and the terminal device on water, etc., and so on.

[0130] Network devices can be equipped with multiple-input multiple-output (MIMO) antennas, such as massive MIMO antennas. MIMO antennas enable beamforming, an antenna signal processing technique. By adjusting the signal phase and amplitude of each antenna element in an antenna array, the signals transmitted by multiple antennas are coherently amplified at a specific point in space (i.e., the target user's location) and coherently canceled out at other points.

[0131] To enable network devices to transmit information more accurately to user equipment (UE), the communication system, including both network devices and UEs, can perform channel sounding to obtain downlink channel state information (CSI) for the UE. Channel sounding is the process in wireless communication systems of actively transmitting a known reference signal and estimating the characteristics of the wireless channel by analyzing changes in the received signal. In TDD systems, uplink and downlink use the same frequency, and the path characteristics of electromagnetic wave propagation (such as attenuation, reflection, and scattering) can be considered identical in both directions. That is, there is strong reciprocity between the uplink and downlink channels. Therefore, the communication system, including network devices and UEs, can perform channel sounding such as... Figure 2 The communication method is used to determine the downlink CSI of the UE.

[0132] Figure 2 This is a schematic flowchart of a communication method in a TDD system.

[0133] In step S201, the network device sends indication information to the first UE, the indication information including reference signal resources.

[0134] Reference signal resources can indicate the sequence of reference signals and time-frequency resources, etc.

[0135] In step S202, the first UE sends a reference signal according to the reference signal resource in the instruction information.

[0136] The first UE can transmit the sequence of reference signals indicated by the first indication information on the time and frequency resources indicated by the first indication information to complete the transmission of the reference signals.

[0137] The reference signal can be a preset digital sequence. The reference signal transmitted by the first UE can be a pilot sequence, such as a sounding reference signal (SRS).

[0138] Step S203: The network device determines the uplink CSI of the first UE.

[0139] The uplink CSI of the first UE includes the amplitude and phase of the reference signal received at each port of the network device's antenna array. In other words, the uplink CSI of the first UE can include the complex channel gain of each port in the network device. The complex channel gain of each port can represent amplitude attenuation and phase shift.

[0140] An antenna port is a logical concept, representing the endpoint of the signal processing link in a network device. In the antenna array of a network device, hundreds or thousands of physical antenna elements can be grouped and mapped to a much smaller number of logical receive ports. During propagation in the wireless channel, the reference signal undergoes fading (amplitude variation) and phase shift (phase change). The uplink CSI can represent the wireless link state from the first UE to each antenna port in the network device.

[0141] In step S204, the network device transposes the uplink CSI of the first UE to obtain the downlink CSI of the first UE.

[0142] Therefore, in a TDD system, network devices can measure the uplink channel and determine the downlink CSI of the UE based on the uplink channel measurement results.

[0143] In FDD systems, uplink and downlink use different frequencies, and the uplink and downlink channels are not reciprocal. During channel sounding, the UE can measure the reference signal sent by the network device to determine the downlink CSI and then send the downlink CSI back to the network device.

[0144] For MIMO antennas, especially massive MIMO antennas, the number of ports is large. Network devices transmit reference signals through each port, consuming significant communication resources. Furthermore, the network device feeds back each received reference signal, resulting in substantial feedback overhead. To reduce communication resource consumption, the network device can transmit reference signals through only some of the multiple ports and determine the UE's downlink CSI based on the UE's measurement of the received reference signals.

[0145] Figure 3 This is a schematic flowchart of a communication method in an FDD system.

[0146] Step S301: The network device sends an indication message, which indicates the reference signal resource.

[0147] For example, a network device can send instruction information via broadcast. UEs within the network device's coverage area can receive the instruction information.

[0148] Reference signal resources can indicate multiple ports that transmit reference signals, as well as the sequence and time-frequency resources of the reference signals transmitted by each of these ports.

[0149] A port is a logical endpoint for signal transmission and reception. It serves as the logical interface between the base station and the terminal for wireless resource scheduling, channel measurement, and data transmission. In other words, an antenna port can be implemented by one or more antenna elements.

[0150] When beamforming weight vectors are applied to an antenna array, they form a beam with a specific radiation pattern in space. The codebook recorded by the network device can include multiple codewords, with different codewords representing different beamforming weight vectors.

[0151] During channel probing, network devices can configure a specific codeword (i.e., a specific predefined beam) for each port used for probing. Different ports can correspond to different codewords. A port can be understood as the logical entity through which the codeword is applied.

[0152] When a network device transmits a reference signal through this port, it can be understood as applying the beamforming weight vector represented by the codeword corresponding to that port to the antenna array; that is, using the beamforming weight vector corresponding to the codeword to drive the antenna array to transmit. The signals transmitted by each antenna element in the antenna array coherently superimpose in space, amplifying in some directions and canceling out in others, thus forming a directional beam in space. In other words, during channel detection, different ports correspond to different beam directions. Different ports can also be understood as logical identifiers for different detection beams. Figure 4 (a) shows the beam directions corresponding to multiple ports in the network device.

[0153] The codebook provides limited and standardized options for beamforming. The codebook can be stored in both network devices and UEs. This allows network devices and UEs to easily identify the ports of network devices and the beams used, reducing communication overhead.

[0154] A network device may include multiple ports, each corresponding to a different beam direction. The port that transmits the reference signal may be all or some of the multiple ports of the network device. Figure 4 (b) shows the beam directions corresponding to the multiple ports used to transmit the reference signal.

[0155] In step S302, the network device sends a reference signal according to the reference signal resource.

[0156] The reference signals transmitted from any two ports are orthogonal. For example, the reference signals can be time-division orthogonal, frequency-division orthogonal, or code-division orthogonal. Thus, the UE can distinguish signals from different ports.

[0157] Frequency division orthogonality (FDI) can be understood as dividing the total bandwidth into multiple non-overlapping, narrower sub-bands, with each signal occupying its own sub-band for transmission. Time division orthogonality (TDI) can be understood as dividing the time axis into periodically repeating frames, with each frame further subdivided into several time slots. Each signal is transmitted in a different time slot. Code division orthogonality (CDO) is a core technology in wireless communication that enables multiple signals to coexist on the same time and frequency without interference. Two sequences being code-digonal means that their inner product is zero.

[0158] The reference signal transmitted by the port can be a pilot sequence, such as a channel state information reference signal (CSI-RS). The signal strength of the reference signal transmitted by different ports is the same.

[0159] Step S303: The first UE measures the complex channel gain of each port that transmits the reference signal.

[0160] The first UE can be a UE within the coverage area of ​​the network device.

[0161] The UE receives a reference signal and, utilizing orthogonality, separates the signal from each port through correlation detection. The UE estimates the equivalent channel gain for each port. The equivalent channel gain can be represented by a complex number. The equivalent channel gain for each port can also be called the complex channel gain of that port.

[0162] During propagation in a wireless channel, the reference signal undergoes fading (amplitude change) and phase shift (phase change). The complex channel gain of a port can include amplitude and phase parameters. The amplitude parameter represents the amplitude attenuation of the reference signal transmitted by that port received by the UE. The phase parameter represents the phase shift of the reference signal transmitted by that port received by the UE.

[0163] In step S304, the first UE sends a precoding matrix indicator (PMI) to the network device based on the complex channel gain of each port transmitting the reference signal.

[0164] PMI can include the complex channel gain for each port that transmits the reference signal.

[0165] Because the signals transmitted by each port are directional, among the signals received by the UE from each port, some ports have higher signal power, while most other ports transmit signals with very low power. Figure 5 The diagram shows the power ratio of the total power of the signals from the m highest-power ports received by the first UE to the total power of all transmitted reference signals received by the first UE, when the number of ports transmitting reference signals is 32. As m increases, the power ratio increases, but the rate of change of the power ratio with respect to m decreases. When m is 6, the power ratio reaches 0.9; when m is 10, the power ratio exceeds 0.95. However, as m increases from 10 to 32, the power ratio increases by less than 0.05.

[0166] In other words, CSI exhibits sparsity. The channel responses of the L ports with higher power can more accurately reflect the downlink CSI of the UE.

[0167] To reduce data transmission volume and thus reduce the consumption of communication resources, the first UE can provide feedback on the channel state of some ports. That is, the PMI can include the complex channel gain of L reference signals received by the first UE. The PMI can also indicate the L ports that respectively transmit the L reference signals. Figure 4 (c) shows the beam directions corresponding to the L ports indicated by the PMI sent by the first UE.

[0168] The first UE sorts multiple ports in the network device that transmit reference signals according to the magnitude of amplitude attenuation represented by the complex channel gain, and selects the L ports with the smallest amplitude attenuation, i.e., the ports with the largest amplitude and strongest signal power of the reference signal received by the first UE. L is less than the number of ports that transmit reference signals, and L can be configured by the protocol or the network device.

[0169] To further reduce data transmission volume, the first UE can quantize the complex channel gain. For the port amplitude parameter, the first UE can use logarithmic quantization; for the port phase parameter, the first UE can use uniform quantization. The PMI can include the quantization results of the complex channel gain of L reference signals.

[0170] Step S305: The network device determines the downlink CSI of the first UE based on the PMI.

[0171] The network device can inverse quantize the quantization results of the complex channel gains of L ports in the PMI to obtain the downlink CSI of the first UE. The downlink CSI of the first UE includes the inverse quantized complex channel gain of each of the L ports.

[0172] After step S204 or step S305, the network device can perform precoding based on the downlink CSI of the first UE, and control the antenna array to transmit information according to the precoding result, so that the signal energy is accurately focused on the first UE.

[0173] For electromagnetic wave frequencies below 6 gigahertz (GHz), the longer wavelengths result in stronger diffraction and penetration capabilities, leading to a rich array of signal propagation paths from network devices to the user equipment (UE) (strong multipath effect). This means that signal energy does not originate from just one or two main directions (sparse), but is distributed across multiple different angles of arrival. Consequently, the angular power spectrum of the channel is diffuse and non-sparse. Therefore, for electromagnetic wave frequencies below 6 GHz, L can be an integer greater than 1 to allow the precoding results to better suppress multi-user interference and achieve greater accuracy.

[0174] The precoding result can be represented as a precoding matrix. The precoding matrix can be determined based on the downlink CSI of the first UE. For example, the precoding matrix can be calculated based on the downlink CSI using algorithms such as maximum ratio transmission, zero-forcing precoding, or minimum mean square error (MMSE) precoding. MMSE precoding can also be called regularized zero-forcing precoding. Unlike zero-forcing precoding, which simply aims to completely eliminate all interference at the receiver, the MMSE precoding algorithm minimizes the mean square error between the transmitted signal vector and the estimated value of the received signal vector after processing. Transmitting signals using the precoding matrix determined by the MMSE precoding algorithm achieves an optimal balance between enhancing the useful signal and suppressing errors introduced by channel distortion, inter-user interference, and noise amplification.

[0175] like Figure 6 As shown, UEs 121 to 123 are located at different positions relative to network device 110. Network device 110 transmits reference signals through multiple ports. Figure 6 The multiple beam directions shown correspond to ports 60 to 67, respectively. The L ports with the largest amplitudes determined by UEs 121 to UE123 are different. For example, UE 121 determines the two ports with the largest amplitudes to be ports 62 and 63, UE 122 determines the two ports with the largest amplitudes to be ports 64 and 65, and UE 123 determines the two ports with the largest amplitudes to be ports 60 and 61. Therefore, in order for the network device to serve UEs located in different positions, the network device needs to transmit the reference signal through more ports, i.e., to enable the UE to measure the channel response of more ports.

[0176] Figure 7 This shows the variation in the accuracy of the downlink CSI determined by the network device based on the PMI fed back by the UE. Figure 7 In the coordinate system shown, the horizontal axis represents the quantized bit width, and the vertical axis represents the normalized mean square error (NMSE). As the quantization bit width increases from 16 bits to 32 bits, the amount of data in the PMI increases, but the decrease in NMSE is limited. In other words, a higher quantization bit width has little effect on improving the accuracy of the downlink CSI determined by the network device for the first UE. However, as the number of ports transmitting reference signals increases from 4 to 32, the NMSE decreases significantly. That is, increasing the number of ports transmitting reference signals improves the accuracy of the downlink CSI determined by the network device. However, transmitting reference signals through more ports consumes more communication resources.

[0177] Furthermore, typically, the beams transmitted from multiple ports sending reference signals can be evenly distributed in space. However, the beams from the UE are not evenly distributed. For example... Figure 1 As shown, UE 121 and UE 122 are located in region D1, UE 123 is located in region D3, and there may be no electronic devices in regions D2 and D4.

[0178] The reference signal is transmitted through a port with a uniformly distributed beam in space, directed towards a port where there is no UE area, resulting in a very low feedback frequency. In other words, transmitting the reference signal through this port consumes significant power and has little impact on improving the accuracy of downlink CSI. This represents a waste of reference signal transmission resources.

[0179] In view of this, this application provides a communication method that can improve the accuracy of downlink CSI determined by network devices while consuming less communication resources.

[0180] The following detailed explanation of the solution provided in this application, in conjunction with the corresponding flowcharts, illustrates the method in detail. It is understood that the illustrative flowcharts provided in this application primarily use different devices (e.g., UE, network devices) as examples of the execution entities for this interactive illustration, but this application does not limit the execution entities of the interactive illustrations. For example, the devices (e.g., UE, network devices) in the illustrative flowcharts can also be chips, chip systems, or processors that support the implementation of this method on the device, or logical modules or software capable of implementing all or part of the device's functions.

[0181] As a general statement, the message or signaling interactions involved in the interaction process of this application embodiment can be standard messages or signaling or newly introduced messages or signaling. This application embodiment does not make specific limitations on this.

[0182] Figure 8 This is a schematic flowchart of a communication method provided in an embodiment of this application. Figure 8 The method shown can be applied to communication systems, specifically during channel probing. The communication system includes network equipment and a first UE. The network equipment can provide services to multiple UEs within its coverage area, and the first UE can be any one of these multiple UEs. It is understood that... Figure 8 The first UE in the middle can be Figure 1 The term "UE" can refer to any UE, or it can refer to a device within the UE (such as a processor, chip, or chip system). Network devices can be... Figure 1 The term "network equipment 110" can also refer to devices within network equipment (such as processors, chips, or chip systems). Figure 8 The method shown includes steps S810 to S850.

[0183] In step S810, the network device sends a first reference signal, which is associated with M ports of the network device.

[0184] Accordingly, UEs within the coverage area of ​​the network equipment can receive the first reference signal.

[0185] For example, a network device can send a first reference signal through each of M ports, where M is an integer greater than 1. The port associated with the reference signal can be understood as the port corresponding to the reference signal, the port that sends the reference signal.

[0186] The reference signal can be CSI-RS, and all M ports can be CSI-RS ports.

[0187] The first reference signals transmitted from any two ports can be orthogonal. During channel detection, different ports in the network device can correspond to different beam directions.

[0188] A network device may include N ports. The M ports that transmit the first reference signal may be some of these N ports. That is, N is an integer greater than M.

[0189] In step S820, the first UE sends a first CSI report. The first CSI report is determined by the first UE based on the first reference signal. The first CSI report is associated with N ports in the network device, where N is greater than M, and both M and N are integers greater than 1.

[0190] Correspondingly, the network device can receive the first CSI report sent by the first UE.

[0191] All UEs within the coverage area of ​​the network device can send a CSI report based on the first reference signal. The first UE can be any UE within the coverage area of ​​the network device.

[0192] The N ports of a network device may include all or some of the M ports that transmit the first reference signal. The following explanation uses an example where the N ports of a network device include the M ports that transmit the first reference signal.

[0193] The first CSI report represents the downlink CSI of the first UE. The first CSI report associates N ports in the network device; it can be understood that the first CSI report is used by the network device to determine the channel information of those N ports. The channel information of a port can be its amplitude parameters or complex channel gain parameters, etc.

[0194] The first UE can measure the first reference signal transmitted by each port to obtain the first channel response information. The first channel response information includes the complex channel gain parameters of K ports out of M ports, where K is a positive integer.

[0195] The complex channel gain parameter for each port represents the complex channel gain of that port. The complex channel gain parameter for each port can include an amplitude parameter and a phase parameter. The amplitude parameter represents amplitude attenuation, and the phase parameter represents phase offset.

[0196] K can be less than or equal to M. For example, the K ports can be the K ports with the smallest amplitude attenuation among the M ports that transmit the first reference signal.

[0197] Feature extraction from a set of channel information, including first-order channel response information, yields channel response features. This feature extraction can be performed using a feature extraction model. By processing the channel information set through this model, the channel response features can be obtained.

[0198] When K equals M, the first channel response information can more accurately reflect the channel state of the first UE. Therefore, the channel response features obtained by feature extraction based on the channel information set are more accurate.

[0199] The signal transmitted through the port exhibits power leakage, meaning that the signal energy radiates beyond the beam direction corresponding to that port. The first channel response information includes the complex channel gain parameters of more ports. In the subsequent process of determining the CSI of the first UE based on the channel information set including the first channel response information, the communication system can determine the CSI of the first UE based on the power leakage of more ports, making the determined CSI of the first UE more accurate.

[0200] Network devices can transmit reference signals periodically or aperiodically. For example, a network device can transmit reference signals at multiple times. A first UE can measure the reference signals transmitted by the network device at each time to obtain channel response information for that time. The first channel response information can be understood as the channel response information obtained by the first UE measuring the reference signals transmitted by the network device at the current time.

[0201] The amount of data for channel response features can be less than or equal to the amount of data in the channel information set. In other words, by processing the channel information set through the feature extraction model, data compression of the channel information set can be achieved, thereby reducing the communication resource consumption of the first UE's feedback of channel information.

[0202] For example, the data volume of the channel response features can be smaller than the data volume of the first channel response information. Therefore, compared to the method where the first UE sends the first channel response information to the network device, the network device extracts features from the channel information set including the first channel response information, and processes the extracted features to obtain the downlink CSI of the first UE, sending the channel response features to the network device by the first UE can reduce the communication resource consumption of the first UE's feedback of channel information.

[0203] The first prediction model can be used to process the channel response characteristics to obtain the downlink CSI of the first UE. The downlink CSI of the first UE includes the complex channel gain parameters of N ports in the network device, where N is greater than M, and the N ports include the M ports.

[0204] The downlink CSI of the first UE determined by the first prediction model can include the complex channel gain parameters of N ports in the network device, realizing channel extrapolation, that is, predicting the complex channel gain parameters of each of the N ports based on the complex channel gain parameters of some ports among the N ports.

[0205] In some embodiments, the first UE may send a first CSI report to the network device, which may include channel response features. That is, the first UE may extract features from the channel information set to obtain channel response features. The network device may process the channel response features using a first prediction model to obtain the downlink CSI of the first UE.

[0206] The network device transmits a first reference signal through each of the M ports, allowing the UE to determine the complex channel gain parameters of the K ports and obtain channel response characteristics. The network device processes the received channel response characteristics from the first UE using a first prediction model to obtain the first UE's downlink CSI. The first UE's downlink CSI includes the complex channel gain parameters of the N ports. N is greater than M, meaning the network device transmits the reference signal through fewer ports, resulting in lower power consumption. Furthermore, compared to the number M ports used to transmit the first reference signal, the network device, through the first prediction model, determines the first UE's CSI by including more ports corresponding to the complex channel gain parameters, making the first UE's downlink CSI more accurate. The network device can then communicate with the first UE based on the first UE's downlink CSI, enabling the first UE to receive the information transmitted by the network device more accurately.

[0207] In other embodiments, the first UE may send a first CSI report to the network device, which may include first channel response information. Therefore, the network device may perform feature extraction on the channel information set including the first channel response information, and process the extracted channel response features to obtain the downlink CSI of the first UE.

[0208] In some other embodiments, the first UE may send a first CSI report to the network device, which may include the first UE's downlink CSI. That is, the first UE may perform feature extraction on the channel information set and process the channel response features obtained from the feature extraction to obtain the first UE's downlink CSI.

[0209] After obtaining the downlink CSI of the first UE, the network device can perform precoding based on the downlink CSI of the first UE, and send information to the first UE based on the precoding matrix obtained by precoding.

[0210] In some other embodiments, the first UE may send a first CSI report to the network device, which may include a first PMI. That is, the first UE may perform feature extraction on the channel information set, process the extracted channel response features to obtain the first UE's downlink CSI, and precode based on the first UE's downlink CSI to obtain the first PMI. The network device may then send information to the first UE based on the first PMI.

[0211] The network device can transmit reference signals at multiple times. The reference signal transmitted at time t is associated with a first port group, and the reference signal transmitted at time t-1 is associated with a second port group, where t is an integer greater than 1. The ports in the first port group are different from the ports in the second port group. The M ports associated with the first reference signal are the ports in the first port group. The first UE can determine the first CSI report based on the reference signals from these multiple times.

[0212] At multiple times, network devices transmit reference signals through port groups including different ports. The first UE determines the first CSI report based on the reference signals at multiple times, thus the first CSI report can contain channel information from more ports, making the first CSI report more accurate.

[0213] The port groups associated with reference signals transmitted by a network device at multiple times can be selected cyclically from Q port groups, where Q is an integer greater than 1. The Q port groups can include a first port group and a second port group.

[0214] Network devices use a cyclic selection of port groups to transmit reference signals, making it more flexible to determine the port associated with the reference signal at each time.

[0215] The network device can send indication information to the first UE. The indication information can instruct the UE to report a CSI report based on reference signals from the plurality of times. For example, the indication information can instruct the UE to send a CSI report for each time period based on the reference signals already received from the plurality of times. Alternatively, the indication information can instruct the UE to send a CSI report based on the most recently received reference signal before receiving reference signals from Q times among the plurality of times, and instruct the UE to send a CSI report based on the already received reference signals after receiving reference signals from Q times among the plurality of times.

[0216] The indication information can also indicate to the UE the port groups associated with the reference signals transmitted by the network devices at the multiple times. The indication information can also indicate to the UE the reference signal resources of the reference signals transmitted by the network devices at the multiple times.

[0217] Network devices can transmit reference signals periodically or non-periodically. Reference signals transmitted by a network device at different times can be understood as reference signals transmitted by the network device in different measurement periods. A first UE can measure the reference signals transmitted by the network device in each measurement period to obtain channel response information for each measurement period. The first channel response information is the channel response information for the t-th measurement period, where t is an integer greater than 1. The t-th measurement period can be understood as the current measurement period. The first UE can determine the channel information set based on the channel response information of one measurement period. Alternatively, the first UE can determine the channel information set based on the channel response information of multiple measurement periods. The channel response information for any one measurement period includes the complex channel gain parameters of all or part of the ports through which the network device transmits the reference signal in that measurement period.

[0218] In each measurement cycle, the power of the reference signal transmitted by the network device through each port can be equal. That is, the amplitude of the reference signal transmitted by the network device through each port can be the same. When the channel response information for a measurement cycle includes the complex channel gain parameters of some ports through which the network device transmits the reference signal in that measurement cycle, these ports can be the ones with the largest amplitude, i.e., the ones with the smallest amplitude attenuation.

[0219] Based on channel response information from multiple measurement periods, a channel information set is determined, thereby making the CSI of the first UE determined by the network device based on the channel information set more accurate.

[0220] The channel response information used to determine the channel information set over multiple measurement periods includes the complex channel gain parameter of the first port (excluding the M ports) among the N ports of the network device. In other words, over multiple measurement periods, the network device transmits reference signals through ports other than the M ports.

[0221] In other words, the first port is the port where the network device sends the reference signal in the t-1 measurement period before the t-th measurement period, and the first port does not belong to the M ports where the network device sends the first reference signal in the t-th measurement period.

[0222] In the (t-1) periods preceding the current measurement period (i.e., the t-th measurement period), the network device transmitted a reference signal through a first port in addition to the M ports. The first channel response information set also includes the complex channel gain parameter of the first port. The channel information set includes the complex channel gains of more ports. The channel information set is determined based on the channel response information from multiple measurement periods, taking into account the influence of the complex channel gain parameter of the first port determined in other measurement periods. Therefore, the downlink CSI of the first UE determined by the network device based on the channel response characteristics of the first channel response information in the channel information set is more accurate.

[0223] The channel information set may include channel response information from multiple measurement periods. Alternatively, the channel information set may include the channel response information from the t-th measurement period, i.e., the first channel response information, and the complex channel gain parameters of the first port.

[0224] For the M ports that transmit the first reference signal in the t-th measurement period, the latest measurement result in the channel information set is used as the complex channel gain parameter for each port, which has higher real-time performance. As a result, the downlink CSI of the first UE determined by the network device based on the channel response characteristics of the first channel response information in the channel information set is more accurate.

[0225] In the sub-6 GHz frequency band, the channel exhibits significant multipath effects, and accurate channel measurements require multi-port measurements. Furthermore, the important ports differ depending on the location. Therefore, network devices need to transmit reference signals through a relatively large number of ports. To reduce the number of ports used by the network device to transmit reference signals in each measurement cycle, the ports used for transmitting reference signals in different measurement cycles can be different.

[0226] In the channel information set, the complex channel gain parameter of the first port can be the measurement result of the first UE on the first port when the network device last sent the reference signal through the first port in the t-1 measurement period before the t-th measurement period, that is, the last measurement result of the first UE on the first port.

[0227] The first UE may move, causing its location to change and resulting in channel outdated information. In other words, the complex channel gain of a port obtained by measuring the reference signal transmitted by the first UE to a certain port before the t-th measurement period cannot accurately represent the channel characteristics of the UE in the t-th measurement period.

[0228] The first UE can determine the channel response prediction information for the t-th measurement period using the second prediction model, based on the channel response information for the (t-1)-th measurement period. The channel response prediction information for the t-th measurement period may include the complex channel gain parameters of P ports out of N ports of the network device. These P ports include the ports corresponding to the complex channel gain parameters in the channel response information of the (t-1)-th measurement period. In other words, these P ports may include all or some of the ports through which the network device transmitted the reference signal during the (t-1)-th measurement period.

[0229] When the P ports include the first port, the channel information set may include the complex channel gain parameter of the first port. The first port belongs to the P ports, but does not belong to the M ports through which the network device transmits the first reference signal in the t-th measurement period.

[0230] In other words, after the first UE determines the channel response prediction information for the t-th measurement period through the second prediction model, it can compare the P ports in the channel response prediction information for the t-th measurement period with the ports in the channel response information for the t-th measurement period, and add the complex channel gain parameters of the ports that are different from those in the channel response information for the t-th measurement period to the channel information set.

[0231] Based on the channel response information of the (t-1)th measurement period, the predicted channel response information for the tth measurement period is determined. Feature extraction is performed on the channel information set to obtain the channel response features. The channel information set includes the first channel response information and the complex channel gain parameters of the other ports among the P ports that do not belong to the current measurement period and transmit the first reference signal.

[0232] Based on the predicted results of the complex channel gain parameters of other ports in the channel response information obtained from the measurement of the reference signal in the previous measurement period, and the measurement results of the complex channel gain parameters of multiple ports in the current measurement period, the channel information set is determined.

[0233] The channel information set includes the complex channel gain parameters of more ports, which makes the CSI of the first UE determined by the network device based on the channel information set more accurate. At the same time, the number of ports that the network device sends reference signals in a measurement cycle is less, reducing the occupation of communication resources by the transmission of reference signals.

[0234] Sending reference signals through numerous ports consumes significant communication resources. Furthermore, the location of the first UE may change during different measurement periods, causing the channel response information from the previous measurement period to fail to accurately reflect the current channel response status of the first UE.

[0235] The channel response information for the (t-1)th measurement period is processed by the second prediction model to obtain the channel response prediction information for the tth measurement period. This prediction information accurately reflects the channel responses of P ports within the tth measurement period. In other words, the channel information set determined based on the channel response prediction information for the tth measurement period more accurately reflects the channel response of the first UE. Consequently, the CSI of the first UE subsequently determined by the network device based on this channel information set has higher accuracy.

[0236] To further improve the accuracy of the channel response prediction information in the t-th measurement period, the channel response prediction information in the (t-1)-th measurement period and the channel response information in the (t-1)-th measurement period can be processed by the second prediction model to obtain the channel response prediction information in the t-th measurement period, where t is greater than 2.

[0237] For the P ports in the channel response prediction information of the t-th measurement period, they can include the ports in the channel response information of the (t-1)-th measurement period, and also the ports in the channel response prediction information of the (t-1)-th measurement period. That is, these P ports can be understood as the union of the ports corresponding to the complex channel gain parameters in the channel response information of the (t-1)-th measurement period and the ports corresponding to the complex channel gain parameters in the channel response prediction information of the (t-1)-th measurement period.

[0238] The second prediction model can be a gated recurrent unit (GRU). This model processes the channel response prediction information from measurement period (t-1) to obtain the channel response prediction information for measurement period t. The channel response prediction information from measurement period (t-1) and the channel response information from measurement period (t-1) are then input into the GRU. The GRU determines the weight of each complex channel gain parameter in the channel response prediction information from measurement period (t-1), as well as the weight of each complex channel gain parameter in the channel response information from measurement period (t-1). The second prediction model can then perform a weighted sum of the complex channel gain parameters corresponding to the same port in the channel response prediction information from measurement period (t-1) and the channel response information from measurement period (t-1) to obtain the complex channel gain parameter for that port in the channel response prediction information from measurement period (t). For complex channel gain parameters that do not overlap with the corresponding port in the channel response prediction information of the (t-1)th measurement period, the complex channel gain parameter can be multiplied by the corresponding weight to obtain the complex channel gain parameter of that port in the channel response prediction information of the tth measurement period.

[0239] The channel response prediction information for the (t-1)th measurement period is determined by the second prediction model based on the channel response information for the (t-2)th measurement period, where t is greater than 2. When t is 3, the channel response prediction information for the (t-1)th measurement period can be obtained by the first UE processing the channel response information for the (t-2)th measurement period using the second prediction model. When t is greater than 3, the channel response prediction information for the (t-1)th measurement period is obtained by processing both the channel response information for the (t-2)th measurement period and the channel response prediction information for the (t-2)th measurement period using the second prediction model. The channel response prediction information for the (t-2)th measurement period can be determined based on the channel response information for the (t-3)th measurement period.

[0240] The channel response information and channel response prediction information for the (t-1)th measurement period are processed by the second prediction model to obtain the channel response prediction information for the t-th measurement period. The ports corresponding to the complex channel gain parameters in the channel response prediction information for the t-th measurement period include both the ports corresponding to the complex channel gain parameters in the channel response information for the (t-1)th measurement period and the ports corresponding to the complex channel gain parameters in the channel response prediction information for the (t-1)th measurement period. Therefore, the channel response prediction information for the t-th measurement period can more accurately reflect the complex channel gain parameters of the ports transmitting reference signals in multiple measurement periods, resulting in higher accuracy of the CSI of the first UE determined based on the channel response prediction information for the t-th measurement period.

[0241] In the case where the target ports in the channel information set can include K ports in the channel response information of the t-th measurement period, as well as the first port, that is, when the channel information set does not include the complex channel gain parameters of the ports that overlap with the channel response information of the t-th measurement period in the channel response prediction information, but includes the complex channel gain parameters of other ports in the channel response prediction information of the t-th measurement period excluding the ports that overlap with the channel response information of the t-th measurement period, the complex channel gain parameters in the channel information set can be arranged according to the beam direction corresponding to each port, or in order of amplitude, or in order of the sequence size of the port identifier.

[0242] Before the first UE extracts features from the channel information set using the feature extraction model, it can sort the complex channel gain parameters of multiple target ports in the channel information set according to their amplitude. For example, the first UE can use a sorting layer to sort the complex channel gain parameters of multiple target ports in the channel information set according to their amplitude.

[0243] For UEs in different locations, the port with the least amplitude attenuation can be different. Furthermore, the channel information set is sparsity-dependent; that is, complex channel gain parameters with higher amplitudes are concentrated in a few ports, while the complex channel gain parameters of most other ports have lower amplitudes. By sorting the complex channel gain parameters of multiple target ports in the channel information set according to their amplitude, the feature extraction model can extract features from relatively stable sparse data, resulting in more accurate extracted channel response features.

[0244] The first UE can also send port information to the network device. The port information represents the target port corresponding to each port channel feature in the channel response features. The port information can be carried in the same or different messages as the channel response features.

[0245] The complex channel gain parameters of the target ports exhibit sparsity, meaning that the amplitude attenuation of the reference signals transmitted from multiple target ports to the first UE varies significantly. Feature extraction from sparse data results in low accuracy and stability. By sorting the complex channel gains in the channel information set according to their amplitude and performing feature extraction on the sorted channel information set, the downlink CSI accuracy of the first UE determined by the channel response characteristics is significantly improved.

[0246] Network devices can extract features from the received channel response characteristics using a first prediction model.

[0247] Alternatively, the network device can determine the expanded channel response characteristics based on port information and channel response characteristics. Bingna then processes the expanded channel response characteristics using the first prediction model to obtain the CSI of the first UE.

[0248] The expanded channel response characteristics include the port characteristics of N ports of the network device. In the expanded channel response characteristics, the port characteristics of these N ports are arranged according to the beam direction corresponding to the port. The port characteristics of the target port are the port channel characteristics of the target port in the channel response characteristics, and the port characteristics of other ports are preset values.

[0249] Depending on the beam direction corresponding to the port, it can be rotated clockwise or counterclockwise according to the beam direction.

[0250] Network devices can fill the position corresponding to any target port in the initial features with the port channel characteristics from the channel response characteristics, based on the correspondence between ports and positions and port information, to obtain the expanded channel response characteristics. The multiple positions in the initial features are arranged according to the beam direction of the port corresponding to that position.

[0251] The initial feature can have N positions with initial values. The port channel feature of a target port in the channel response feature is filled into the position corresponding to that target port in the initial feature; alternatively, the initial value of the position corresponding to that target port in the initial feature can be modified to the port channel feature of that target port.

[0252] Alternatively, no preset values ​​may be set in the initial features. After writing the port channel features of each target port into the corresponding positions of the target ports in the channel response features, the positions where port channel features were not written can be set to preset values ​​to obtain the expanded channel response features. Preset values ​​can also be called initial values ​​or preset features. A preset value can be 0.

[0253] Therefore, the expanded channel response characteristics include the port characteristics of each target port, and the characteristics of other ports besides the target ports are preset values. The expanded channel response characteristics can also be called the sorted channel response characteristics.

[0254] The signal transmitted from a port exhibits power leakage, meaning that signal energy radiates beyond the intended main beam direction. The degree of power leakage in each direction is correlated with the directional difference between the beam direction corresponding to that port and the other directions. Before extracting features from the received channel response features, the network device can sort and expand the complex channel gain parameters in the channel response features according to the beam direction corresponding to the port. This ensures that the expanded channel response features include the port features of each of the network device's N ports, and the port features of the N ports are arranged according to their corresponding beam directions. Therefore, the downlink CSI of the first UE obtained by processing the expanded channel response features has higher accuracy.

[0255] For example, a network device can sort and expand multiple port channel features in the channel response features according to the beam direction corresponding to the target port represented by the port information by filling the neck network, so as to obtain the expanded channel response features.

[0256] For example, the network device may record the beam direction corresponding to each port. Therefore, the network device can also sort the channel characteristics of multiple ports in the channel response characteristics, as well as the initial values ​​of the other ports among the N ports besides those recorded in the channel response characteristics, according to the beam direction corresponding to the port, to obtain the expanded channel response characteristics.

[0257] Alternatively, network devices can distinguish individual ports using port identifiers. Port information can represent the port identifier of the target port corresponding to each port channel feature in the channel response characteristics. The sequence numbers of the port identifiers of N ports can be arranged according to the beam direction corresponding to the port. Thus, network devices can also sort multiple port channel features in the channel response characteristics, along with the initial values ​​of other ports, according to the port identifier sequence numbers.

[0258] Before step S810, the network device determines the M ports from the included N ports.

[0259] For example, in each measurement cycle, the port from which the network device transmits the reference signal can be randomly determined by the network device. Or, for example, before proceeding to step S810, the network device can determine Q port groups and a first order for those Q port groups. The network device can then transmit the reference signal by cyclically selecting port groups from the Q port groups according to the first order. In other words, the port groups from which the reference signal is transmitted in each measurement cycle are cyclically selected by the network device from the Q port groups according to the first order.

[0260] The Q port groups include a first port group and a second port group. The ports in the first port group are different from those in the second port group. For example, all ports in each of the Q port groups are different. Alternatively, some ports in some of the Q port groups may be the same. The first port group includes M ports. That is, the ports in the first port group can be the M ports that transmit the first reference signal.

[0261] The method of determining the port to send the reference signal in each measurement cycle by cyclically selecting port groups from the Q port groups in the first order makes the processing simpler.

[0262] The first order and the Q port groups can be preset. Alternatively, the first order and the Q port groups can be determined by the network device.

[0263] The Q port group may include multiple candidate ports. The number of these candidate ports is less than N. The Q port group may or may not include other ports besides these candidate ports. These candidate ports may be preset or determined by the network device.

[0264] Network devices can divide multiple candidate ports into Q port groups. For details on how network devices divide multiple candidate ports, please refer to [link to relevant documentation]. Figure 11 Explanation.

[0265] In the multiple time intervals during which the network device transmits reference signals, the length of Q time intervals is less than a preset duration. The port groups associated with the reference signals transmitted by the network device in the Q time intervals include Q port groups.

[0266] The first UE may move, causing a change in its location and resulting in channel aging. The network device can cyclically select port groups from Q port groups to transmit reference signals. Transmitting reference signals associated with the Q port groups within a shorter preset time reduces the likelihood of channel aging and improves the accuracy of the first CSI report.

[0267] For example, the Q ports may include multiple candidate ports, with the candidate ports having a higher importance than other ports among the N ports of the network device. Reference signals transmitted by the network device at multiple times are associated with the candidate ports with higher importance, thereby making the first UE more accurate in determining the first CSI report to be transmitted.

[0268] A network device can determine a second order of importance for N ports based on CSI reports from multiple user devices. The network device can then identify multiple candidate ports from these N ports based on this second order. In this second order, each candidate port is considered more important than any other port among the N ports.

[0269] The multiple UEs may or may not include the first UE. In the CSI reports of the multiple UEs, the number of CSI reports for each UE may be one or more.

[0270] Each UE's CSI report can represent the UE's downlink CSI. The UE's downlink channel state information includes the complex channel gain parameters of multiple ports out of N ports of the network device. For example, the network device can determine the downlink CSI of multiple UEs based on the CSI reports of multiple UEs, and determine a second order of the N ports of the network device according to their importance based on the downlink CSI of multiple UEs.

[0271] When the number of ports transmitting reference signals in a network device is limited, the first UE measures the reference signals transmitted by the candidate ports with higher importance in the network device. This allows the measurement results of the first UE to more accurately reflect the channel state between the network device and the base station. Based on the measurement results of the first UE, the downlink CSI of the first UE is determined, making the downlink CSI of the first UE more accurate.

[0272] In some embodiments, during the process of determining a second order of importance for N ports, where each downlink CSI in the plurality of UEs includes amplitude parameters of the N ports of the network device, the N ports can be sorted according to the magnitude of the amplitude parameters based on each downlink CSI to obtain the port sorting result corresponding to that downlink CSI. For example, based on each downlink CSI, the N ports can be sorted in ascending order of amplitude attenuation represented by the amplitude parameters. The index of each of the N ports of the network device in the port sorting result corresponding to each downlink CSI can be the average index of that port. Sorting the N ports in ascending order of the average index can be understood as sorting the N ports in descending order of importance.

[0273] In other embodiments, by processing the downlink CSIs of multiple UEs using a port model, the importance of each of the N ports can also be obtained. When processing using a port model, each downlink CSI of the multiple UEs can include the complex channel gain parameters of all or some of the N ports of the network device.

[0274] The port model can be a neural network model. The structure of the port model can be found in [reference needed]. Figure 12 The explanation is as follows. The port model can be achieved through... Figure 13 The training method shown is used to train the equipment.

[0275] Network devices select high-importance ports as candidate ports and transmit reference signals through these candidate ports during periodic transmission. Therefore, the channel response characteristics are determined based on the complex channel gain parameters of the high-importance ports, resulting in a more accurate downlink CSI for the first UE determined from these channel response characteristics.

[0276] By dividing multiple candidate ports with high importance into Q port groups, and cyclically selecting the port group that sends the reference signal in each measurement cycle in the Q port groups in a first order, the first UE within the coverage area of ​​the network device can perform measurement on each candidate port in Q cycles.

[0277] The importance of a port is determined based on the downlink CSI of multiple UEs. Each UE's downlink CSI is influenced by its location and surrounding environment. Therefore, the importance of a port reflects the demand of the network device's environment on each port. Consequently, when the number of measurement periods t is greater than or equal to Q, the network device transmits reference signals through candidate ports with higher importance. The transmission of reference signals by the network device is adapted to its environment, thus making the determined downlink CSI of the first UE more accurate.

[0278] The network device can instruct the M ports within its coverage area to send the first reference signal during the t measurement period.

[0279] Before transmitting a reference signal through a port in the port group during each measurement cycle, the network device sends a first measurement indication message. This first measurement indication message indicates which port in the port group will transmit the reference signal during the current measurement cycle. Thus, UEs within the network device's coverage area can measure the reference signal during the current measurement cycle based on the first measurement indication message, determining the complex channel gain parameters of each port transmitting the reference signal.

[0280] The network device can send a second measurement indication message, which can indicate the Q port groups and the first sequence. Thus, the UE within the network device's coverage area can, based on the second measurement indication message, determine the ports in the port group that transmit the reference signal in each measurement cycle, measure the reference signal in each measurement cycle, and determine the complex channel gain parameters of each port transmitting the reference signal in that measurement cycle.

[0281] If a network device determines the importance of each of the N ports using a port model, it can also divide the multiple candidate ports into Q port groups based on the importance of each first candidate port.

[0282] If the differences in importance between candidate ports are small, repeatedly setting ports with higher importance may reduce the frequency of reference signal transmission on other ports, potentially resulting in lower accuracy of the downlink CSI for the first UE determined by the network device.

[0283] After determining the importance of each port, the network device can determine whether a second and third port exist among multiple candidate ports. The first importance of the second port is greater than the second importance of the third port, and the difference between the first and second importance is greater than or equal to a preset difference threshold.

[0284] If neither a second nor a third port exists among multiple candidate ports, each candidate port can be set in one port group. If both a second and a third port exist among multiple candidate ports, the number of port groups including the second port is greater than the number of port groups including the third port. In other words, the second port can be set in multiple port groups out of Q port groups.

[0285] In the Q port groups, the number of ports in each port group can be the same or different. The number of ports in each port group can be preset. Alternatively, the network device can also configure the number of ports in each port group. In addition to configuring the number of ports in each port group, the network device can also configure the number of port groups Q.

[0286] In some cases, among the N ports of a network device, one or more ports are of far greater importance than the others. For these ports, the network device transmits reference signals at a higher frequency, which can improve the accuracy of the downlink CSI of the determined first UE.

[0287] Under preset conditions, the network device can determine the number of ports in each port group as the number of ports in the first cycle. If the preset conditions are not met, the network device can determine the number of ports in each port group as the number of ports in the second cycle. The preset conditions are: the data transmission volume of the network device is less than a preset transmission volume threshold, and / or, the movement speed of the UE within the coverage area of ​​the network device is greater than or equal to a preset speed threshold. The number of ports in the first cycle is greater than the number of ports in the second cycle.

[0288] The movement speed of user equipment within the network equipment's coverage area can be the maximum, minimum, or central tendency of the movement speed of each UE. The central tendency of the data can be understood as a representative value of the data, such as the mean, median, or mode.

[0289] In a measurement cycle, sending reference signals through more ports can improve the accuracy of the downlink CSI of the UE obtained in that cycle. When the network device's data transmission volume is less than a preset transmission volume threshold, its communication resources are relatively abundant, and measuring more ports in a single cycle will not affect data transmission. However, when the network device's data transmission volume is high, exceeding or equal to the preset transmission volume threshold, sending reference signals through more ports will consume more communication resources and impact data transmission.

[0290] The CSI estimate is valid during the channel coherence time when the channel characteristics remain essentially unchanged. The downlink CSI of the UE determined by the network device can be understood as the CSI estimate. Rapid UE movement causes rapid changes in the length and angle of the wireless transmission path. This leads to an increase in the Doppler frequency shift of the received signal and rapid changes in the channel impulse response, resulting in a shorter channel coherence time.

[0291] When the UE's movement speed within the network device's coverage area is high, the channel coherence time is short. The network device needs to use more ports to transmit reference signals in each measurement cycle to ensure accurate downlink CSI by transmitting reference signals through various candidate ports within a short time. Conversely, when the UE's movement speed within the network device's coverage area is low, the channel coherence time is long. The network device needs to use fewer ports to transmit reference signals in a measurement cycle to achieve a more accurate downlink CSI. Using fewer ports to transmit reference signals in a measurement cycle also reduces the consumption of communication resources.

[0292] Based on preset conditions, the network device can adjust the number of ports in the port group, thus providing greater flexibility in setting the number of ports that send reference signals, and making the number of ports that send reference signals in each measurement cycle more adaptable to the current situation.

[0293] The first UE can also quantize the channel response features and send the quantized channel response features to the network device. The quantized channel response features can include the quantized port channel features of each of the multiple target ports. By quantizing the channel response features, the first UE can further reduce the amount of data sent to the network device for port channel status feedback.

[0294] The network device can expand the quantized port channel features in the quantized channel response features to obtain expanded channel response features, and process the expanded channel response features to obtain the downlink CSI of the first UE. In other words, the network device does not need to perform inverse quantization processing.

[0295] The communication method provided in this application involves a network device sending a first reference signal associated with M ports. The network device also receives a first CSI report, determined based on the first reference signal, sent by a first UE. This first CSI report is associated with N ports in the network device, where N > M. Therefore, the first CSI report includes channel information from more ports, more accurately reflecting the downlink channel situation between the first UE and the network device. The network device then sends information to the first UE based on the first CSI report, enabling the first UE to receive the information more accurately and efficiently.

[0296] It should be understood that network devices can determine candidate ports periodically or non-periodically. Network devices can update candidate ports according to a candidate port update cycle and then re-determine port groups based on the updated candidate ports. For example... Figure 9 As shown, each candidate port update period can include T measurement periods. T can be a positive integer greater than 1.

[0297] Different candidate port update periods correspond to different candidate ports. These different ports correspond to different beam directions. In other words, different candidate port update periods can be understood as sending reference signals in different directions. The measurement period can be determined based on the Doppler frequency shift. The length of the candidate port update period can be much longer than the length of the measurement period.

[0298] For each candidate port update cycle, the communication system can perform... Figure 10 The communication method shown.

[0299] Figure 10 This is a schematic flowchart of a communication method provided in an embodiment of this application. Figure 10 The method shown includes steps S1001 to S1010.

[0300] In step S1001, the network device determines multiple candidate ports based on the downlink CSIs of multiple UEs determined in the i-1 candidate port update period.

[0301] After determining the downlink CSI of each UE within the coverage area of ​​the Tth measurement period in the (i-1)th candidate port update period, the network device can determine multiple candidate ports for the i-th candidate port update period based on the downlink CSI of the UE determined in the Tth measurement period of the (i-1)th candidate port update period.

[0302] In some cases, network devices may need a certain amount of time to determine multiple candidate ports. Alternatively, network devices may determine multiple candidate ports for the i-th candidate port update period based on the downlink CSI of the UE already determined in the i-1 candidate port update period, within a first preset time period before the end of the i-1 candidate port update period. The first preset time period may be an integer multiple of the measurement period length.

[0303] Network devices can use a port model to process the downlink CSI of multiple UEs identified in the (i-1)th candidate port update cycle to determine the importance of N ports in the network device. The network device can then identify the multiple candidate ports with the highest importance from these N ports. The number of candidate ports is a preset number.

[0304] like Figure 11 As shown in (a), multiple candidate ports may include port 0 to port 5.

[0305] In step S1002, the network device divides multiple candidate ports into Q port groups, where Q is a positive integer greater than 1.

[0306] In the Q port groups, different port groups may or may not include the same candidate ports.

[0307] If no two ports are the same across the Q port groups, meaning each candidate port is located in one of the Q port groups, the network device can randomly assign multiple candidate ports to any of the Q port groups. Alternatively, the network device can divide the multiple candidate ports according to their importance.

[0308] In some embodiments, the importance of ports in the Q port groups can be uniform. That is, dividing multiple candidate ports can balance the importance of each port group as much as possible.

[0309] For example, in descending order of importance, the first Q ports from multiple candidate ports can be set in the Q port group, the (Q+1)th to the 2Qth ports from multiple candidate ports can be set in the Q port group, and so on, until all multiple candidate ports are set in the Q port group.

[0310] like Figure 11 As shown in (b), ports 0 through 5 are arranged in descending order of importance. The Q port groups can include port group q1 and port group q2. Port group q1 can include port 0, port 2, and port 4, while port group q2 can include port 1, port 3, and port 5.

[0311] When the differences in importance among candidate ports are relatively balanced—for example, when the difference in importance between any two adjacent candidate ports is less than or equal to a preset difference—the partitioning of candidate ports ensures that the importance of candidate ports within each port group is balanced. This relatively balanced difference in the importance of candidate ports can be understood as a weak sparsity among the candidate ports of the network device.

[0312] When the candidate ports of a network device are highly sparsity-oriented, meaning there is a significant difference in importance between the candidate ports—for example, some candidate ports are far more important than others—these ports can be assigned to a larger number of port groups. In other words, the number of port groups that include any one of these ports can be greater than or equal to the number of port groups that include any other port.

[0313] For example, in some embodiments, ports of higher importance can be set in some port groups, while ports of lower importance can be set in other port groups. Repeating the setting of ports of higher importance in some port groups results in Q port groups. That is, in some of the Q port groups, the ports in multiple port groups are the same. The ports in these port groups can have a higher importance than the ports in other port groups.

[0314] like Figure 11 As shown in (c), the Q port groups can include port group q3, port group q4, and port group q5. Port group q3 and port group q4 are identical, each including the three most important ports from port 0 to port 5, namely port 0, port 1, and port 2. Port group q5 includes ports 3, port 4, and port 5, which are of lower importance from port 0 to port 5.

[0315] In some other embodiments, candidate ports with higher importance can be set in multiple port groups of the Q port groups, while each candidate port with lower importance can be set in one port group of the Q port groups.

[0316] Except for multiple candidate ports that are repeatedly set in multiple port groups, other candidate ports can be evenly distributed among the Q ports according to their importance. For example, in order of decreasing importance, the first Q candidate ports (excluding those repeatedly set in multiple port groups) can be set in Q port groups, the (Q+1)th to the 2Qth ports can be set in Q port groups, and so on, until all candidate ports are set in the Q port groups.

[0317] like Figure 11 As shown in (d), the Q port groups can include port group q6 and port group q7. Both port group q6 and port group q7 can include ports 0 and 1, which have higher importance. Port group q6 can also include ports 2 and 4, and port group q7 can also include ports 3 and 5.

[0318] In some cases, if the UE detects that the amplitude attenuation represented by the complex channel gain parameter for one or more ports is significantly less than that for other ports, the UE can send important port information to the network device. Important port information can indicate the specific port(s). Within Q port groups, the number of port groups including the port indicated by the important port information can be higher than the number of port groups including other candidate ports.

[0319] The number of ports in each port group can be determined by the network device based on data transmission volume and / or the movement speed of the UE within the coverage area. The number of port groups can be determined by the network device based on the number of ports in each port group.

[0320] For example, under the condition that a preset condition is met, the number of ports in each of the Q port groups can be the number of ports in the first cycle; conversely, under the condition that the preset condition is not met, the number of ports in each of the Q port groups can be the number of ports in the second cycle. The number of ports in the first cycle is greater than the number of ports in the second cycle. The preset condition can be: the data transmission volume of the network device is less than a preset transmission volume threshold, and / or, the movement speed of the user equipment within the coverage area of ​​the network device is greater than or equal to a preset speed threshold.

[0321] For example, when the data transmission volume of the network device is relatively small, each port group of Q port groups contains 3 ports, and the Q port groups can include, for example... Figure 11Port groups q1 and q2 are shown in (b) above. When the data transmission volume of the network device is large, each of the Q port groups contains 2 ports, such as... Figure 11 In (e), the Q port groups can include port group q8, port group q9, and port group q10. Port group q8 can include port 0 and port 5, port group q9 can include port 1 and port 4, and port group q10 can include port 2 and port 3. For example... Figure 11 The importance of the ports in the Q-port group shown in (e) is uniform.

[0322] Step S1003: The network device sends the reference signal resource for the j-th port group.

[0323] The reference signal resources can indicate the time-frequency resources and signal sequence of the reference signals transmitted by each port. When step S1003 is executed for the first time after step S1002, j is 1.

[0324] In step S1004, the network device sends a reference signal through the j-th port group according to the reference signal resources of the j-th port group.

[0325] Step S1005: Determine the channel information based on the reference signal resources of the j-th port group.

[0326] The first UE can measure the reference signal transmitted by each party in the j-th port group based on the reference signal resources of the j-th port group, and determine the channel information based on the measurement results. In each measurement period, the measurement result of the first UE can be the channel response information for that measurement period. The channel response information for each measurement period can include the complex channel gain parameters of each port through which the network device transmits the reference signal in that measurement period.

[0327] Channel information may include the complex channel gain of all or some ports in the j-th port group. Alternatively, channel information may indicate the port with the smallest amplitude attenuation in the j-th port group. Or, the first UE may determine the channel response characteristics based on measurement results. Channel information can be channel response characteristics.

[0328] Step S1006: The first UE sends channel information to the network device.

[0329] In step S1007, the network device determines the downlink CSI of the first UE based on the channel information.

[0330] Step S1008: The network device determines whether j is greater than or equal to the number of port groups Q.

[0331] When j < Q, the network device can proceed to step S1009. When j ≥ Q, the network device can proceed to step S1010.

[0332] In step S1009, the network device increments j by 1.

[0333] In other words, network devices can set j to j+1.

[0334] In step S1010, the network device sets j to 1.

[0335] After step S1009 or step S1010, the network device may perform steps S1003 and S1004. It should be understood that steps S1003 and S1004 may be performed periodically. Steps S1003 and S1004 may be performed once in each measurement cycle.

[0336] Steps S1008 to S1010 may be performed after step S1004. For example, the network device may perform steps S1008 to S1010 after step S1006 or step S1007.

[0337] For example, the network device may also record the number of measurement cycles elapsed after the start of the i-th candidate port update cycle. The number of measurement cycles elapsed can be understood as the number of times the reference signal is transmitted, i.e., the number of times step S1004 is performed. If the number of measurement cycles elapsed after the start of the i-th candidate port update cycle is greater than or equal to T, the network device may perform step S1001 again to determine multiple candidate ports in the (i+1)-th candidate port update cycle, and perform steps S1002 to S1010 again, thereby determining the downlink CSI of the first UE based on the multiple candidate ports in the (i+1)-th candidate port update cycle.

[0338] During step S1001, the network device can process the downlink CSI of multiple UEs using a port model to determine multiple candidate ports. The structure of the port model can be found in [reference needed]. Figure 12 Explanation.

[0339] Figure 12 This is a schematic structural diagram of a port model provided in an embodiment of this application.

[0340] Port model 1200 can be a neural network model. The port model may include a batch normalization layer 1210, multiple convolutional layers 1221 to 122n, a pooling layer 1230, and a fully connected layer 1240.

[0341] The batch normalization layer is used to normalize the data of each batch input port model, thereby solving the problem of data distribution offset.

[0342] The batch normalization layer 1210 is used to adjust the amplitude parameters in the downlink CSI of multiple UEs in the input port model 1200, so that the mean and variance of the adjusted downlink CSI are within a preset range. The data distribution of the adjusted downlink CSI is more stable, thus making the processing results of the adjusted downlink CSI more accurate.

[0343] Each convolutional layer from 1221 to 122n is used to perform convolution processing on the input data, thereby achieving feature extraction. When a neural network has multiple convolutional layers, the initial convolutional layers (such as convolutional layer 1221) tend to extract more general features, which can also be called low-level features. As the depth of the neural network increases, the features extracted by later convolutional layers (such as 122n) become more and more complex, such as high-level semantic features. Features with higher semantic levels are more suitable for the problem to be solved.

[0344] Because it is often necessary to reduce the number of training parameters, pooling layers are often introduced periodically after convolutional layers. For example... Figure 12 Each layer in the convolutional layers 1221 to 122n in the example of the mid-port model 1200 can be a convolutional layer followed by a pooling layer, or multiple convolutional layers followed by one or more pooling layers.

[0345] The purpose of pooling layer 1230 can be to reduce the spatial size of the data. Pooling layer 1230 may include average pooling operators and / or max pooling operators to sample the input data to obtain data of smaller size. The average pooling operator can calculate the average value of the element values ​​in the data within a certain range. The max pooling operator can take the element value with the largest value within a certain range as the result of max pooling.

[0346] Feature data can be obtained by processing the adjusted downlink CSIs through convolutional layers 1221 to 122n and pooling layer 1230.

[0347] The fully connected layer 1240 can be used to process feature data to obtain the importance of each of the N ports of the network device.

[0348] In other words, during the i-1 candidate port update period, even if the network device does not transmit a reference signal through a certain port, the importance of that port can be determined by the downlink CSI of multiple UEs determined by the network device in the i-1 candidate port update period through the port model.

[0349] Considering the temporal correlation of the UE's downlink CSI, a port with higher importance in the (i-1)th candidate port update period is also likely to have higher importance in the ith candidate port update period. Therefore, a port with higher importance in the (i-1)th candidate port update period can be used as the candidate port for the ith candidate port update period.

[0350] Port model 1200 can be a trained neural network model. Below, we will combine... Figure 13 The training method for the port model is explained.

[0351] Figure 13 This is a schematic flowchart of a training method provided in an embodiment of this application. Figure 13 The training method shown includes steps S1301 to S1303.

[0352] Step S1301: Obtain the second training data, which includes channel information samples and importance order labels.

[0353] The channel information sample includes the training downlink CSI of multiple training UEs. The training downlink CSI of any one of these training UEs includes the complex channel gain parameters of N1 of the N third training ports of the second training network device determined by that training UE.

[0354] The importance ranking label represents the order of importance of the N1 third training ports out of the N third training ports. N1 is less than or equal to N, and N1 is an integer greater than 1. The channel information sample includes the complex channel gain of the N1 third training ports.

[0355] The N third training ports of the second training network device are the same as the N ports of the network device.

[0356] In other words, the second training network device can have the same antenna array as the network device. Furthermore, the codebooks corresponding to the N third training ports in the second training network device are the same as those corresponding to the N ports in the network device. Therefore, the beam directions corresponding to the N third training ports in the second training network device are the same as those corresponding to the N ports in the network device.

[0357] The second training network device can transmit reference signals through all or some of the N third training ports. Training UEs within the coverage area of ​​the second training network device can determine the complex channel gain parameter of each third training port transmitting the reference signal in the second training network through measurement. The training downlink CSI of each training UE can be determined based on the measurement results of that training UE.

[0358] For example, the training downlink CSI for each training UE may include all or part of the complex channel gain parameters of the third training port that transmits the reference signal in the second training network device.

[0359] For example, in the case where the second training network device transmits reference signals through part of the third training port, via Figure 14 The channel information processing system shown can determine the complex channel gain parameter of each of the N third training ports. Therefore, when the second training network device transmits reference signals through some of the third training ports, the training downlink CSI of each training UE can also include the complex channel gain parameter of each of the N1 third training ports.

[0360] For each of the N1 third training ports whose importance is ranked by the importance order label, each training downlink CSI includes the complex channel gain parameter of each of the N1 third training ports.

[0361] Step S1302: Process the channel information samples using the initial port model to obtain the training importance value of each third training port.

[0362] Step S1303: Adjust the parameters of the initial port model according to the training importance values ​​and importance order labels of the N1 third training ports to obtain the port model.

[0363] The initial port model after parameter adjustment can be used as the port model.

[0364] In step S1303, the N1 third training ports can be sorted according to their training importance values ​​to obtain a training importance order. Based on the difference between the training importance order and the importance order labels, the parameters of the initial port model are adjusted to obtain the port model. The difference between the training importance order and the importance order labels can be represented as the first loss value.

[0365] Alternatively, the first loss value can be positively correlated with the difference parameter of each of the N1 third training ports, excluding the third training port with the lowest importance. The difference parameter of any of the N1 third training ports, excluding the third training port with the lowest importance, is the sum of at least one sub-difference value of that third training port.

[0366] The i-th sub-difference of any third training port is determined based on the training importance value of that third training port and the i-th training importance value of the i-th target training port corresponding to that third training port. The importance of the i-th target training port corresponding to any third training port mentioned in the importance order label is lower than the importance of the third training port itself.

[0367] If the first difference obtained by subtracting a preset difference value from the training importance value of any third training port is greater than or equal to the i-th training importance value, then the i-th sub-difference of any third training port is 0. If the first difference is less than the i-th training importance value, then the i-th sub-difference is the second difference obtained by subtracting the first difference from the i-th training importance value.

[0368] In other words, for any one of the N1 third training ports, and the i-th target training port whose importance in the importance order label is lower than that third training port, if the training importance value of any third training port is greater than or equal to the sum of the i-th training importance value of the i-th target training port and a preset difference value, then the i-th sub-difference of that third training port is 0. Conversely, if the training importance value of any third training port is less than the sum of the i-th training importance value of the i-th target training port and a preset difference value, then the i-th sub-difference of that third training port is the second difference obtained by subtracting the training importance value of that third training port from the sum of the i-th training importance value of the i-th target training port and the preset difference value.

[0369] For any third training port, each target training port with a lower importance in the importance order label is designated as the i-th target training port corresponding to that third training port, and the i-th sub-difference of that third training port is determined. The sum of all i-th sub-differences of that third training port is calculated to obtain the difference parameter of that third training port.

[0370] Apart from the third training port with the lowest importance in the importance order labels, after determining the difference parameter of each third training port, a first loss value can be determined, which is positively correlated with the difference parameter of each third training port.

[0371] By setting a preset difference value during training, the trained port model can more accurately distinguish ports of different importance, thereby improving the data processing accuracy of the port model.

[0372] The first loss value can be the sum of the difference parameters of each of the third training ports. Alternatively, the first loss value can be obtained by weighted summing of the difference parameters of each of the N third training ports except for the third training port with the lowest importance. The weight of the difference parameter of any third training port is positively correlated with the importance of that third training port as represented by the importance order label.

[0373] Therefore, the first loss value It can be represented as:

[0374] in, To preset the difference value, This represents the training importance value of the a-th third training port in the importance order label. Let be the training importance value of the b-th third training port in the importance order label. The a-th third training port has a higher importance value than the b-th third training port in the importance order label. R represents all groups of third training ports that satisfy the condition that the a-th third training port has a higher importance value than the b-th third training port in the importance order label. Each group of third training ports includes the a-th and b-th third training ports. This represents the number of elements in set R, which is the number of the third training port group. As weight.

[0375] In the importance ranking labels, the third training ports can be arranged in descending order of importance. The set R can be represented as... , This represents the index of the a-th third training port in the importance order label. This indicates the index of the b-th third training port in the importance order label.

[0376] During training, for the calculation of the first loss value, ports with higher importance in the importance ranking labels are assigned higher weights. This ensures that the initial port model prioritizes the accuracy of its processing results for these ports, leading to higher accuracy in handling them. During inference, the port model's processing results are used by the network device to determine candidate ports. Candidate ports are the highest-importance ports among the network device's N ports. The higher accuracy of the port model in processing these important ports results in more accurate candidate port selection by the network device.

[0377] The second training data may also include a map representing the environment surrounding the second training network device. This map could be a traffic map, topographic map, or radio environment map (REM) of the area where the second training network device is located. Therefore, during the inference phase, the port model can be used to process the downlink CSI of multiple UEs and the map representing the environment surrounding the network device to obtain the importance of each of the N ports of the network device.

[0378] The second training data can also include the location information of each of the multiple training UEs. Therefore, during the inference phase, the port model can also be used to process the location information of each of the multiple UEs.

[0379] The second training data may also include measurement results of reference signals transmitted by multiple UEs to at least one fourth training port of the second training network device. The number of antenna elements in the fourth training port is less than the number of antenna elements in the third training port. That is, the network device can transmit a wide beam through the fourth training port, and the network device can transmit a narrow beam through the third training port.

[0380] Therefore, during the inference phase, the port model can also be used to measure the reference signals transmitted by multiple UEs to the fifth port of the network device. The antenna elements in the fifth port are the same as those in the fourth training port, and the beam directions corresponding to the fifth port and the fourth training port can be the same.

[0381] The trained port model can be applied to... Figure 8 , Figure 10 or Figure 20 The communication method shown.

[0382] In step S1005, the channel information determined by the first UE can be channel response features. The feature extraction model used by the first UE to determine the channel response features, and the first prediction model used by the network device to process the channel response features to obtain the downlink CSI of the first UE, can both be models in the channel information processing system. The following will combine... Figure 14 The channel information processing system is described.

[0383] Figure 14 This is a schematic structural diagram of a channel information processing system provided in an embodiment of this application.

[0384] The channel information processing system 1400 includes a second prediction model 1410, an encoder 1420, and a decoder 1430. The channel information processing system 1400 can be installed in one or more electronic devices. For example, the channel information processing system 1400 can be installed in a UE or a network device. Alternatively, the second prediction model 1410 and the encoder 1420 can be installed in the UE, and the decoder 1430 can be installed in the network device.

[0385] When t is greater than 2, the second prediction model 1410 is used to process the channel response information and the channel response prediction information of the (t-1)th measurement period to obtain the channel response prediction information of the tth measurement period. When t is greater than 1, the second prediction model 1410 is used to process the channel response information of the (t-1)th measurement period to obtain the channel response prediction information of the tth measurement period. The second prediction model 1410 can be a GRU or other neural network.

[0386] In the first measurement period, the second prediction model 1410 may not require processing. For example... Figure 15 As shown, in the second measurement period, the second prediction model 1410 can process the channel response information of the first measurement period to obtain the channel response prediction information of the second measurement period. In the third measurement period, the second prediction model 1410 can process both the channel response information of the second measurement period and the channel response prediction information of the second measurement period to obtain the channel response prediction information of the third measurement period.

[0387] When t is 1, encoder 1420 is used to encode the channel response information of the t-th measurement period to obtain the channel response features of the t-th measurement period. When t is greater than 1, encoder 1420 is used to encode the channel response prediction information and the channel response information of the t-th measurement period to obtain the channel response features of the t-th measurement period.

[0388] Decoder 1430 is used to decode the channel response characteristics of the t-th measurement period to obtain the t-th downlink CSI.

[0389] The encoder 1420 may include a splicing layer 1421, a first sorting layer 1422, a feature extraction model 1423, and a quantization layer 1424.

[0390] The channel response information for the t-th measurement period includes the complex channel gain parameters for each of the M ports through which the network device transmits the reference signal in the t-th measurement period.

[0391] The splicing layer 1421 is used to add the complex channel gain parameter of at least one first port that does not belong to the M ports in the channel response prediction information of the t-th measurement period to the channel response information of the t-th measurement period, so as to obtain the channel information set of the t-th measurement period.

[0392] The first sorting layer 1422 is used to sort the complex channel gain parameters in the channel information set of the t-th measurement period according to the magnitude, so as to obtain the sorted channel information set of the t-th measurement period.

[0393] Feature extraction model 1423 is used to extract features from the sorted channel information set after the t-th measurement period to obtain the channel response features of the t-th measurement period. The channel response features of the t-th measurement period include the port channel features of each of the multiple target ports. Feature extraction model 1423 can be an MLP or other neural network.

[0394] The quantization layer 1424 is used to quantize the channel response features of the t-th measurement period to obtain the quantized channel response features of the t-th measurement period. The quantized channel response features of the t-th measurement period include the quantized port channel features of each target port.

[0395] The decoder 1430 may include a second sorting layer 1431 and a first prediction model 1432.

[0396] The second sorting layer 1431 is used to expand the quantized port channel features of each target port in the quantized channel response features of the t-th measurement period according to the beam direction corresponding to the port, so as to obtain the expanded channel response features of the t-th measurement period.

[0397] The first prediction model 1432 is used to process the expanded channel response characteristics in the t-th measurement period to obtain the downlink CSI of the first UE in the t-th measurement period. The t-th downlink CSI includes the complex channel gain parameters of the N ports of the network device. The first prediction model 1432 can be a CNN or other neural network.

[0398] The parameter t can increase over time. The network device can initialize parameter t, setting it to an initial value. After initializing parameter t, the network device can also send initialization information to instruct UEs within its coverage area to set parameter t to the initial value. The network device can initialize parameter t periodically or non-periodically. The network device can initialize parameter t when multiple candidate ports change. The initialization information may, for example, be update periodicity information, instructing the UE to no longer determine channel response prediction information based on channel response information determined before receiving the update periodicity information.

[0399] like Figure 9 As shown, the network device can set the parameter t to an initial value of 1 in each candidate port update cycle. Each candidate port update cycle can include T measurement cycles, where T is a positive integer. The candidate ports corresponding to different candidate port update cycles can be different.

[0400] During each candidate port update cycle, the network device can use multiple candidate ports corresponding to that update cycle. The network device can divide these multiple candidate ports into Q port groups. Generally, the number of measurement cycles T in a candidate port update cycle is much larger than the number of port groups Q.

[0401] In each measurement cycle of this candidate port update cycle, the network device can cyclically select a port group from Q port groups to transmit reference signals. For example, Figure 9 In the i-th candidate port update period shown, the number of port groups Q determined by the network device is 2, and the Q port groups are port group q1 and port group q2.

[0402] For the (i+1)th candidate port update period following the (i-th)-th candidate port update period, the network device can re-determine Q port groups based on the multiple candidate ports corresponding to the (i+1)-th candidate port update period. The Q port groups in the (t+1)-th candidate port update period may include port group q1' and port group q2'.

[0403] The second prediction model 1410 processes the channel response prediction information for the (t-1)th measurement period and the channel response information for the (t-1)th measurement period to obtain the channel response prediction information for the t-th measurement period. The channel response prediction information for the (t-1)th measurement period may include the complex channel gain parameters of M1 ports in the network device, and the channel response information for the (t-1)th measurement period may include the complex channel gain parameters of M2 ports in the network device. The channel response prediction information for the t-th measurement period may include the complex channel gain parameters of M3 ports. These M3 ports can be the union of the M1 ports and the M2 ports. That is, the M3 ports include both the M1 ports and the M2 ports.

[0404] like Figure 16 As shown, the M1 ports can include port 0, port 2, and port 4. The M2 ports can include port 0 and port 3. The M3 ports can include port 0, port 2, port 3, and port 4.

[0405] In step S1001, the network device processes the downlink CSI of the UE determined in the (i-1)th candidate port update cycle using the port model to determine multiple candidate ports in the ith candidate port update cycle. Before each candidate port update cycle, the network device can determine candidate ports using the port model. That is, the candidate ports can continuously change.

[0406] In network devices, the number of ports can be far greater than the number of candidate ports. If, at the beginning of each candidate port update period, t is not set to an initial value, meaning that in a new candidate port update period, the second prediction model 1410 still determines the channel response prediction information for the current measurement period based on the channel response prediction information and channel response information of the previous measurement period, and determines the channel information set based on the channel response prediction information and channel response information of the current measurement period, then the channel information set may include complex channel gain parameters for a large number of ports. Furthermore, as the candidate port update period increases, the number of ports in the channel information set continuously increases.

[0407] The channel response features sent by the first UE to the network device include port channel features of each port in the channel information set. As the number of ports in the channel information set increases, the amount of data in the channel response features sent by the first UE to the network device increases. Furthermore, the accuracy of the complex channel gain parameters for ports in the channel information set that have not transmitted reference signals for extended periods is relatively low.

[0408] Therefore, at the beginning of each candidate port update cycle, the network device can set the parameter t to the initial value 1. That is, the first UE no longer determines the channel information set based on the measurement results of the reference signals sent by each port of the network device in the previous candidate port update cycle. This reduces the occupation of communication resources while making the channel response characteristics more accurate.

[0409] For example, in the process of Figure 10 During the process illustrated, the network device can also send information related to the candidate port update period and measurement period to indicate the duration of the candidate port update period and measurement period, etc. For example, the network device can send update period information at the beginning of each candidate port update period to instruct the UE not to determine the channel response prediction information based on the channel response information determined before receiving the update period information.

[0410] After Q measurement cycles have elapsed since the start of each candidate port update cycle, the channel information set determined by the first UE can include the complex channel gain parameters of each candidate port. Therefore, after Q measurement cycles have elapsed since the start of each candidate port update cycle, the downlink CSI of the first UE determined by the network device takes into account the channel conditions of each candidate port, making the downlink CSI determined by the network more accurate.

[0411] When the first UE sorts the complex channel gain parameters in the channel information set according to the beam direction corresponding to the port, multiple port channel features in a channel response feature are arranged according to the beam direction corresponding to the port. Before the start of each candidate port update period and after Q measurement periods, the port channel features of other candidate ports besides the target port in the channel response features sent by the first UE can be set to preset values. After the start of each candidate port update period and after Q measurement periods, the channel response feature includes the port channel features of each candidate port. Thus, the channel response feature includes the port channel features of each candidate port, and multiple port channel features are arranged according to the beam direction corresponding to the port. Therefore, the network device can determine the port corresponding to each port channel feature in the channel response feature.

[0412] When the first UE sorts the complex channel gain parameters in the channel information set according to their amplitude, it can send port information to the network device to indicate the target port corresponding to each port channel feature of the channel response characteristics. If the channel response characteristics include port channel features of ports other than those in the current candidate port update period, the first UE can indicate the port corresponding to each port channel feature using the port information used to uniquely identify each port in the network device. Conversely, if the channel response characteristics include port channel features of multiple candidate ports in the current candidate port update period but not port channel features of other ports, the UE can indicate the candidate port corresponding to each port channel feature using the port information used to uniquely identify multiple candidate ports.

[0413] In network devices, the total number of ports, N, is much larger than the number of candidate ports. Therefore, the bit width of the port information used to uniquely identify multiple candidate ports can be smaller than the bit width of the port information used to uniquely identify each port in the network device. Consequently, when the channel response characteristics do not include the port channel characteristics of ports other than the multiple candidate ports in the current candidate port update period, i.e., when the channel information set does not include the complex channel gain parameters of ports other than the multiple candidate ports in the current candidate port update period, the port information sent by the first UE to the network device can have a smaller data volume.

[0414] During each candidate port update cycle, the network device can send candidate port information to the UE to indicate multiple candidate ports for that candidate port update cycle.

[0415] For example, in some embodiments, after the first UE sorts the complex channel gain parameters in the channel information set according to their amplitude and extracts features from the sorted channel information set using a feature extraction model to obtain channel response features, it can further sort the port channel features of multiple target ports in the channel response features according to the beam direction corresponding to the port to obtain reordered channel response features. The reordered channel response features may include the port channel features of each candidate port. Therefore, the first UE can send the reordered channel response features to the network device instead of sending the port information.

[0416] The network device can set the port characteristics of ports other than candidate ports in the reordered channel response characteristics to preset values ​​to obtain expanded channel response characteristics. The network device can process the expanded channel response characteristics using a first prediction model to obtain the downlink CSI of the first UE. The downlink CSI of the first UE includes the complex channel gain parameters of N ports in the network device.

[0417] The second prediction model 1410, feature extraction model 1423, and first prediction model 1432 in the channel information processing system 1400 are all neural network models, and all can be trained. The following section combines... Figure 17 The training process of the channel information processing system 1400 is explained.

[0418] Figure 17 This is a schematic flowchart of a training method provided in an embodiment of this application. Figure 17 The training method shown includes steps S1701 to S1703.

[0419] Step S1701: Obtain first training data, which includes a set of training channel information and channel state labels.

[0420] The training channel information set includes the first training channel response information.

[0421] The first training channel response information is information determined by the first training user equipment based on the first training reference signal sent by the first training network equipment. The first training reference signal is associated with M' first training ports of the first training user equipment. In other words, the first training channel response information can be information obtained by the first training user equipment measuring the first training reference signal sent by the first training network equipment through each of the M' first training ports. The first training channel response information includes the complex channel gain parameters of K' of the M' first training ports.

[0422] The channel state label includes the complex channel gain parameters of the N first training ports of the first training network device. The N first training ports of the first training network device include the M' first training ports, where M' and K' are both positive integers greater than 1.

[0423] The N first training ports of the first training network device and Figure 8 , Figure 10 The network devices in the method shown have N identical ports.

[0424] In other words, the first training network device and the network device can have the same antenna array. Furthermore, the codebooks corresponding to the N first training ports in the first training network device are the same as those corresponding to the N ports in the network device.

[0425] Therefore, the beam directions corresponding to the N first training ports in the first training network device are the same as the beam directions corresponding to the N ports in the network device. When the beamforming weight vectors represented by the same codewords are applied to the antenna arrays of the first training network device and the network device respectively, the directional beam formed by the first training network device in space is the same as the directional beam formed by the network device in space.

[0426] Step S1702: The training channel information set is processed through the initial feature extraction model to obtain the training channel response features.

[0427] Step S1703: Process the training channel response features using the first initial prediction model to obtain the training channel state.

[0428] Step S1704: Based on the difference between the training channel state and the channel state label, adjust the parameters in the initial feature extraction model and the first initial prediction model to obtain the feature extraction model and the first prediction model.

[0429] Before step S1702, the complex channel gain parameters of multiple training target ports in the training channel information set can be sorted according to their amplitude to obtain a sorted training channel information set. The multiple training target ports include the K' first training ports. The processing of the training channel information set through the initial feature extraction model can be the processing of the sorted training channel information set through the initial feature extraction model.

[0430] The complex channel gain parameters of different ports vary significantly, and those with smaller amplitude attenuation, represented by amplitude parameters, are concentrated in a few ports among the multiple training target ports. The initial feature extraction model faces considerable difficulty in stably extracting features from complex channel gain parameters with large amplitude differences. Sort the complex channel gain parameters of multiple training target ports in the training channel information set according to their amplitude magnitude. This improves the stability of the training channel response features extracted by the initial feature extraction model and facilitates its training.

[0431] The training channel response features can include the port channel features of each training target port among multiple target ports. Before proceeding to step S1703, the port channel features of each target port in the training channel response features can be filled into the positions corresponding to each training target port in the initial features to obtain the expanded training channel response features. The N positions in the initial features are arranged according to the beam direction of the corresponding first training port. In the expanded training channel response features, the positions of the first training ports other than the training target ports are all preset values.

[0432] In step S1703, the sorted training channel response features can be processed using the first initial prediction model to obtain the training channel state. The sorted training channel response features may include the port channel features of N first training ports. The port channel features of the N first training ports are arranged according to the beam directions corresponding to the N first training ports.

[0433] The signal transmitted through the port exhibits power leakage, meaning that signal energy radiates beyond the intended main beam direction. The degree of power leakage in each direction is related to the directional difference between the beam direction corresponding to that port and the respective directions. For example... Figure 18 As shown, the signal transmitted through a port has its energy mainly concentrated in the beam direction corresponding to that port. However, the signal can also be detected in other directions far from that beam direction. The energy of the signal in these other directions is much lower than its energy in the beam direction corresponding to that port.

[0434] By processing the sorted training channel response features using the first initial prediction model, the first initial prediction model can learn the impact of power leakage at the target port on the channel state during the training process, thereby making the data processing of the trained first prediction model more accurate.

[0435] The trained feature extraction model and the first prediction model are configured, with the feature extraction model installed in the UE and the first prediction model installed in the network device. To reduce the amount of data the UE sends to the network device, a quantization layer can be added after the feature extraction model. This allows the UE to send quantized channel response features to the network device.

[0436] During training, the use of quantization layers can lead to gradient truncation in neural networks. Quantization is a piecewise constant function that maps continuous values ​​to discrete steps. In most regions, the derivative of this function is zero. At step transition points, the derivative is infinite or undefined. This means that in standard backpropagation, gradients cannot effectively flow back through the quantization layer, preventing preceding network layers from receiving effective gradient updates.

[0437] During training, the quantization layer can be replaced with a smooth, differentiable function to approximate the quantization function, making its gradient computationable. In other words, during training, a continuous, smooth function can be used to approximate a discrete, non-differentiable operation.

[0438] For example, the function used to replace the quantization layer during training can be a stacked sigmoid function.

[0439] A single sigmoid function can only produce a smooth "S"-shaped curve with an output range of (0, 1), and cannot directly simulate a quantizer with multiple discrete steps. To simulate a multi-level quantizer, theoretically, multiple sigmoid functions with different offsets can be linearly combined to construct a curve with multiple smooth steps. The quantization layer can be a multi-level quantizer. For example... Figure 19 As shown, with the same input, the difference between the output of the stacked sigmoid function and the output of the quantization layer is small.

[0440] Steps S1701 to S1704 can be used to train the feature extraction model and the first prediction model. During the training of the feature extraction model and the first prediction model, the number of first training data points used can be one or more. The number M' of the first training reference signals associated with the first training channel response information in different first training data points can be the same or different.

[0441] The second prediction model can also be trained.

[0442] In some embodiments, the second prediction model may be trained based on other training data.

[0443] The training data used to train the second prediction model may include a first training sample and channel response labels for the m-th training measurement period. The first training sample includes training channel response information for the (m-1)-th training measurement period. The first training sample may also include training channel response prediction information for the (m-1)-th training measurement period.

[0444] The first training sample is processed using the second initial prediction model to obtain the training channel response prediction information for the m-th training measurement period.

[0445] Based on the difference between the predicted channel response information of the m-th training measurement period and the channel response label of the m-th training measurement period, the parameters of the second initial prediction model are adjusted to obtain the second prediction model.

[0446] In other embodiments, the second prediction model, the feature extraction model, and the first prediction model may be trained in an end-to-end manner. The initial channel information processing system may include an initial feature extraction model, a first initial prediction model, and a first initial prediction model.

[0447] The training channel information set used during the training of the feature extraction model and the first prediction model may include the complex channel gain parameters of the target first training port in the training channel response prediction information of the m-th training measurement period. The first training channel response information is the training channel response information of the m-th training measurement period. The training channel response prediction information of the m-th training measurement period is determined by the second initial prediction model based on the training channel response information of the (m-1)-th training measurement period. The training channel response prediction information of the m-th measurement period includes the complex channel gain parameters of P' first training ports out of the N first training ports, where the target first training port belongs to these P' first training ports and does not belong to these M' first training ports, and m is an integer greater than 1.

[0448] The second initial prediction model can also be adjusted based on the difference between the training channel state and the channel state label.

[0449] In other words, the training data for the second prediction model, the feature extraction model, and the first prediction model can include second training samples and channel state labels. The second training samples include the training channel response information for the (m-1)th training measurement period and the training channel response information for the mth training measurement period.

[0450] For example, when m is 2, the training channel response information of the (m-1)th training measurement period can be processed using the second initial prediction model to obtain the training channel response prediction information of the m-th measurement period. When m is greater than 2, the training samples also include the training channel response prediction information of the (m-1)th training measurement period. By processing the training channel response information of the (m-1)th training measurement period and the training channel response prediction information of the (m-1)th training measurement period using the second initial prediction model, the training channel response prediction information of the m-th measurement period can be obtained.

[0451] For example, the splicing layer can process the training channel response prediction information of the m-th measurement period output by the second prediction model and the first training channel response information to obtain a training channel information set.

[0452] During training, the second prediction model learns the temporal correlation between channel response information from two adjacent training measurement periods. This allows it to predict the channel response information for the current training measurement period based on the channel response information from the previous training measurement period, thus determining the predicted channel response information for the current training measurement period. The training measurement periods used in the training data for the second prediction model can be equal to the measurement periods used in the inference process of the trained second prediction model. Consequently, the processing results of the second prediction model have high accuracy.

[0453] Multiple sub-models designed to achieve a specific goal are trained separately. Each sub-model optimizes its own intermediate objective, and each module may reach a local optimum on its own task. However, there may be objective conflicts between modules, resulting in suboptimal performance for the combined system. In contrast, end-to-end training involves all parameters working together to optimize the final global objective, and the system tends to find the globally optimal solution from input to output.

[0454] The difference between the training channel state and the channel state label can be represented by a second loss value. The second loss value can be the normalized mean squared error (NMSE).

[0455] A channel information processing system can be trained based on multiple training datasets. These training datasets can be divided into multiple batches. During training, the initial channel information processing system can process a second training sample from one batch of training data at a time. The second loss value can represent the difference between the training channel state obtained by the initial channel information processing system for each second training sample in a batch and the channel state label corresponding to that second training sample. The second loss value can be expressed as:

[0456] in, This represents the training channel state obtained by the initial channel information processing system after processing the second training sample in the u-th training data of a batch. This represents the channel state label in the u-th training data of this batch, where U represents the total number of training data in this batch. This represents the sum of squares of the components in x.

[0457] During end-to-end training, the convergence speed of the second initial prediction network can be higher than that of the initial feature extraction model and the first initial prediction model. Different learning rates can be set for the second prediction network and the feature extraction model. For example, the learning rate of the second initial prediction network could be 5 × 10⁻⁶. -4 The learning rate of the initial feature extraction model and the first initial prediction model can be 2×10. -3 .

[0458] To determine a more accurate UE downlink CSI with less communication resource consumption, embodiments of this application also provide, as follows: Figure 20 The communication method shown.

[0459] Figure 20 This is a schematic flowchart of a communication method provided in an embodiment of this application. Figure 20 The method shown can be applied to a communication system, specifically during channel probing. The communication system includes network equipment and a first UE. The network equipment can provide services to multiple UEs, and the first UE can be any one of those UEs. Figure 20 The communication method shown includes steps S2010 to S2070.

[0460] In step S2010, the network device acquires downlink CSI of multiple UEs. The downlink CSI of each UE includes the complex channel gain parameters of multiple target ports among the N ports of the network device. Different ports in the network device correspond to different beam directions, and N is a positive integer greater than 1.

[0461] In step S2020, the network device processes the downlink CSI of multiple UEs through the port model to obtain the importance of each of the N ports.

[0462] In step S2030, the network device determines multiple candidate ports from the N ports based on the importance of each port, with each candidate port having a higher importance than the other ports in the N ports besides the candidate ports.

[0463] In step S2040, the network device sends reference signals through multiple candidate ports, and the reference signals sent by any two candidate ports are orthogonal.

[0464] Network devices can send reference signals through these multiple candidate ports during a measurement cycle.

[0465] Alternatively, the network device can transmit a reference signal through a subset of candidate ports in each measurement cycle. After multiple measurement cycles, the network device can transmit the reference signal through all candidate ports. For example, the network device can divide multiple candidate ports into Q port groups, and cyclically select port groups from these Q groups, transmitting the reference signal through a port in one of these port groups in each measurement cycle.

[0466] In step S2050, the first UE measures the reference signal sent by each candidate port to obtain channel information.

[0467] The first UE can be any UE within the coverage area of ​​the network device. In the downlink CSI of multiple UEs used by the network device in determining the importance of each of the N ports, the multiple UEs may or may not include the first UE.

[0468] Channel information can be determined based on channel response information. Channel response information may include complex channel gain parameters for all or part of the ports transmitting the reference signal. Channel information can include channel response information. Alternatively, the first UE can perform feature extraction on the channel response information to obtain channel response features. These channel response features can serve as channel information.

[0469] When a network device periodically sends a reference signal through a port, the channel information can be determined based on the channel response information of one or more measurement periods.

[0470] When a network device transmits a reference signal through all candidate ports in a measurement cycle, the channel information that the first UE can transmit to the network device in each measurement cycle can be determined by the channel response information obtained by measuring the reference signal in the current measurement cycle.

[0471] When the network device transmits reference signals through a subset of candidate ports in each measurement cycle, the channel information that the first UE can transmit to the network device in each measurement cycle can be determined based on the channel response information of one or more measurement cycles.

[0472] For example, the first UE can extract features from the channel response information of the current measurement period using a feature extraction model. The channel response features obtained from feature extraction can be used as channel information.

[0473] For example, the first UE can also determine the predicted channel response information for the t-th measurement period based on the channel response information for the (t-1)-th measurement period using a second prediction model. The first UE can determine the channel information set based on the predicted channel response information and the channel response information for the t-th measurement period. The channel information set includes the channel response information for the t-th measurement period and the complex channel gain parameter of the first port. The channel response information for the t-th measurement period does not include the complex channel gain parameter of the first port, and the predicted channel response information for the t-th measurement period includes the complex channel gain parameter of the first port. The first UE can extract features from the channel information set using a feature extraction model. The channel response features obtained from the feature extraction can be used as channel information.

[0474] In step S2060, the first UE sends channel information to the network device.

[0475] Correspondingly, the network device receives the channel response information. The first UE sending channel information to the network device can be understood as PMI reporting.

[0476] In step S2070, the network device determines the downlink CSI of the first UE based on the channel response information.

[0477] When the channel information is a channel response feature, the network device can process the channel response feature through the first prediction model to obtain the downlink CSI of the first UE.

[0478] For details on the UE's determination of channel response characteristics and the network device's processing of these characteristics, please refer to [link to relevant documentation]. Figure 8 ,as well as Figures 14 to 16 Explanation.

[0479] The communication method provided in this application embodiment allows the network device to process the downlink CSI of multiple UEs through a port model, obtain the importance of each port among the N ports of the network device, and determine multiple candidate ports among the N ports based on the importance of each port. The importance of the candidate ports is higher than that of other ports. The downlink CSI of each UE includes the multiple channel gain parameters of the N ports.

[0480] The importance of a port is determined based on the downlink CSI of multiple UEs. Each UE's downlink CSI is influenced by its location and surrounding environment. Therefore, the importance of a port reflects the demand of the network environment on each port. Based on the importance of each port, multiple candidate ports are determined from these N ports, allowing for a non-uniform, adaptive selection of the port for transmitting the reference signal.

[0481] The higher the importance of a port, the denser the number of UEs (User Equipments) are near the propagation direction of the signal transmitted from that port. In densely populated UE areas, using more ports whose propagation direction passes through that area to transmit reference signals yields the downlink CSI (Channel Sounding Index) for each UE. This allows for communication with each UE based on its downlink CSI, effectively eliminating inter-user interference. Conversely, in sparsely populated UE areas, fewer reference signals from candidate ports are needed within the area traversed by the main propagation path, resulting in better communication quality. This leads to efficient channel information acquisition.

[0482] Network devices transmit reference signals through candidate ports of higher importance, making the transmission of reference signals by the network devices adapt to the environment. That is, the beam direction corresponding to the candidate port matches the actual channel environment and user distribution, thereby making the downlink CSI of the first UE more accurate.

[0483] Because the location of UEs within the coverage area of ​​network equipment may change in real time, the distribution trend of UEs is different at different times. For example, during working hours, UEs are more concentrated in office areas such as office buildings and factories; while during non-working hours, UEs are more concentrated outdoors and in residential areas.

[0484] To improve the accuracy of the downlink CSI of the UE determined by the network device, the network device can determine the importance of N ports and re-determine the candidate ports. For example, the network device can process the downlink CSI of the UE obtained in one or more recent candidate port update cycles using a port model, thereby updating the candidate ports.

[0485] The following example demonstrates the effectiveness of the solution provided in this application, with a cell radius rc of 300 meters (m), a base station height h of 25 meters, 5 UEs within the network equipment coverage area, 32 ports N in the network equipment, 1 antenna element per UE, a carrier frequency fc of 3.5 GHz for the reference signal, coexistence of indoor and outdoor users, an indoor user probability Pin of 0.8, coexistence of line-of-sight (LOS) and non-line-of-sight (NLOS) UEs, 20 paths in each NLOS cluster, a downlink measurement signal-to-noise ratio (SNR) of 30 dB, a measurement period of 5 milliseconds (ms), a candidate port update period of 100 ms, and 2 port groups Q.

[0486] Figure 21(a) to (c) in the figure represent three different UE location examples. The dashed lines in the figure indicate the direction of the UE relative to the network device. When the network device transmits a reference signal through eight ports with evenly distributed beam directions, allocating a large number of ports for reference signal transmission in areas without users wastes communication resources. However, in densely populated user areas, the beam directions corresponding to the ports transmitting the reference signal are sparsely distributed, leading to insufficient accuracy in user measurement results.

[0487] Network devices process downlink CSI from multiple UEs using a port model to determine candidate ports. When the number of ports is 8, in... Figure 21 In the UE location examples shown in (a) to (c), the beam directions corresponding to the candidate ports determined by the network device are respectively as follows: Figure 21 As shown in (d) to (f).

[0488] Communication systems can be Figure 3 The method shown is used for channel detection. Figure 22 The diagram illustrates the error in the downlink CSI of the UE determined by the network device when transmitting reference signals through ports with uniform beam distribution and through candidate ports. The error in the downlink CSI of the UE can be expressed as NMSE. Figure 22 Method 'a' in the diagram indicates that the network device transmits reference signals through ports that are evenly distributed in the beam direction. Figure 22 In mode b, the network device transmits the reference signal through all candidate ports in one measurement cycle. In modes a and b, the number of ports through which the network device transmits the reference signal in one measurement cycle is Ns. The UE can send the quantized complex channel gain parameters of L ports out of the Ns ports to the network device.

[0489] In three scenarios where the number of ports (Ns) transmitting reference signals in each measurement cycle are 4, 8, and 16, respectively, the network device effectively reduces the error of the determined downlink CSI of the UE by transmitting reference signals through candidate ports. As the number of ports (L) corresponding to the complex channel gain parameters that the UE can transmit to the network device increases, the error of the downlink CSI determined by the network device decreases. Furthermore, as the number of ports (Ns) through which the network device transmits reference signals in a measurement cycle increases, the error reduction in method b, i.e., transmitting reference signals through candidate ports, becomes more significant.

[0490] Figure 23This paper illustrates the relationship between the normalized downlink CSI error of the UE when the network device transmits reference signals through 24 ports evenly distributed along the beam direction as the standard value, and the number of ports transmitting reference signals as the reference signal is transmitted through candidate ports. By using method b, i.e., transmitting reference signals through candidate ports, and determining 12 candidate port reference signals, the accuracy of the UE's downlink CSI can be achieved, which is the same as when the network device transmits reference signals through 24 ports evenly distributed along the beam direction. This reduces communication resource consumption by 50% while still obtaining a high-accuracy UE downlink CSI. Under the same UE downlink CSI accuracy, transmitting reference signals through candidate ports can effectively reduce the communication resource consumption of channel probing.

[0491] Figure 24 and Figure 25 The error of the downlink CSI of the UE determined by the network device through four methods, from method a to method d, is shown.

[0492] In mode a, during each measurement cycle, the network device transmits reference signals through Ns ports evenly distributed along the beam direction. The UE measures the reference signals transmitted from each port to determine its downlink CSI. The UE then transmits its downlink CSI to the network device. The downlink CSI transmitted by the UE includes the quantized complex channel gain parameters of the L ports with the smallest amplitude attenuation among the Ns ports.

[0493] In mode b, during each measurement period, the network device transmits reference signals through Ns determined candidate ports. The UE measures the reference signals transmitted by each port to determine the UE's downlink CSI. The UE transmits its downlink CSI to the network device. The downlink CSI transmitted by the UE includes the quantized complex channel gain parameters of the L ports with the smallest amplitude attenuation among the Ns ports.

[0494] In mode c, during each measurement cycle, the network device transmits reference signals through Ns determined candidate ports. The UE measures the reference signals transmitted by each port to determine the complex channel gain parameter for each candidate port. The UE extracts features from the complex channel gain parameters of the L ports with the smallest amplitude attenuation among the Ns ports using an encoder, obtaining channel response features, and transmits these features to the network device. The network device determines the UE's downlink CSI based on the channel response features, and the network device's determination includes the complex channel gain parameters of all ports within the network device.

[0495] In mode d, the network device determines 2Ns candidate ports and divides them into two port groups. Each port group includes Ns candidate ports. During each measurement cycle, the network device cyclically selects a port from one of the two port groups to transmit a reference signal. The UE measures the reference signal transmitted by each port to determine the channel response information for the current measurement cycle. The channel response information for the current measurement cycle includes the complex channel gain parameters of the L ports with the smallest amplitude attenuation among the Ns ports transmitting the reference signal. The UE processes the channel response information from the previous measurement cycle using a second prediction model to determine the predicted channel response information for the current measurement cycle. The UE processes the predicted channel response information and the channel response information for the current measurement cycle using an encoder to obtain channel response characteristics and transmits these characteristics to the network device. The network device determines the UE's downlink CSI based on the channel response characteristics. The determined downlink CSI includes the complex channel gain of all ports in the network device.

[0496] like Figure 24 and Figure 25 As shown, with the increase of the number of ports Ns transmitting reference signals in a measurement cycle, the error of the downlink CSI of the UE determined by methods a to d all decreases. Figure 24 With the number of ports Ns transmitting reference signals remaining constant during a measurement cycle, the error of the UE's downlink CSI determined by methods a to d decreases sequentially.

[0497] In modes a and b, the downlink CSI transmitted by the UE includes L port-quantized complex channel gain parameters. In modes c and d, the channel response features transmitted by the UE to the network device may include multiple port-quantized port channel features. For example... Figure 25 As shown, the increase in downlink CSI error of the UE is not significant as the quantized bit width decreases. In other words, methods a through d are all robust to quantization operations.

[0498] In the cases of method c and method d, the sorting layer may or may not be used before feature extraction is performed through the feature extraction model in the encoder. Figure 26The diagram illustrates the error of the downlink CSI of the UE determined by the trained channel information processing system in two scenarios during training: sorting the complex channel gain parameters by amplitude and sorting them by the beam direction corresponding to the port. The error of the downlink CSI of the UE processed by the trained channel information processing system decreases with increasing training epochs. Compared to the scenario where the complex channel gain parameters are sorted by beam direction corresponding to the port during both training and inference, and then processed by the feature extraction model, the method of sorting the complex channel gain parameters by amplitude during training and inference, followed by processing by the feature extraction model, significantly reduces the error of the downlink CSI of the UE obtained during inference.

[0499] The network device performs precoding based on the determined downlink CSI of the UE and communicates with the UE, with spectral efficiency as follows: Figure 27 As shown. Spectral efficiency represents the number of bits that can be reliably transmitted per unit bandwidth per unit time. It is a fundamental indicator for measuring the advancement of wireless communication technology. The higher the accuracy of the UE's downlink CSI, the higher the spectral efficiency in communication based on the UE's downlink CSI. Spectral efficiency improves as the number of epochs in the training process of the channel information processing system increases. Furthermore, compared to sorting the complex channel gain parameters according to the beam direction corresponding to the port, sorting the complex channel gain parameters according to amplitude during training and inference, and then processing them through a feature extraction model, results in higher spectral efficiency for communication based on the determined UE downlink CSI.

[0500] Compared to method a, determining the UE's downlink CSI using method d reduces the number of ports requiring reference signal transmission by 50% while achieving the same downlink CSI accuracy, thus reducing pilot overhead. Compared to method a, with the same number of ports transmitting reference signals per measurement cycle, method d improves the UE downlink CSI accuracy by 2.92 dB and enhances channel estimation accuracy. Considering inter-user interference, compared to method a, network devices communicating with the UE based on the downlink CSI determined by method d achieve a 22.2% improvement in spectral efficiency.

[0501] Compared to method a, methods b, c, and d transmit reference signals through determined candidate ports. The amplitude attenuation represented by the amplitude parameter in the complex channel gain parameter of the channel response information measured by the UE is smaller, that is, the codeword utilization is higher. Therefore, the accuracy of the downlink CSI determined based on the channel response information measured by the UE is improved.

[0502] For example, network devices can also determine the downlink CSI of the UE based on the partial heterogeneity of uplink and downlink channels and the time correlation of the channels, thereby achieving more efficient and accurate CSI determination.

[0503] To improve communication efficiency, this application also provides a communication method.

[0504] Figure 28 This is a schematic flowchart of a communication method provided in an embodiment of this application. Figure 28 The method shown can be applied to a communication system, and specifically, to the process of channel probing within the communication system. The communication system includes network equipment and a first UE.

[0505] In step S2801, the network device sends a reference signal at Q1 time intervals. During these Q1 time intervals, the reference signals sent at different time intervals are associated with different port groups. The port groups associated with the reference signals sent at these Q1 time intervals include multiple candidate ports, and Q1 is an integer greater than 1.

[0506] For example, a first reference signal sent by the network device at a first time is associated with a third port group, and a reference signal sent by the network device at a second time is associated with a fourth port group. The ports in the third port group and the fourth port group are different. The ports in both the third port group and the fourth port group can be candidate ports. Exemplarily, each port group may or may not include other ports besides candidate ports.

[0507] The reference signal transmitted at each of the Q1 time periods can be a CSI-RS.

[0508] Correspondingly, UEs within the coverage area of ​​the network equipment can receive the reference signal in Q1 time.

[0509] The port group associated with the reference signal sent by the network device at each time in the Q1 time period can also be understood as the port used by the network device to send the reference signal.

[0510] In step S2802, the first UE sends a first CSI report to the network device. The first CSI report is associated with a set of ports, which includes the multiple candidate ports.

[0511] Accordingly, the network device can receive the first CSI report sent by the first UE. The first CSI report can be determined based on a reference signal at time Q1.

[0512] Both the third port group and the fourth port group can be subsets of the port set.

[0513] For example, if the port set includes two port groups, namely the third port group and the fourth port group, Q1 can be 2. After the first time and the second time, the first UE can send a first CSI report to the network device. The ports in the third port group and the fourth port group are different.

[0514] The port set associated with the first CSI report can be understood as the port set corresponding to the first CSI report, the port set covered by the first CSI report, or the port set used by the first CSI report. For example, the first CSI report may include measurement information corresponding to the port set. The measurement information corresponding to the port set can also be understood as the measurement information of the first UE on the reference signal associated with the port in the port set.

[0515] The port set may or may not include ports other than the candidate ports. For example, the port set may include N ports in a network device, which include multiple candidate ports, and N is greater than the number of candidate ports. The N ports in the network device may, for example, be all the ports of the network device.

[0516] The first CSI report may include at least one of the following: precoding matrix information, codebook information, feature information, and downlink channel response information.

[0517] The precoding matrix information can be associated with a port set. When the port set includes N ports from the network device, the first UE can perform feature extraction on the channel information set, process the extracted channel response features to obtain the first UE's downlink CSI, and perform precoding based on the first UE's downlink CSI to obtain the first PMI. The precoding matrix information can be the first PMI determined by the first UE.

[0518] Codebook information can be associated with a set of ports. Codebook information can be used to indicate the first PMI. For example, codebook information can be an index of the first PMI.

[0519] Feature information can be associated with a set of ports. When the port set includes N ports from the network device, the first UE can extract features from the channel information set to obtain channel response features. The network device can process the channel response features using a first prediction model to obtain the downlink CSI of the first UE. Feature information can be channel response features.

[0520] Downlink channel response information is associated with a set of ports. When the port set includes N ports in the network device, the network device can extract features from the channel information set including the first channel response information, and process the extracted channel response features to obtain the downlink CSI of the first UE. The downlink channel response information can represent the channel information set.

[0521] In some embodiments, the first user equipment (UE) may also send Q1-1 second CSI reports to the network device. The Q1-1 second CSI reports each correspond to the first Q-1 times within the Q times. For each of the first Q-1 times, after receiving the reference signal sent by the network device at each of the first Q-1 times, the first UE may send the second CSI report corresponding to that time to the network device. The port group associated with the second CSI report corresponding to each of the first Q-1 times is the port group associated with the reference signal sent by the network device at that time.

[0522] For example, if the port set includes two port groups, a third port group and a fourth port group, and if the first time of sending the first reference signal is before the second time of sending the second reference signal by the network device, then after the first time, the first UE can send a second CSI report associated with the third port group.

[0523] After receiving the reference signal sent by the network device at time Q1, the first user equipment (UE) may or may not send a second CSI report to the network device. For example, after the second time, the first UE may or may not send a second CSI report associated with the fourth port group.

[0524] A network device may include Q port groups, where the ports in each of the Q1 port groups are distinct. Q1 can be less than or equal to Q. The network device may cyclically select port groups from the Q port groups to transmit reference signals. Q1 time intervals may represent the time taken for the network device to select the Q1 port groups with distinct ports from the Q port groups to transmit reference signals.

[0525] For the first Q-1 time intervals, the transmission of the second CSI report by the first user equipment (UE) can be understood as a cold start process. The cold start is completed after Q1 time intervals. After Q1 time intervals, the network device can also transmit reference signals. For each reference signal transmitted by the network device after Q1 time intervals, the first UE can send a first CSI report. Each first CSI report transmitted by the first UE is associated with a set of ports.

[0526] For each reference signal transmitted by the network device after Q time intervals, the first CSI report transmitted by the first UE can be determined based on the reference signals transmitted at least Q times.

[0527] The network device can also send indication information, which can instruct the first UE to report a CSI report associated with the port set.

[0528] The indication information can also indicate the port groups associated with the reference signals transmitted at each of the Q1 times. For example, the indication information can indicate the third port group associated with the first reference signal transmitted at the first time and the fourth port group associated with the second reference signal transmitted at the second time.

[0529] The indication information can also indicate the association between each port group and the port set. For example, the indication information can also indicate the association between the third port group, the fourth port group, and the port set. The association between each port group and the port set can also be understood as the ports within each port group into which the port set is divided.

[0530] For example, the port set may include N ports in a network device, and the N ports in the network device include the plurality of candidate ports. The number N ports in the network device may be greater than the number of candidate ports.

[0531] Therefore, the port associated with the reference signal sent by the network device is a candidate port, and the number of candidate ports is less than the number of ports associated with the first CSI report sent by the first UE. The network device can establish a connection with the first UE and send information based on the channel information of more ports.

[0532] Multiple candidate ports can be preset or determined by the network device. For example, the network device can determine the port with the highest importance from N ports as candidate ports based on their importance. The number of candidate ports can be preset. That is, the network device selects a preset number of ports from the N ports, sorted by importance from highest to lowest, as candidate ports. The importance of the candidate ports is higher than that of the other ports in the N ports.

[0533] By associating the transmitted reference signal with a port of higher importance, the first CSI report becomes more accurate.

[0534] The network device can configure the following information for the first UE user equipment: the reference signal resources corresponding to the port group associated with the reference signal transmitted in the Q1 time period.

[0535] Reference signal resources can include time-frequency resources, such as resource elements (REs) or resource blocks (RBs). A resource element, also called a resource particle, is the smallest unit of physical layer resources. An RE can occupy one symbol in the time domain and one subcarrier in the frequency domain. A resource block is the basic unit for resource scheduling and allocation. An RB contains multiple consecutive symbols within a time slot in the time domain and multiple consecutive subcarriers in the frequency domain.

[0536] Network devices can send information to configure reference signal resources for UEs within their coverage area. Network devices can configure reference signal resources for a first reference signal, reference signal resources for a second reference signal, and so on.

[0537] The number of port groups associated with the reference signals transmitted in the Q1 time intervals is Q1. The frequency domain resources in the reference signal resources configured by the network device for each port group can be the same or different. For example, there may be a fifth port group and a sixth port group associated with the reference signals transmitted in the Q1 time intervals. The reference signal resources corresponding to the fifth port group and the reference signal resources corresponding to the sixth port group include different frequency domain resource locations. For example, the fifth port group and the sixth port group could be the third port group and the fourth port group, respectively.

[0538] Wireless channels exhibit frequency-selective fading, meaning that channel quality can vary significantly across different frequencies. By distributing the reference signal across multiple frequency domain locations, a more comprehensive and accurate sampling of the channel state across the entire system bandwidth can be achieved.

[0539] The network device can also configure the reference signal sequence corresponding to the port group associated with the reference signal transmitted in the Q1 time for the first user equipment.

[0540] The reference signal sequences corresponding to the Q1 port groups can be the same or different. For example, among the port groups associated with the reference signals transmitted in these Q1 time intervals, there are a seventh port group and an eighth port group, and the reference signal sequence corresponding to the seventh port group is different from the reference signal sequence corresponding to the eighth port group. For example, the seventh port group and the eighth port group can be the third port group and the fourth port group, respectively.

[0541] The Q1 times belong to a time period whose duration can be less than or equal to a preset duration. In other words, the time span of the Q1 times can be less than or equal to the preset duration. For example, the time span from the first time to the second time can be less than or equal to the preset duration.

[0542] The first UE may move, causing a change in its location and resulting in channel timeout. Transmitting reference signals associated with each candidate port within a short, preset time period improves the accuracy of the first CSI report.

[0543] Network devices can transmit reference signals periodically or non-periodically. When a network device transmits reference signals periodically, the Q1 times can belong to different measurement periods. For example, the first time and the second time can be located in different measurement periods.

[0544] It should be understood that Figures 1 to 28 The flowcharts or scene diagrams shown are for illustrative purposes only and are not intended to limit the embodiments of this application to the examples illustrated. In fact, those skilled in the art can interpret the embodiments based on... Figures 1 to 28 The examples in the document can be transformed into equivalent ways to obtain more implementations.

[0545] The above text combined Figures 1 to 28 This document describes in detail the communication method provided in the embodiments of this application. The following will combine... Figures 29 to 30 The device embodiments of this application are described in detail below. It should be understood that the communication device of this application embodiment can execute the various communication methods of the foregoing embodiments of this application, that is, the specific working processes of the various products below can be referred to the corresponding processes in the foregoing method embodiments.

[0546] In the embodiments described above, the user equipment may execute some or all of the steps in each embodiment; the network device may execute some or all of the steps in each embodiment. These steps or operations are merely examples, and the embodiments of this application may also perform other operations or variations thereof. Furthermore, the steps may be executed in different orders as presented in the embodiments, and it is not necessary to execute all the operations in the embodiments of this application. Moreover, the sequence number of each step does not imply the order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0547] Figure 29 This is a schematic block diagram of a communication device provided in an embodiment of this application. Figure 29 As shown, the communication device 2600 may include a communication module 2620. The communication module 2620 can implement corresponding communication functions, which can be internal communication functions of the communication device 2600 or communication functions between the communication device 2600 and other devices. Optionally, the communication module 2620 may also be referred to as a communication interface or transceiver module. Optionally, the communication device 2600 further includes a processing module 2610. The processing module 2610 can implement corresponding processing functions.

[0548] Optionally, the communication device 2600 further includes a storage module, which can be used to store instructions and / or data; the processing module 2610 can read the instructions and / or data in the storage module so that the communication device 2600 can implement the aforementioned method embodiments.

[0549] In one possible design, the communication device 2600 may correspond to the network device in the above method embodiments, or to a component (such as a circuit, chip, or chip system) configured in a user equipment. The communication device 2600 may be used to perform the steps or processes performed by the network device in any of the above method embodiments.

[0550] For example, the communication module 2620 is configured to send a first reference signal associated with M ports of the network device.

[0551] Communication module 2620 is further configured to receive a first CSI report sent by a first user equipment, wherein the first CSI report is determined by the first user equipment based on the first reference signal, and the first CSI report is associated with N ports in the network device, where N is greater than M, and M and N are both integers greater than 1.

[0552] For example, the processing module 2610 is used to acquire downlink channel state information of multiple user equipments. The downlink channel state information of each user equipment includes the complex channel gain parameters of multiple ports among N ports of the network device. Different ports in the network device correspond to different beam directions, and N is a positive integer greater than 1.

[0553] The communication module 2620 is used to process the downlink channel state information of multiple user equipments through a port model to obtain the importance of each of the N ports.

[0554] The processing module 2610 is used to determine multiple candidate ports from the N ports according to the importance of each port, wherein the importance of each candidate port is higher than the importance of other ports in the N ports besides the multiple candidate ports.

[0555] The communication module 2620 is also used to send reference signals through the plurality of candidate ports, wherein the reference signals sent by any two candidate ports are orthogonal.

[0556] The communication module 2620 is further configured to receive channel information sent by the first user equipment, the channel information being determined by the first user equipment based on the complex channel gain parameters of the port in the network device that transmits reference signals.

[0557] The processing module 2610 is used to determine the downlink channel state information of the first user equipment based on the channel information.

[0558] In another possible design, the communication device 2600 may correspond to the first UE in the above method embodiments, or to a component (such as a circuit, chip, or chip system) configured in a network device. The communication device 2600 may be used to perform the steps or processes performed by the first UE in any of the above method embodiments.

[0559] For example, the communication module 2620 is configured to receive a first reference signal sent by a network device, the first reference signal being associated with M ports of the network device, where M is an integer greater than 1.

[0560] Communication module 2620 is further configured to send the first CSI report to the network device, wherein the first CSI report is associated with N ports in the network device, and N is an integer greater than M.

[0561] The above are merely examples; for detailed steps or procedures, please refer to the descriptions in the foregoing embodiments.

[0562] The term "unit" here can be implemented in software and / or hardware, without specific limitation. For example, a "unit" can be a software program, hardware circuit, or a combination of both that implements the above-described functions. The hardware circuit may include application-specific integrated circuits (ASICs), electronic circuits, processors (e.g., shared processors, proprietary processors, or group processors) and memory for executing one or more software or firmware programs, combined logic circuits, and / or other suitable components that support the described functions.

[0563] Therefore, the units of the various examples described in the embodiments of this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented 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 this application.

[0564] Figure 30 This is another schematic block diagram of the communication device 2700 provided in the embodiments of this application. The communication device 2700 may be a chip, chip system, or processor, etc., used by a user equipment or network device to implement the above-described methods. The communication device 2700 can be used to implement the methods described in the above-described method embodiments; for details, please refer to the descriptions in the above-described method embodiments.

[0565] like Figure 30 As shown, the communication device 2700 may include one or more processors 2710, which may also be referred to as processing units or processing modules, and can implement certain control functions. The processor 2710 may be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, while the central processing unit can be used to control the communication device 2700 (e.g., a base station, baseband chip, user, user chip), execute software programs, and process data from the software programs.

[0566] In an alternative design, the processor 2710 may also store instructions and / or data that can be executed by the processor 2710 to cause the communication device 2700 to perform the methods described in the above method embodiments.

[0567] In another alternative design, the communication device 2700 may include a communication interface 2720 for implementing receiving and transmitting functions. For example, the communication interface 2720 may be a transceiver circuit, interface, interface circuit, or transceiver. The transceiver circuit, interface, interface circuit, or transceiver for implementing receiving and transmitting functions may be separate or integrated. The aforementioned transceiver circuit, interface, interface circuit, or transceiver may be used for reading and writing code / data, or it may be used for transmitting or relaying signals.

[0568] Optionally, the communication device 2700 may include one or more memories 2730, which may store instructions that can be executed on the processor 2710, causing the communication device 2700 to perform the methods described in the above method embodiments. Optionally, the memories 2730 may also store data. Optionally, the processor 2710 may also store instructions and / or data. The processor 2710 and the memories 2730 may be provided separately or integrated together.

[0569] It should be understood that, in one possible design, the steps in the method embodiments provided in this application can be implemented by integrated logic circuits in the processor's hardware or by instructions in software form. The steps of the methods disclosed in the embodiments of this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are not provided here.

[0570] In one implementation, the communication device 2700 may correspond to the user equipment in the above method embodiments and may be used to execute the various steps and / or processes executed by the user equipment in the above method embodiments. The processor 2710 may be used to execute instructions stored in the memory 2730, and when the processor 2710 executes the instructions stored in the memory, the processor 2710 is used to execute the various steps and / or processes of the above method embodiments corresponding to the user equipment.

[0571] In another implementation, the communication device 2700 may correspond to the network device in the above method embodiments and may be used to execute the various steps and / or processes executed by the network device in the above method embodiments. The processor 2710 may be used to execute instructions stored in the memory 2730, and when the processor 2710 executes the instructions stored in the memory, the processor 2710 is used to execute the various steps and / or processes of the above method embodiments corresponding to the network device.

[0572] It should be understood that the aforementioned processing device can be one or more chips. For example, the processing device can be a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a system-on-chip (SoC), a central processor unit (CPU), a network processor (NP), a digital signal processor (DSP), a microcontroller unit (MCU), a programmable logic device (PLD), or other integrated chips.

[0573] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0574] According to the method provided in the embodiments of this application, this application also provides a chip system, which includes one or more processors for calling and executing instructions stored in memory, thereby causing the method described in the embodiments of this application to be executed. The chip system may be composed of chips or may include chips and other discrete devices.

[0575] The chip system may include input circuits or interfaces for transmitting information or data, and output circuits or interfaces for receiving information or data.

[0576] According to the method provided in the embodiments of this application, this application also provides a communication system, which includes the aforementioned network device and user equipment.

[0577] According to the method provided in the embodiments of this application, this application also provides a computer program product, which includes: computer program code, which, when run on a computer, causes the computer to execute the various steps or processes executed by the network device or user equipment in any of the foregoing method embodiments.

[0578] According to the method provided in the embodiments of this application, this application also provides a computer-readable storage medium storing program code, which, when run on a computer, causes the computer to execute the various steps or processes executed by the network device or user equipment in any of the foregoing method embodiments.

[0579] The computer-readable storage medium may be the aforementioned volatile memory or non-volatile memory, or it may include both volatile memory and non-volatile memory.

[0580] In the embodiments of this application, the terms and English abbreviations are exemplary examples given for ease of description and should not be construed as limiting the application in any way. This application does not preclude the possibility of defining other terms that can achieve the same or similar functions in existing or future agreements.

[0581] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When these computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated.

[0582] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0583] In the description of this application, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance, or a specific order or sequence. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0584] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0585] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented 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 this application.

[0586] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0587] In summary, the above description is merely a preferred embodiment of the technical solution of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A communication method, characterized in that, Applied to network devices, the method includes: Send a first reference signal, which is associated with M ports of the network device; The system receives a first channel state information report sent by a first user equipment. The first channel state information report is determined by the first user equipment based on the first reference signal. The first channel state information report is associated with N ports in the network device, where N is greater than M, and M and N are both integers greater than 1.

2. The method according to claim 1, characterized in that, The transmission of the first reference signal includes: Reference signals are transmitted at multiple times. The reference signal transmitted at time t is associated with a first port group, and the reference signal transmitted at time t-1 is associated with a second port group. The ports in the first port group are different from the ports in the second port group. The M ports are the ports in the first port group. The first channel state information report is determined by the first user equipment based on the reference signals at the multiple times, where t is an integer greater than 1.

3. The method according to claim 2, characterized in that, The port groups associated with the reference signals sent by the network device at the multiple times are port groups that the network device cyclically selects from Q port groups, where Q is an integer greater than 1.

4. The method according to claim 3, characterized in that, The method further includes: Sending instruction information, the instruction information instructing the first user equipment to send a channel state information report based on the reference signals of the plurality of times.

5. The method according to claim 3 or 4, characterized in that, The length of Q of the multiple times is less than a preset duration, and the port groups associated with the reference signals sent by the network device at the Q times respectively include the Q port groups.

6. The method according to claim 3 or 4, characterized in that, The Q ports include multiple candidate ports, and the importance of the candidate ports is higher than that of the other ports among the N ports besides the multiple candidate ports.

7. The method according to claim 6, characterized in that, The method further includes: Based on the channel status information reports from multiple user equipment, a second order is determined for the N ports, ranked according to their importance.

8. The method according to claim 7, characterized in that, The step of determining the second order of the N ports according to their importance based on channel state information reports from multiple user equipment includes: Based on the channel state information reports from the multiple user equipment, the importance of each of the N ports is determined using the port model. When a second port and a third port exist among the multiple candidate ports, the number of port groups including the second port is greater than the number of port groups including the third port in the Q port groups, the first importance of the second port is greater than the second importance of the third port, and the difference between the first importance and the second importance is greater than or equal to a preset importance difference threshold.

9. The method according to claim 8, characterized in that, The channel state information report of any one of the multiple user equipments represents the downlink channel state information of that user equipment, and the downlink channel state information of that user equipment includes the complex channel gain parameters of multiple ports among the N ports; The port model is trained based on second training data, which includes channel information samples and importance order labels. The channel information samples include training downlink channel state information of multiple training user equipments. The importance order labels represent the importance ranking results of N1 third training ports among the N third training ports. The training downlink channel state information of any one of the multiple training user equipments includes the complex channel gain parameters of the N1 third training ports among the N third training ports of the second training network device determined by the any one training user equipment. The N third training ports are the same as the N ports, and N1 is an integer greater than 1.

10. The method according to claim 9, characterized in that, The port model is obtained by adjusting the parameters of the initial port model based on the loss value. The initial port model is used to process the channel information samples to obtain the training importance value of each third training port. The loss value is positively correlated with the difference parameter of each of the N1 third training ports except for the third training port with the lowest importance. The difference parameter of any third training port except for the third training port with the lowest importance is the sum of at least one sub-difference of that third training port. The i-th sub-difference of any third training port is based on the training importance value of that third training port and the sum of the differences between the two sub-differences. The importance value of the i-th target training port corresponding to any third training port is determined. In the importance order label, the importance of the i-th target training port corresponding to any third training port is lower than the importance of any third training port. If the first difference obtained by subtracting the preset difference value from the training importance value of any third training port is greater than or equal to the i-th training importance value, the i-th sub-difference of any third training port is 0. If the first difference is less than the i-th training importance value, the i-th sub-difference of any third training port is the second difference obtained by subtracting the first difference from the i-th training importance value. The preset difference value is greater than 0.

11. The method according to claim 10, characterized in that, The loss value is obtained by weighted summation of the difference parameters of each of the N1 third training ports except for the third training port with the lowest importance. The weight of the difference parameter of any third training port is positively correlated with the importance of any third training port represented by the importance order label.

12. The method according to any one of claims 3, 4, 7-11, characterized in that, The number of first ports is greater than the number of second ports. The number of first ports represents the number of ports in each of the Q port groups when a preset condition is met. The number of second ports represents the number of ports in each of the Q port groups when the preset condition is not met. The preset condition is that the data transmission volume of the network device is less than a preset transmission volume threshold, and / or the movement speed of the user equipment within the coverage area of ​​the network device is greater than or equal to a preset speed threshold.

13. The method according to any one of claims 3, 4, 7-11, characterized in that, The first channel state information report includes channel response features, which are obtained by feature extraction from a set of channel information, and the set of channel information includes complex channel gain parameters of K ports out of the M ports; The method further includes: processing the channel response features through a first prediction model to obtain downlink channel state information of the first user equipment, wherein the downlink channel state information of the first user equipment includes complex channel gain parameters of N ports in the network device.

14. The method according to claim 13, characterized in that, The channel response features are obtained by the first user equipment from the sorted channel information set, wherein the complex channel gain parameters of the multiple target ports in the sorted channel information set are arranged according to their amplitude. The method further includes: Receive port information sent by the first user equipment, wherein the port information represents the target port corresponding to each port channel feature among the multiple port channel features of the channel response feature; Based on the port information and the channel response characteristics, the expanded channel response characteristics are determined. In the expanded channel response characteristics, the port characteristics of the N ports are arranged according to the beam direction corresponding to the port. The port characteristics of the target port are the port channel characteristics of the target port in the channel response characteristics, and the port characteristics of other ports besides the target port are preset values. The step of processing the channel response features using a first prediction model to obtain the downlink channel state information of the first user equipment includes: The expanded channel response features are processed using the first prediction model to obtain the downlink channel state information of the first user equipment.

15. A communication method, characterized in that, Applied to a first user equipment, the method includes: Receive a first reference signal sent by a network device, wherein the first reference signal is associated with M ports of the network device, where M is an integer greater than 1; The first channel status information report is sent to the network device. The first channel status information report is associated with N ports in the network device, where N is an integer greater than M.

16. The method according to claim 15, characterized in that, The method further includes: The system receives indication information sent by the network device. The indication information is used to instruct the first user equipment to report channel state information associated with the N ports based on reference signals at multiple times. The indication information is also used to indicate the port groups associated with the reference signals sent by the network device at the multiple times, wherein the reference signal sent by the network device at time t is associated with the first port group, the reference signal sent by the network device at time t-1 is associated with the second port group, the ports in the first port group are different from the ports in the second port group, and the M ports are the ports in the first port group. The first reference signal sent by the receiving network device includes: Receive reference signals sent by the network device at multiple times.

17. The method according to claim 15, characterized in that, The first reference signal received by the network device includes: Receive reference signals sent by the network device at multiple times; The method further includes: Based on the reference signals at the plurality of times, channel response information for each time is determined. The channel response information for any time among the plurality of times includes the complex channel gain parameter of each port in the port group associated with the reference signal transmitted by the network device at that time. The t-th time when the first reference signal is transmitted is the latest time among the plurality of times. Based on the channel response information at each time, a channel information set is determined. The channel information set includes the complex channel gain parameters of the M ports determined based on the first reference signal, and the complex channel gain parameters of the first port among the N ports. The port group associated with the reference signals at multiple times includes the first port, and the first port does not belong to the M ports. The first channel state information report is determined based on the channel information set.

18. The method according to claim 17, characterized in that, The different times mentioned belong to different measurement periods. The step of determining the channel information set based on the channel response information at each time period includes: Using the second prediction model, channel response prediction information for the t-th measurement period is determined based on the channel response information of the (t-1)-th measurement period. The channel response prediction information for the t-th measurement period includes the complex channel gain parameters of P ports out of the N ports. The P ports include the first port. The channel information set includes the complex channel gain parameters of the first port in the channel response prediction information for the t-th measurement period. The ports associated with the reference signals sent by the network device in the (t-1)-th measurement periods prior to the t-th measurement period include the P ports.

19. The method according to claim 18, characterized in that, The step of determining the channel response prediction information for the t-th measurement period using the second prediction model, based on the channel response information for the (t-1)-th measurement period, includes: The channel response prediction information for the (t-1)th measurement period and the channel response information for the (t-1)th measurement period are processed by the second prediction model to obtain the channel response prediction information for the (t-1)th measurement period. The channel response prediction information for the (t-1)th measurement period is determined by the second prediction model based on the channel response information for the (t-2)th measurement period, where t is greater than 2.

20. The method according to claim 18 or 19, characterized in that, The method further includes: The channel information set is subjected to feature extraction using a feature extraction model to obtain channel response features. The channel information set includes the complex channel gain parameters of K ports out of the M ports. The first channel state information report includes the channel response features. The channel response features are used by the network device to process the data through a first prediction model to obtain the downlink channel state information of the first user equipment. The downlink channel state information of the first user equipment includes the complex channel gain parameters of N ports in the network device.

21. The method according to claim 20, characterized in that, The feature extraction model and the first prediction model are obtained by adjusting the parameters of the initial feature extraction model and the first initial prediction model respectively based on the difference between the training channel state and the channel state label. The training channel state is obtained by processing the training channel response features through the first initial prediction model. The training channel response features are obtained by processing the training channel information set through the initial feature extraction model. The training channel information set includes first training channel response information, which is information determined by the first training user equipment based on the first training reference signal sent by the first training network equipment. The first training reference signal is associated with M' first training ports of the first training user equipment. The first training channel response information includes the complex channel gain parameters of K' first training ports among the M' first training ports. The channel state label includes the complex channel gain parameters of N first training ports of the first training network equipment. The N first training ports include the M' first training ports, where M' and K' are both positive integers greater than 1, and the N first training ports are the same as the N ports.

22. The method according to claim 21, characterized in that, The second prediction model is obtained by adjusting the parameters of the second initial prediction model based on the difference. The first training channel response information is the training channel response information of the m-th training measurement period. The training channel information set also includes the complex channel gain parameter of the target first training port in the training channel response prediction information of the m-th training measurement period. The training channel response prediction information of the m-th training measurement period is determined by the second initial prediction model based on the training channel response information of the (m-1)-th training measurement period. The channel response prediction information of the m-th measurement period includes the complex channel gain parameter of P' first training ports among the N first training ports. The target first training port belongs to the P' first training ports, and the first training port does not belong to the M' first training ports, where m is an integer greater than 1.

23. The method according to claim 20, characterized in that, The method further includes: The complex channel gain parameters of multiple target ports in the channel information set are sorted according to the magnitude of the amplitude to obtain the sorted channel information set, wherein the multiple target ports include the K ports; The process of extracting features from the channel information set using a feature extraction model to obtain channel response features includes: The channel response features are obtained by extracting features from the sorted channel information set using the feature extraction model. The channel response features include the port channel features of each target port. The method further includes: The network device sends port information, which is used to indicate the target port corresponding to each port channel feature in the channel response features.

24. The method according to claim 20, characterized in that, The K ports are the same as the M ports.

25. A communication method, characterized in that, Applied to network devices, the method includes: The downlink channel state information of multiple user equipments is obtained. The downlink channel state information of each user equipment includes the complex channel gain parameters of multiple ports among the N ports of the network device. Different ports in the network device correspond to different beam directions, and N is a positive integer greater than 1. By using a port model, the downlink channel state information of multiple user equipments is processed to obtain the importance of each of the N ports; Based on the importance of each port, multiple candidate ports are determined from the N ports, and the importance of each candidate port is higher than that of the other ports in the N ports besides the multiple candidate ports; Reference signals are transmitted through the multiple candidate ports, and the reference signals transmitted by any two candidate ports are orthogonal. The system receives channel information sent by a first user equipment, the channel information being determined by the first user equipment based on the complex channel gain parameters of the port in the network device that transmits reference signals. Based on the channel information, the downlink channel state information of the first user equipment is determined.

26. A communication system, characterized in that, It includes the network device and the first user equipment as described in any one of claims 1-24, or it includes the network device and the first user equipment as described in claim 25.

27. An electronic device, characterized in that, The device includes at least one processor coupled to a memory storing a program or instructions, the processor executing the program or instructions to cause the electronic device to perform the method as claimed in any one of claims 1 to 14, or the method as claimed in any one of claims 15 to 24, or the method as claimed in claim 25.

28. A chip system, characterized in that, The chip system includes one or more processors, which are configured to retrieve and execute instructions stored in memory, such that the method as claimed in any one of claims 1 to 14 is executed, or the method as claimed in any one of claims 15 to 24 is executed, or the method as claimed in claim 25 is executed.

29. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program for implementing the method of any one of claims 1 to 14, or the method of any one of claims 15 to 24, or the method of claim 25.

30. A computer program product, characterized in that, When the computer program product is run on an electronic device, the electronic device is used to perform the method of any one of claims 1 to 14, or the method of any one of claims 15 to 24, or the method of claim 25.

Images