CSI feedback method, CSI recovery method, device, and medium

Abstract

A channel state information (CSI) feedback method, a CSI recovery method, a device, and a medium. The CSI feedback method comprises: receiving first indication information sent by a network device, the first indication information being used for indicating a first frequency domain resource; on the basis of a frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, determining a target feedback mode for CSI feedback, the second frequency domain resources being frequency domain resources corresponding to a model on a terminal side, and the at least one group of second frequency domain resources corresponding to the model on the terminal side; on the basis of the target feedback mode, determining CSI feedback information fed back to the network device; and sending the CSI feedback information to the network device.

Description

CSI feedback method, CSI recovery method, device and medium

[0001] The present disclosure claims priority to the Chinese patent application No. 202411384047.2, filed on September 30, 2024, and entitled "CSI feedback method, CSI recovery method, device and medium", the whole content of which is incorporated herein by reference. TECHNICAL FIELD

[0002] The present disclosure relates to the technical field of communication, and in particular to a CSI feedback method, a CSI recovery method, a device and a medium. BACKGROUND

[0003] Channel state information (CSI) feedback based on artificial intelligence (AI) or machine learning (ML) is one of the important applications of AI / ML technology in the physical layer, aiming to replace channel prediction or codebook-based CSI feedback through inference of an AI / ML model, so as to improve the accuracy of CSI feedback and thus improve system performance.

[0004] An AI / ML model (hereinafter referred to as an AI model) is a data-driven algorithm that needs to be trained through a suitable data set first, and then can infer or predict a specific form of output according to a specific form of input. For a two-end model, i.e., the AI model is deployed on the base station side and the terminal side, the model on the base station side needs to match the model on the terminal side, otherwise the model cannot work; and the input of the terminal side model is the CSI on a fixed number of subbands or physical resource blocks (PRBs), and the output of the base station side model is the recovered CSI on these subbands / PRBs, once the input of the terminal side model changes, the model has the risk of not working or performance degradation. That is, the current AI model-based CSI feedback is limited, resulting in inflexible CSI feedback. SUMMARY

[0005] The present disclosure aims to provide a CSI feedback method, a CSI recovery method, a device and a medium to solve the problem of limited model-based CSI feedback in related technologies, resulting in inflexible CSI feedback.

[0006] To achieve the above-mentioned purpose, in a first aspect, the present disclosure provides a CSI feedback method applied to a terminal, comprising:

[0007] receiving first indication information sent by a network device, the first indication information being used to indicate first frequency domain resources;

[0008] determine a target feedback mode of channel state information (CSI) feedback based on a frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resources being frequency domain resources corresponding to a model at a terminal side, and at least one group of the second frequency domain resources corresponding to one model at the terminal side;

[0009] determine CSI feedback information to be fed back to the network device based on the target feedback mode;

[0010] send the CSI feedback information to the network device.

[0011] In a second aspect, the embodiments of the present disclosure further provide a CSI recovery method applied to a network device, comprising:

[0012] send first indication information to a terminal, the first indication information being used to indicate a first frequency domain resource, and the first frequency domain resource being used for channel state information (CSI) feedback of the terminal;

[0013] receive CSI feedback information sent by the terminal;

[0014] recover the CSI according to the CSI feedback information;

[0015] The CSI feedback information is obtained by the terminal based on a target feedback mode, and the target feedback mode is determined by the terminal based on a frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resources being frequency domain resources corresponding to a model at a terminal side, and at least one group of the second frequency domain resources corresponding to one model at the terminal side.

[0016] In a third aspect, the embodiments of the present disclosure further provide a terminal, comprising a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the transceiver is used to receive first indication information sent by a network device, the first indication information being used to indicate a first frequency domain resource;

[0017] The processor is used to read the program in the memory and execute the following processes:

[0018] determine a target feedback mode of channel state information (CSI) feedback based on a frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resources being frequency domain resources corresponding to a model at a terminal side, and at least one group of the second frequency domain resources corresponding to one model at the terminal side;

[0019] determine CSI feedback information to be fed back to the network device based on the target feedback mode;

[0020] The transceiver is further configured to send the CSI feedback information to the network device.

[0021] In a fourth aspect, the embodiments of the present disclosure further provide a terminal, comprising:

[0022] A first receiving unit is configured to receive first indication information sent by a network device, the first indication information being used to indicate first frequency domain resources;

[0023] A first processing unit is configured to determine a target feedback mode of channel state information (CSI) feedback based on a frequency domain location relationship between the first frequency domain resources and at least one group of second frequency domain resources, the second frequency domain resources being frequency domain resources corresponding to a terminal-side model, and at least one group of the second frequency domain resources corresponding to one terminal-side model.

[0024] A second processing unit is configured to determine CSI feedback information to be fed back to the network device based on the target feedback mode.

[0025] A first sending unit is configured to send the CSI feedback information to the network device.

[0026] In a fifth aspect, the embodiments of the present disclosure further provide a network device, comprising a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; the transceiver is configured to send first indication information to a terminal, the first indication information being used to indicate first frequency domain resources, the first frequency domain resources being used for channel state information (CSI) feedback of the terminal;

[0027] receive CSI feedback information sent by the terminal;

[0028] The processor is configured to read the program in the memory and perform the following process:

[0029] recover CSI according to the CSI feedback information;

[0030] The CSI feedback information is obtained by the terminal based on a target feedback mode; the target feedback mode is determined by the terminal based on a frequency domain location relationship between the first frequency domain resources and at least one group of second frequency domain resources, the second frequency domain resources being frequency domain resources corresponding to a terminal-side model, and at least one group of the second frequency domain resources corresponding to one terminal-side model.

[0031] In a sixth aspect, the embodiments of the present disclosure further provide a network device, comprising:

[0032] A second sending unit is configured to send first indication information to a terminal, the first indication information being used to indicate first frequency domain resources, the first frequency domain resources being used for channel state information (CSI) feedback of the terminal;

[0033] a second receiving unit, configured to receive CSI feedback information sent by the terminal;

[0034] a third processing unit, configured to recover the CSI according to the CSI feedback information;

[0035] The CSI feedback information is obtained by the terminal based on a target feedback mode, and the target feedback mode is determined by the terminal based on a frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources. The second frequency domain resource is a frequency domain resource corresponding to a terminal-side model, and at least one group of the second frequency domain resources corresponds to one terminal-side model.

[0036] In a seventh aspect, the embodiments of the present disclosure further provide a non-transitory readable storage medium, wherein the non-transitory readable storage medium stores a program, and the program is used to execute the CSI feedback method in the first aspect or execute the CSI recovery method in the second aspect.

[0037] The above technical solutions of the present disclosure have at least the following beneficial effects:

[0038] In the above technical solutions of the embodiments of the present disclosure, first, the first indication information sent by the network device is received, and the first indication information is used to indicate the first frequency domain resource. Then, based on the frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, the target feedback mode of channel state information (CSI) feedback is determined, the second frequency domain resource is a frequency domain resource corresponding to a terminal-side model, and at least one group of the second frequency domain resources corresponds to one terminal-side model. Then, based on the target feedback mode, the CSI feedback information fed back to the network device is determined. Finally, the CSI feedback information is sent to the network device. In this way, the terminal determines the actual feedback mode to be executed based on the indication of the network device and the model of the terminal itself, the feedback mode is flexible, and the flexibility of the CSI feedback can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0039] FIG. 1 is a flow diagram of a CSI feedback method according to an embodiment of the present disclosure;

[0040] FIG. 2 is a diagram of a CSI feedback step according to an embodiment of the present disclosure;

[0041] FIG. 3 is a diagram of a CSI feedback step according to an embodiment of the present disclosure;

[0042] FIG. 4 is a diagram of a CSI feedback step according to an embodiment of the present disclosure;

[0043] FIG. 5 is a diagram of a CSI feedback step according to an embodiment of the present disclosure;

[0044] FIG. 6 is a diagram of a CSI feedback step according to an embodiment of the present disclosure;

[0045] FIG. 7 is a flow diagram of a CSI recovery method according to an embodiment of the present disclosure;

[0046] FIG. 8 is a block diagram of a terminal according to an embodiment of the present disclosure;

[0047] FIG. 9 is a block diagram of a terminal according to an embodiment of the present disclosure;

[0048] FIG. 10 is a block diagram of a network device according to an embodiment of the present disclosure;

[0049] FIG. 11 is a block diagram of a network device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present disclosure.

[0051] In the embodiments of the present disclosure, the term "and / or" describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the following three cases: A exists alone, A and B exist together, and B exists alone. The character " / " generally represents an "or" relationship between the associated objects before and after it.

[0052] In the embodiments of the present disclosure, the term "multiple" refers to two or more, and other quantifiers are similar.

[0053] The content related to the scheme provided by the embodiments of the present disclosure will be introduced first.

[0054] Currently in the research of 3rd Generation Partnership Project (3GPP), the CSI compression uses a two-end AI model, which is deployed at the base station side and the terminal side respectively. The input of the terminal side model is usually the CSI on a fixed number of subbands / PRBs. These subbands / PRBs can be continuous or have a fixed interval. The input of the base station side model is the CSI recovered on these subbands / PRBs. If the input of the terminal side model changes, the model may not work or the performance will decrease. In addition, when training an AI model, the data of the training set can only reflect the data characteristics under limited change conditions. To make the AI model learn more complex features, on the one hand, the model complexity will increase, thereby increasing the burden of device computing power and slowing down the inference speed; on the other hand, more training data may be needed, making the model training more difficult. Therefore, in practical applications, it is difficult to avoid the situation that the data characteristics of the model inference stage are different from those of the training stage, resulting in a decrease in model performance.

[0055] The existing codebook feedback-based algorithm has obvious advantages over the AI model in the above two problems, i.e., the input and output forms of the algorithm and the conditions used by the algorithm. Moreover, it does not involve the matching problem of the two-end model. For example, the R16 eType II codebook can flexibly adjust the number of subbands and the size of subbands that need to be fed back, and can also work normally when the deployment scene (such as Urban Macro (UMa), Urban Micro (UMi), carrier frequency, terminal moving speed, and other conditions change.

[0056] Currently, there is a standardized method for the base station to instruct the terminal to perform CSI feedback on which subbands. The currently supported subband size values are related to the bandwidth part (BandWidth Part, BWP) configured by the base station. The base station can select one value from two optional values as the subband size for the terminal to feed back the CSI according to the number of PRBs contained in each BWP configured for the terminal. Each subband is continuous PRBs, i.e., the subband size in the following table Table 5.2.1.4-2. The i-th BWP is divided into several subbands according to the subband size, denoted as represents the starting PRB number of the BWP, represents the number of PRBs contained in the BWP, and the size of each subband is as follows:

[0057] The size of the first subband is

[0058] The size of the last subband is:

[0059] When , take Otherwise take

[0060] The remaining subband size is

[0061] Table 5.2.1.4-2: Configurable subband sizes

[0062] The base station indicates which subbands the UE needs to feedback on through the higher layer parameters csi-ReportingBand and subbandSize. Among them, subbandSize represents the subband size mentioned above; csi-ReportingBand is a 3-19 bit string, each bit indicates whether a subband needs to be fed back.

[0063] In summary, under the existing mechanism, the introduction of AI model-based CSI feedback may have the following problems:

[0064] 1. The existing mechanism does not have steps to verify the model matching of the terminal and the base station.

[0065] 2. The base station only configures the subbands that need to be fed back according to the existing mechanism, which may cause some AI models to be unable to work or even no available models on the terminal side. Specifically:

[0066] a) The input and output of the AI model (such as the terminal-side model in the double-end CSI compression model) may conflict with the existing subband size, for example, the AI model intelligently outputs the CSI of the subband size of 8 PRB, but the base station configures the subband size of 4 PRB.

[0067] b) The AI model input and output do not match the number of subbands configured by the base station. For example, the terminal-side AI model intelligently compresses the CSI on 12 subbands (48 PRBs), but the base station configures the number of subbands to be 13 subbands (52 PRBs), at this time, the remaining 1 CSI information cannot be obtained through the AI model.

[0068] c) The AI model performs poorly on the subbands configured by the base station. For example, the base station configures the terminal to feed back on 12 subbands except the 4th subband, but the AI model only has good performance when inputting 12 adjacent or equally spaced subbands, then the terminal cannot complete the CSI feedback through the AI model.

[0069] 3. The terminal needs to fall back to the codebook feedback mechanism when it cannot use the AI model to complete the CSI feedback. The reason why the model cannot be used may be:

[0070] a) the current inference condition is too different from the model training set, and the model has insufficient generalization ability (e.g., less than the codebook feedback accuracy under similar overhead)

[0071] b) the current device has insufficient computing power.

[0072] In summary, the CSI feedback based on the AI model is currently limited, resulting in the problem of inflexible CSI feedback.

[0073] To solve the above technical problems, the present disclosure provides a CSI feedback method, a CSI recovery method, a device and a medium, wherein the method and the device are based on the same application concept. Since the principles of the method and the device for solving problems are similar, the implementation of the device and the method can be mutually referred to, and the repeated parts will not be described here.

[0074] As shown in FIG. 1, it is a flowchart of the CSI feedback method provided by the present disclosure. The method is applied to a terminal, that is, it is executed by the terminal. The method specifically includes:

[0075] Step 101, receiving first indication information sent by a network device, wherein the first indication information is used to indicate a first frequency domain resource.

[0076] In some embodiments, the network device is a base station. The first frequency domain resource is the frequency domain resource required by the base station to instruct the terminal to perform CSI feedback, that is, the terminal needs to perform CSI feedback on which frequency domain resources (such as subbands / PRBs).

[0077] It should be noted that before receiving the first indication information sent by the network device, the terminal and the network device can perform information interaction of both models. On the one hand, it provides a basis for determining the first indication information for the network device; on the other hand, it provides a basis for the terminal to use which model or models for CSI feedback. The implementation process of the terminal and the network device performing information interaction of both models can be illustrated by the following embodiments.

[0078] In some embodiments, before step 101, the method of the present disclosure further includes:

[0079] receiving first information related to the model on the network device side sent by the network device; and determining a model on the terminal side matched with the model on the network device side according to the first information.

[0080] In some embodiments, the first information includes a model identification ID, a training data set of the model, an ID of the training data set of the model, model transmission information, or designated auxiliary information. Of course, depending on the specific circumstances, the first information can also include other information, which is not limited here.

[0081] In some embodiments, the first information is sent by the network device when the terminal accesses, or the first information is sent by the network device at a time of transmitting high-layer signaling.

[0082] In this embodiment, the terminal determines the terminal-side model matched with the network-device-side model according to the first information sent by the network device, that is, determines which model or models match the network-device-side model, so as to subsequently perform CSI feedback based on the model, thereby improving the CSI feedback accuracy.

[0083] In some embodiments, before step 101, the method of the present disclosure further comprises:

[0084] sending, to the network device, second information related to the terminal-side model, the second information being used to determine the terminal-side model matched with the network-device-side model.

[0085] In some embodiments, the second information includes a model identification ID, a training data set of the model, an ID of the training data set of the model, model transmission information, or designated auxiliary information. Of course, the second information can also include other information according to specific circumstances, which is not limited here.

[0086] In some embodiments, the second information is sent by the terminal when the terminal accesses (specifically, the second information can be carried in the terminal capability information reported by the terminal), or the second information is sent by the terminal after receiving the high-layer signaling sent by the network device (specifically, the high-layer information can indicate that the terminal reports the second information).

[0087] The second information can also include the frequency domain resource corresponding to the terminal-side model.

[0088] Here, the matching of the terminal-side model with the network-device-side model means that the network-device-side model can infer and recover the CSI information based on the output of the terminal-side model.

[0089] In this embodiment, the network device determines which model or models of the terminal side match the network-device-side model according to the second information sent by the terminal, and can also know the frequency domain resource corresponding to the terminal-side model matched with the network-device-side model, so that the network device can determine how to configure the frequency domain resource (such as sub-band / PRB) that needs to be fed back based on the above-obtained information and in combination with its own needs (such as the CSI of which sub-band is needed, what kind of frequency domain configuration is needed for the input of the network-device-side model), that is, determine the first indication information, so that the terminal can more easily perform inference or prediction based on the model, thereby improving the CSI feedback accuracy.

[0090] At step 102, a target feedback mode of channel state information (CSI) feedback is determined based on a frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resources being frequency domain resources corresponding to a model at a terminal side, and the at least one group of second frequency domain resources corresponding to one model at the terminal side.

[0091] The model in the embodiments of the present disclosure can be an AI model, an AI / ML model, an encoder (at the network device side), a decoder (at the terminal side), a CSI generation part (at the terminal side), a CSI reconstruction part (at the network device side), and the like. These can be collectively referred to as a model.

[0092] The frequency domain resource corresponding to the model refers to that, when the CSI on the frequency domain resource is input, the model can complete one inference or prediction.

[0093] It should be understood that different frequency domain location relationships correspond to different target feedback modes of CSI feedback. In some embodiments, the target feedback mode includes one of the following modes:

[0094] Mode one: performing one model inference or prediction to obtain first CSI on the first frequency domain resource, and feeding back the first CSI;

[0095] Here, mode one corresponds to a case where the terminal performs one model inference or prediction without additional processing. That is, the terminal can obtain the first CSI on the first frequency domain resource (i.e., the frequency domain resource required by the network device) by performing one model inference or prediction. It should be understood that in this case, the second frequency domain resource corresponding to the model for inference or prediction by the terminal is exactly the same as the first frequency domain resource (i.e., the two groups of frequency domain resources include the same subbands / PRBs).

[0096] Mode two: performing one model inference or prediction to obtain second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the second CSI and the frequency domain location corresponding to the third frequency domain resource, or feeding back the second CSI and the frequency domain location corresponding to the second frequency domain resource;

[0097] Mode three: performing one model inference or prediction to obtain second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the first CSI;

[0098] Here, mode two and mode three correspond to a case where the terminal performs one model inference or prediction, but needs to solve the problem that the CSI output by the model corresponds to the third frequency domain resource other than the first frequency domain resource indicated by the network device.

[0099] Specifically, if the second mode is adopted, the network device side needs to handle the extra processing, that is, after the terminal feeds back the second CSI and the frequency domain position corresponding to the third frequency domain resource (i.e., the redundant frequency domain resource), the network device side deletes the CSI on the third frequency domain resource (i.e., the CSI on the third frequency domain resource) after recovering the CSI, and finally obtains the first CSI on the first frequency domain resource.

[0100] If the third mode is adopted, the terminal side handles the extra processing, that is, the terminal distinguishes the first CSI on the first frequency domain resource (the first CSI corresponding to the first frequency domain resource) and the CSI on the third frequency domain resource (the CSI corresponding to the third frequency domain resource) from the second CSI, and then feeds back only the first CSI, that is, the network device side does not need other processing.

[0101] The fourth mode: performing multiple model inferences or predictions to obtain the first CSI, and feeding back the first CSI.

[0102] Here, the fourth mode corresponds to the case where the terminal needs to perform multiple model inferences or predictions to cover the frequency domain resources indicated by the network device side.

[0103] The fifth mode: after the model input is processed by an algorithm, the model performs inference or prediction on one or more groups of the second frequency domain resource, and after the obtained model output is processed by an algorithm, the third CSI is obtained and fed back; wherein the third CSI includes the first CSI.

[0104] Here, the fifth mode corresponds to the case where the frequency domain resources indicated by the network device side do not match the frequency domain resources corresponding to the terminal side model, and some additional calculation processing enables the terminal to complete the model-based CSI feedback.

[0105] The sixth mode: performing codebook-based CSI feedback, or reporting that the model inference or prediction on the terminal side cannot be completed, or skipping this time of CSI feedback.

[0106] Here, the sixth mode corresponds to the case where the terminal cannot perform model inference or prediction. In this case, it can fall back to codebook-based CSI feedback, or report to the network device side that the model inference or prediction on the terminal side cannot be completed, or skip this time of CSI feedback.

[0107] It should be noted that the terminal determines the feedback mode (including one of the above six modes) based on the indication of the network device and the model implementation of the terminal itself, and the feedback mode is flexible, which can improve the flexibility of CSI feedback. In the case where the above modes one to five cannot be implemented, the sixth mode is adopted, that is, the terminal can use the model to perform CSI feedback as much as possible to improve the CSI feedback accuracy and thus improve the system performance.

[0108] In step 103, the CSI feedback information fed back to the network device is determined based on the target feedback mode.

[0109] In some embodiments, the CSI feedback information comprises: the fed back CSI information; or the fed back CSI information and at least one of the following:

[0110] identification information of the model used by the terminal;

[0111] information indicating model-based feedback or codebook-based feedback;

[0112] information indicating a second frequency domain resource corresponding to the model used by the terminal;

[0113] information indicating a redundant frequency domain resource in addition to the first frequency domain resource;

[0114] information indicating all frequency domain resources corresponding to the fed back CSI information.

[0115] It should be noted that, according to the protocol agreement, if the CSI feedback information only comprises the fed back CSI information, it is defaulted that the terminal performs codebook-based CSI feedback. Alternatively, when the target feedback mode is mode one, according to the protocol agreement, if the CSI feedback information only comprises the fed back CSI information, it is defaulted that the terminal performs model-based CSI feedback.

[0116] The information indicating model-based feedback or codebook-based feedback as an optional item can be included in the CSI feedback information when the terminal performs model-based CSI feedback, or can be included in the CSI feedback information when the terminal performs codebook-based CSI feedback, and the remaining information other than the fed back CSI information and the information indicating model-based feedback or codebook-based feedback as an optional item can be included in the CSI feedback information when the terminal performs model-based CSI feedback.

[0117] In step 104, the CSI feedback information is sent to the network device.

[0118] The CSI feedback method of the embodiment of the application determines the actual feedback mode based on the indication of the network device and the model of the terminal itself, and the feedback mode is flexible, which can improve the flexibility of CSI feedback.

[0119] In some embodiments, the first indication information comprises:

[0120] information indicating a subband size;

[0121] information indicating whether CSI needs to be fed back for each subband;

[0122] information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or a sequence number of a starting subband in the BWP;

[0123] information indicating a number of PRBs between two adjacent subbands;

[0124] information indicating a number of PRBs or subbands;

[0125] information indicating a position of a starting PRB or subband in a BWP;

[0126] information indicating a gap between two adjacent PRBs or subbands;

[0127] information indicating a position of a starting PRB of each segment of continuous PRBs in a BWP;

[0128] information indicating a position of a first PRB of each segment of starting subbands in a BWP;

[0129] information indicating a number of each segment of continuous PRBs or subbands;

[0130] information indicating an output form of a terminal-side model.

[0131] The first indication information can include each of the above information, or each of the information can be represented by one parameter, or each of the information can be grouped and represented by one parameter, and the specific form can be determined as appropriate and is not limited.

[0132] It should be understood that the first indication information can include one or more combinations of the above information. Specifically, the following cases can be divided.

[0133] Case one, the first indication information includes information indicating a subband size and information indicating whether each subband needs to feed back CSI.

[0134] Example one, the first indication information includes subbandSize and csi-ReportingBand, wherein subbandSize indicates a subband size, and csi-ReportingBand indicates information indicating whether each subband needs to feed back CSI. Wherein csi-ReportingBand is a 3-19 bit string, and each bit is used to indicate whether the corresponding subband needs to feed back CSI. The above case one is compatible with the existing indication method of the protocol.

[0135] Case two, 1) the first indication information includes information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or a sequence number of a starting subband in the BWP, information indicating a subband size, information indicating a number of PRBs between two adjacent subbands, and information indicating a number of subbands.

[0136] Example two, the first indication information includes p1, p2, p3 and p4, wherein, p1 is used to indicate the sequence number of the starting PRB in the bandwidth part (BWP) (such as the frequency order, the network device and the terminal know the order of the rules), p2 is used to indicate the sub-band size, p2 is greater than or equal to 1, p3 is used to indicate the number of PRBs between adjacent two sub-bands, which can be 0, and p4 is used to indicate the number of sub-bands, that is, how many sub-bands in total.

[0137] Wherein, when p2 = 1, this example two indicates feedback on equally spaced PRBs. When p2 > 1, this example two indicates several sub-bands equally spaced in frequency, and the first sub-band can be at any position.

[0138] Here, if the sequence number of the starting PRB in the BWP is an integer multiple of p2, at this time, p1 is used to indicate the sequence number of the starting sub-band in the BWP.

[0139] 2) The first indication information includes: information indicating the sequence number of the starting PRB in the BWP or indicating the sequence number of the starting sub-band in the BWP, information indicating the sub-band size, and information indicating the number of sub-bands.

[0140] If the sub-bands indicated by the network device side are continuous, the information included in the first indication information is the case described in 2) above, that is, the information indicating the number of PRBs between adjacent two sub-bands can be omitted.

[0141] The above case two can indicate discontinuous but equally spaced sub-bands.

[0142] Case three, the first indication information includes: information indicating the position of the starting PRB in the BWP, information indicating the interval between adjacent two PRBs, and information indicating the number of PRBs. Or, the first indication information includes: information indicating the position of the starting sub-band in the BWP, information indicating the interval between adjacent two sub-bands, information indicating the sub-band size, and information indicating the number of sub-bands.

[0143] Example three, the first indication information includes: an array prb_interval, which includes start_prb, len and len-1 (indicating length); wherein, start_prb indicates the position of the starting PRB in the BWP (frequency low to high order), len indicates the number of PRBs (that is, how many PRBs are reported in total), and len-1 indicates the interval between adjacent two PRBs.

[0144] Example four, the first indication information includes: an array sb_interval, the array sb_interval includes start sb, len, sbsize and len-1 (indicating length); wherein, start sb indicates the position of the starting sub-band in the BWP (frequency from low to high sorting), len indicates the number of sub-bands (i.e. how many sub-bands are reported together), sbsize indicates the size of the sub-band, and len-1 indicates the interval between adjacent two sub-bands.

[0145] The above case three is used to indicate the case where there is no simple rule for the distance between sub-bands / PRBs.

[0146] Case four, the first indication information includes information indicating the position of the starting PRB of each segment of continuous PRBs in the BWP and information indicating the number of each segment of continuous PRBs. Alternatively, the first indication information includes information indicating the position of the first PRB of each segment of starting sub-bands in the BWP, information indicating the number of each segment of continuous sub-bands, and information indicating the size of the sub-band.

[0147] Example five, the first indication information includes: an array start_prb and an array len, wherein the array start_prb indicates the position of the starting PRB of each segment of continuous PRBs in the BWP, and the array len indicates the number of each segment of continuous PRBs. The length of the array start_prb and the array len is the same.

[0148] Example six, the first indication information includes: an array start_sb, an array len and sbsize, wherein the array start_sb indicates the position of the first PRB of each segment of starting sub-bands in the BWP, the array len indicates the number of each segment of continuous sub-bands, and sbsize indicates the size of the sub-band. The length of the array start_sb and the array len is the same.

[0149] The above case four is used to indicate the case where most of the PRBs are continuous but there are a few interruptions in the middle.

[0150] Case five, the first indication information includes: information indicating the output form of the terminal side model and information indicating the position of the starting PRB in the BWP.

[0151] Example seven, the first indication information includes patternID and p1, wherein patternID indicates the output form of the terminal side model, and p1 indicates the position of the starting PRB in the BWP.

[0152] If the network device knows in advance which output is supported by the terminal side model (e.g., the model ID or the data set ID used for model training is determined), the network device can configure the parameter patternID to indicate the output form of a unique terminal side model, and the parameter p1 to indicate the position of the starting PRB in the BWP. Here, the model output form refers to the relative position relationship between the subbands / PRBs corresponding to the terminal side compressed CSI. According to the position of the starting PRB in the BWP indicated by p1 and the relative relationship between all PRBs, the terminal can completely determine all the subbands / PRBs that need to be fed back, i.e., determine the first frequency domain resource.

[0153] Case five is used for a case where the output form of the terminal side model can be determined by standardizing the data set format or model ID, etc., and a more simple indication manner can be implemented.

[0154] The above indication manners of indicating which subbands / PRBs are reported by the UE show that the network device side can more flexibly indicate the frequency domain configuration (the first frequency domain resource) of the CSI that needs to be reported, so as to better adapt to the model supported by the network device side and the terminal side, thereby solving the problem that when the network device expects to recover the CSI information reported by the terminal based on the model, the indication of the network device to the terminal may not match the input and output of the terminal side model, resulting in that the double-end model cannot work.

[0155] In addition, it should be noted that the terminal side may also involve a case where the frequency domain position of the frequency domain resource needs to be reported to the network device side. As can be seen from the specific content included in the above CSI feedback information, if one or more of the information for indicating the second frequency domain resource corresponding to the model used by the terminal, the information for indicating the redundant frequency domain resource in addition to the first frequency domain resource, and the information for indicating the entire frequency domain resource corresponding to the feedback CSI information are included in the CSI feedback information, i.e., the terminal needs to send second indication information to the network device to indicate the frequency domain position of the frequency domain resource. The specific content included in the second indication information sent by the terminal to the network device is described above in the first indication information, which will not be described here.

[0156] Referring to FIG. 2, the general flow of the method of the present disclosure is described from the perspective of the interaction between devices through the following embodiment one.

[0157] Embodiment one

[0158] Step 201, model information interaction between the network device and the terminal.

[0159] That is, before the network device indicates the subbands / PRBs of the CSI feedback of the terminal, the two parties can first obtain a part of the key information of the model of the other party (the first information and the second information described above).

[0160] Here, the time of obtaining the key information and the information in which the key information is contained can refer to the descriptions of the first information and the second information above, which will not be repeated here.

[0161] In step 202, the network device determines the first indication information according to the key information of the model on the terminal side and sends the first indication information to the terminal.

[0162] That is, the network device indicates to the terminal on which sub-band / PRB the terminal needs to perform CSI feedback according to the key information of the model on the terminal side. In step 201, the network device can have known how to configure the sub-band / PRB that needs to be fed back, which can make the inference or prediction of the model on the terminal side easier.

[0163] Wherein, the specific content included in the first indication information can refer to the descriptions of the corresponding part above, which will not be repeated here.

[0164] In step 203, after receiving the first indication information sent by the network device, the terminal determines the feedback mode according to the model implemented and supported by the terminal itself, and determines the CSI feedback information based on the feedback mode.

[0165] That is, to determine whether the feedback mode is based on the model to perform CSI feedback or to fall back to the feedback mode based on the codebook to perform CSI feedback.

[0166] When the CSI feedback is based on the model, the terminal can output the compressed CSI on the sub-band / PRB substantially the same as the frequency domain configuration indicated by the network device through one or more inferences (or predictions) of the model. Then, the terminal can report the model output corresponding to the frequency domain resource indicated by the network device according to the implementation of the model itself, or report the model output including the frequency domain resource corresponding to the excess frequency domain resource. At this time, in addition to reporting one or more model outputs, the terminal can also report to the network device whether to use the model and, when necessary, report the CSI on the excess frequency domain resource in addition to the frequency domain resource indicated by the network device in the model output.

[0167] When the CSI feedback cannot be based on the model, the terminal can fall back to the CSI feedback based on the codebook.

[0168] In step 204, the terminal sends the CSI feedback information to the network device, wherein the CSI feedback information can include the fed back CSI and the feedback mode on the terminal side.

[0169] In step 205, the network device determines to recover the CSI using the model on the network device side or to recover the CSI through the codebook according to the CSI feedback information.

[0170] Wherein, when the terminal performs CSI feedback based on the model, the network device can also need to delete the CSI on the excess sub-band / PRB reported by the terminal.

[0171] Finally, steps 202-205 or steps 203-205 are executed in a loop. Here, when steps 203-205 are executed in a loop, it is explained that the first indication information is no longer changed after being determined by the network device, i.e., the terminal needs to feed back CSI on the same subband / PRB each time.

[0172] The target feedback mode related to the target feedback mode will be described below.

[0173] In some embodiments, the step 102 of determining the target feedback mode of the CSI feedback based on the frequency domain position relationship between the first frequency domain resource and the at least one group of second frequency domain resources comprises:

[0174] In a case where the frequency domain position of the first frequency domain resource is completely the same as the frequency domain position of a group of second frequency domain resources in the at least one group of second frequency domain resources and a first preset condition is met, the target feedback mode is determined as the mode one.

[0175] The first preset condition comprises that a performance index of a first model at the terminal side is good and / or a computing capability of the terminal reaches a complexity of the first model, and the first model is a model corresponding to the second frequency domain resource whose frequency domain position is completely the same as the frequency domain position of the first frequency domain resource.

[0176] Here, the at least one group of second frequency domain resources comprises one or more groups of second frequency domain resources. Corresponding to the present embodiment, when the at least one group of second frequency domain resources comprises one group of second frequency domain resources, the frequency domain position of the first frequency domain resource is completely the same as the frequency domain position of the group of second frequency domain resources (i.e., the two groups of frequency domain resources contain the same subband / PRB). At this time, the group of second frequency domain resources refers to all second frequency domain resources of the model of a given terminal.

[0177] When the at least one group of second frequency domain resources comprises multiple groups of second frequency domain resources, the frequency domain position of the first frequency domain resource is completely the same as the frequency domain position of a group of second frequency domain resources in the multiple groups of second frequency domain resources (i.e., the two groups of frequency domain resources contain the same subband / PRB). At this time, the group of second frequency domain resources refers to all second frequency domain resources of the model of a given terminal.

[0178] It should be understood that the first preset condition is met by the terminal itself. The performance index of the first model at the terminal side is good. Here, the standard for measuring that the performance index of the first model at the terminal side is good is determined by the terminal itself, such as a codebook feedback with performance exceeding overhead.

[0179] Correspondingly, the step 103 of determining the CSI feedback information to be fed back to the network device based on the target feedback mode comprises:

[0180] The model corresponding to the first frequency domain resource is selected, one model inference or prediction is performed through the model to obtain the first CSI, and the CSI feedback information fed back to the network device includes the first CSI.

[0181] The model corresponding to the first frequency domain resource is the model corresponding to a second frequency domain resource whose frequency domain position is completely same as that of the first frequency domain resource.

[0182] The first mode corresponds to a case where the terminal performs one model inference or prediction and does not need additional processing. That is, the terminal can obtain the first CSI on the first frequency domain resource (i.e., the frequency domain resource required by the network device) by performing one model inference or prediction.

[0183] The CSI feedback information can further include information indicating model-based feedback. That is, the CSI feedback information includes the first CSI and the information indicating model-based feedback.

[0184] Of course, the CSI feedback information can only include the first CSI. At this time, it can be defaulted through protocol agreement that if the network device side determines that the CSI feedback information only includes the first CSI, it is defaulted that the terminal side is model-based CSI feedback.

[0185] Further, the method of the present disclosure further includes:

[0186] The frequency domain position of the first frequency domain resource is completely same as that of a group of second frequency domain resources in the at least one group of second frequency domain resources, and the first preset condition is not met, and the target feedback mode is determined as the sixth mode.

[0187] It should be understood that the first preset condition is not met means that the terminal itself does not meet the first preset condition. Wherein, the terminal itself does not meet the first preset condition means that the performance index of the first model on the terminal side is poor and / or the computing capability of the terminal cannot reach the complexity of the first model (i.e., the computing capability of the terminal is insufficient).

[0188] Referring to FIG. 3, the flow of the method of the present disclosure is explained through the following embodiment two when the frequency domain resource indicated by the network device is completely matched with a group of frequency domain resources on the terminal side. In this case, the terminal can directly perform model-based CSI feedback without other operations.

[0189] Embodiment two

[0190] Step 301, model information interaction between the network device and the terminal.

[0191] That is, before the network device indicates the sub-band / PRB of the terminal CSI feedback, the two parties can first obtain a part of the key information of the model of the other party (the first information and the second information described above).

[0192] Here, for the acquisition time of the key information and which information the key information is contained in, please refer to the description of the first information and the second information part above, which will not be repeated here.

[0193] Then, the network device side: based on the model information obtained from the terminal side, confirm which models of the terminal side match the models of the network device side.

[0194] Terminal side: based on the model information obtained from the network device side, confirm which models of the terminal side match the models of the network device side.

[0195] Step 302, the network device determines the first indication information according to the model information of the terminal side and its own needs (such as which sub-band CSI is needed, and what kind of frequency domain configuration the input of the network device side model needs), and sends it to the terminal.

[0196] Wherein, for the specific content included in the first indication information, please refer to the description of the corresponding part above, which will not be repeated here.

[0197] Step 303, the terminal performs CSI feedback.

[0198] Specifically, the terminal determines that the frequency domain position of the first frequency domain resource indicated by the network device is exactly the same as the frequency domain position of a group of second frequency domain resources of the terminal side based on the first indication information. At this time, the following steps 303-1 or 303-2 can be used to realize CSI feedback:

[0199] Step 303-1: the performance index of the first model of the terminal side is good, and the computing capacity of the terminal reaches the complexity of the first model, the terminal feeds back the output content of the model and notifies the network device that the model is used for CSI feedback this time (if necessary, the specific model can be specified by model ID, data set ID of model training, etc.).

[0200] Step 303-2: when the performance index of the first model of the terminal side is not good or the computing capacity of the terminal is insufficient under the current condition, fallback to codebook-based CSI.

[0201] Step 304, the network device recovers the CSI according to the model output of the terminal side fed back by the terminal (corresponding to step 303-1) or recovers the CSI through the codebook (corresponding to step 303-2).

[0202] It should be noted that in this embodiment two, the first frequency domain resource and the second frequency domain resource do not necessarily represent continuous sub-bands / PRBs.

[0203] The above is the description of the content related to mode one and mode six. Next, the content related to mode two or mode three and the content related to mode six is described.

[0204] In some embodiments, the step 102 of determining the target feedback mode of the channel state information (CSI) feedback based on the frequency domain location relationship between the first frequency domain resource and the at least one group of second frequency domain resources comprises:

[0205] In a case where the frequency domain location range of the first frequency domain resource is contained in the frequency domain location range of the target frequency domain resource and a second preset condition is met, the target feedback mode is determined as the second mode or the third mode, and the target frequency domain resource is a group of second frequency domain resources in the at least one group of second frequency domain resources.

[0206] It should be noted that the frequency domain location range of the first frequency domain resource is contained in the frequency domain location range of the target frequency domain resource, that is, the frequency domain location range of the first frequency domain resource is located in the frequency domain location range of the target frequency domain resource. It should be understood that if the frequency domain location range of the target frequency domain resource is a set, the frequency domain location range of the first frequency domain resource is a subset of the set.

[0207] The second preset condition comprises at least one of the following:

[0208] The performance index of the terminal-side second model is good, and the second model is a model corresponding to the target frequency domain resource. Here, the standard for measuring that the performance index of the terminal-side second model is good is determined by the terminal itself (for example, the data set corresponding to the model training of the second model matches the inference condition of the second model).

[0209] The part corresponding to the first frequency domain resource can be obtained from the output of the second model. Here, the output of the second model is the CSI on the first frequency domain resource and other redundant frequency domain resources (subbands / PRBs), and here, the part corresponding to the first frequency domain resource can be obtained from the output of the second model means that the terminal can extract the bits corresponding only to the first frequency domain resource from the output of the second model.

[0210] The network device side model can input the part corresponding to the first frequency domain resource in the output of the second model and restore the first CSI. Here, since the model information of both the terminal and the network device is exchanged, the terminal has learned the relevant information of the model of the network device side, and based on this, it can be determined from the terminal side whether the network device has the above-mentioned capability. Wherein, the network device side model can input the part corresponding to the first frequency domain resource in the output of the second model, it should be understood that the network device side model can input the part corresponding to the first frequency domain resource in the output of the second model as its model input.

[0211] The first specified rule related to the terminal capability, wherein the first specified rule comprises but is not limited to the following rules:

[0212] The second model has a smaller time delay for one inference or prediction than a time delay for codebook-based feedback on the first frequency domain resource indicated by the network device.

[0213] The computing capability of the terminal reaches the complexity of the second model.

[0214] It should be understood that the second preset condition being met means that the terminal itself meets the second preset condition.

[0215] Correspondingly, the step 103 of determining the CSI feedback information fed back to the network device based on the target feedback mode comprises the following steps.

[0216] In step 1031, a model corresponding to the second frequency domain resource is selected, the second frequency domain resource comprises the first frequency domain resource and the third frequency domain resource, one inference or prediction is performed by using the model to obtain the second CSI (the CSI on the second frequency domain resource), and the CSI feedback information fed back to the network device comprises the second CSI and the frequency domain position corresponding to the third frequency domain resource, or the CSI feedback information fed back to the network device comprises the second CSI and the frequency domain position corresponding to the second frequency domain resource.

[0217] Here, the CSI feedback information fed back to the network device comprises the second CSI and the frequency domain position corresponding to the third frequency domain resource, that is, the terminal reports the complete model output, and reports the frequency domain position of the redundant part (that is, the third frequency domain resource) compared with the first frequency domain resource, that is, which subbands / PRBs are the redundant part. Then, the network device can delete the CSI information corresponding to the third frequency domain resource from the CSI according to the frequency domain position corresponding to the third frequency domain resource after recovering the CSI based on the second CSI.

[0218] And the CSI feedback information fed back to the network device comprises the second CSI and the frequency domain position corresponding to the second frequency domain resource, that is, the terminal reports the complete model output, and reports the complete frequency domain position (that is, the frequency domain position of the second frequency domain resource). Then, the network device can determine the frequency domain position of the redundant part (that is, the third frequency domain resource) according to the frequency domain position of the second frequency domain resource and the frequency domain position of the first frequency domain resource (known on the network device side) after recovering the CSI based on the second CSI, and finally delete the CSI information corresponding to the third frequency domain resource from the CSI according to the frequency domain position corresponding to the third frequency domain resource.

[0219] Step 1031 corresponds to mode two, which corresponds to the terminal performing one model inference or prediction, but needs to solve the case that the CSI of the model output corresponds to the third frequency domain resource other than the first frequency domain resource indicated by the network device side. And this mode two is to solve the case that the network device side needs to handle extra.

[0220] The CSI feedback information can further include information indicating model-based feedback and / or identification information used to determine the model used by the terminal.

[0221] Alternatively, 1032, a model corresponding to the selected second frequency domain resource is selected, the second frequency domain resource includes the first frequency domain resource and the third frequency domain resource, and the second CSI (CSI on the second frequency domain resource) is obtained by performing inference or prediction on the model; based on the second CSI, the first CSI (CSI on the first frequency domain resource) is obtained, and the CSI feedback information fed back to the network device includes the first CSI.

[0222] Step 1032 corresponds to mode three, which corresponds to the terminal performing model inference or prediction once, but needs to solve the case that the CSI output by the model corresponds to the third frequency domain resource other than the first frequency domain resource indicated by the network device side. And this mode three is to solve the case that the terminal needs additional processing, that is, the terminal distinguishes the first CSI on the first frequency domain resource and the CSI on the third frequency domain resource from the second CSI, and then only feeds back the first CSI, that is, the network device side does not need other processing.

[0223] The CSI feedback information can further include information indicating model-based feedback and / or identification information used to determine the model used by the terminal.

[0224] Further, the method of the present disclosure further comprises:

[0225] The frequency domain position range of the first frequency domain resource is contained in the frequency domain position range of the target frequency domain resource, and in the case that the second preset condition is not met, the target feedback mode is determined as the mode six.

[0226] It should be understood that not meeting the second preset condition means that the terminal itself does not meet the second preset condition.

[0227] Referring to FIG. 4, the implementation process of the method of the present disclosure is illustrated by the following embodiment three when the frequency domain resource indicated by the network device is a subset of a set of frequency domain resources on the terminal side.

[0228] Embodiment three

[0229] Step 401, model information interaction between the network device and the terminal.

[0230] That is, before the network device indicates the sub-band / PRB of the terminal CSI feedback, the two parties can first obtain a part of the key information of the model of the other party (the first information and the second information described above).

[0231] Here, for the acquisition time of the key information and which information the key information is contained in, please refer to the description in the first information and the second information part above, which will not be repeated here.

[0232] After that, the network device side: based on the model information obtained from the terminal side, confirm which models of the terminal side match the models of the network device side.

[0233] Terminal side: based on the model information obtained from the network device side, confirm which models of the terminal side match the models of the network device side.

[0234] Step 402, the network device determines the first indication information according to the model information of the terminal side and its own needs (such as which sub-band CSI is needed, and what kind of frequency domain configuration the input of the network device side model needs), and sends it to the terminal.

[0235] Among them, for the specific content included in the first indication information, please refer to the description in the corresponding part above, which will not be repeated here.

[0236] Step 403, the terminal finds that any one group of second frequency domain resources is not completely the same as the first frequency domain resource, but at least one group of second frequency domain resources completely contains the second frequency domain resource. The terminal performs CSI feedback according to the following rules.

[0237] Step 403-1, the terminal selects a corresponding model of a second frequency domain resource (the second frequency domain resource contains the first frequency domain resource), and satisfies the following condition a, then performs inference or prediction once through the corresponding model of the group of second frequency domain resources, obtains the model output, and feeds back part of the output of the model (i.e. the first CSI on the first frequency domain resource), and informs the network device that this feedback is based on the model (if necessary, the specific model can be specified through model ID, data set ID for model training, etc.), wherein the condition a includes:

[0238] The performance index of the selected model is good (there is no data set and inference mismatching problem, etc.);

[0239] The output of the selected model can obtain the part corresponding to the first frequency domain resource;

[0240] The model of the network device side can take the part of the terminal side model output corresponding to the first frequency domain resource as the model input, and restore the first CSI;

[0241] Other conditions related to terminal capability, such as:

[0242] The delay of the selected model performing inference or prediction is less than the delay of feedback based on the codebook on the first frequency domain resource indicated by the network device;

[0243] The computing power of the terminal reaches the complexity of the selected model.

[0244] It should be noted that the CSI feedback information obtained in step 403-1 can include the first CSI, information indicating model-based feedback, and identification information (such as a model ID or a data set ID for model training) for determining the model used by the terminal. In step 403-1, the terminal can distinguish that the part of the model output corresponding to the first frequency domain resource depends on the model implementation.

[0245] In step 403-2, the implementation process is similar to step 403-1, but the terminal reports the complete model input (that is, reports the second CSI on the second frequency domain resource), and at the same time reports the second frequency domain resource or the redundant frequency domain resource compared with the first frequency domain resource (that is, which subbands / PRBs are redundant compared with the first frequency domain resource).

[0246] Then, the network device restores the CSI and deletes the redundant CSI information.

[0247] The way of notifying the redundant subbands / PRBs can be referred to the description of the first indication information part above, which will not be repeated here. In step 403-2, the terminal does not need to distinguish the part of the model output corresponding to the first frequency domain resource.

[0248] It should be noted that the CSI feedback information obtained in step 403-2 can include the second CSI and the frequency domain location of the redundant frequency domain resource other than the first frequency domain resource (that is, the third frequency domain resource described above), or include the second CSI and the frequency domain location corresponding to the second frequency domain resource.

[0249] In step 403-3, if the conditions in steps 403-1 and 403-2 are not met, the terminal will fall back to a non-AI / ML model manner (such as a codebook feedback-based manner) to complete the CSI reporting, and inform the network device that the model is not used this time.

[0250] It should be noted that the CSI feedback information obtained in step 403-3 can include the feedback CSI information and information indicating codebook-based feedback.

[0251] In step 404, the network device selects to recover the first CSI on the first frequency domain resource based on the model (corresponding to step 403-1 or step 403-2) or obtains the CSI according to the codebook (corresponding to step 403-3) according to the feedback manner reported by the terminal.

[0252] It should be noted that in this embodiment three, the first frequency domain resource and the second frequency domain resource do not necessarily represent consecutive subbands / PRBs.

[0253] The above is a description of the content related to Mode Two or Mode Three and the content related to Mode Six. Next, the content related to Mode Four and Mode Six is described.

[0254] In some embodiments, the step 102 of determining the target feedback mode of the channel state information (CSI) feedback based on the frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources comprises:

[0255] When the frequency domain location range of the first frequency domain resource is combined from the frequency domain location ranges of multiple groups of second frequency domain resources and a third preset condition is met, the target feedback mode is determined to be Mode Four.

[0256] The third preset condition includes that the performance index of a third model on the terminal side is good, and / or a second specified rule related to the terminal capability, and the third model is a model related to the multiple groups of second frequency domain resources.

[0257] It should be noted that the third model being a model related to the multiple groups of second frequency domain resources can be understood as: each group of second frequency domain resources corresponds to a third model, or multiple groups of second frequency domain resources correspond to a third model, or multiple groups of second frequency domain resources are divided into several sets, and each set corresponds to a third model.

[0258] Here, the frequency domain location range of the first frequency domain resource is combined from the frequency domain location ranges of multiple groups of second frequency domain resources, which should be understood as that the first frequency domain resource is a combination of multiple groups of second frequency domain resources. That is, the first frequency domain resource can be exactly divided into multiple groups of second frequency domain resources.

[0259] It should be understood that the third preset condition being met means that the terminal itself meets the third preset condition. The standard for measuring the performance index of the third model on the terminal side being good is determined by the terminal itself (for example, the data set corresponding to the third model is matched with the inference condition of the third model). Here, if the third model on the terminal side is one, the performance index of the third model is good; if the third model on the terminal side is multiple, the performance index of each third model is good.

[0260] The second specified rule related to the terminal capability includes but is not limited to the following rules:

[0261] The latency of the terminal when completing multiple inferences meets a preset latency requirement (which may involve whether multiple models can be used in parallel, whether a single model can be in inference at the same time);

[0262] The computing capability of the terminal reaches the complexity of the used model.

[0263] It should be noted that mode four corresponds to a case where the terminal needs to perform multiple model inferences or predictions to cover the frequency domain resources indicated by the network device.

[0264] Correspondingly, the step 103 of determining the CSI feedback information fed back to the network device based on the target feedback mode comprises:

[0265] When each group of second frequency domain resources corresponds to a third model, the first CSI is obtained through inference or prediction on the corresponding second frequency domain resource by each third model (i.e., each model performs one inference or prediction), and the CSI feedback information fed back to the network device comprises the first CSI.

[0266] When each group of second frequency domain resources corresponds to a third model, the first CSI is obtained through inference or prediction on the corresponding second frequency domain resource by each third model (i.e., each model performs one inference or prediction), and the CSI feedback information fed back to the network device comprises the first CSI.

[0267] In the above two cases, the CSI feedback information can further comprise information indicating model-based feedback (if necessary, the specific model can be specified by a model ID, a data set ID trained by the model, etc., i.e., the CSI feedback information can further comprise identification information for determining the model used by the terminal) and information indicating the frequency domain location of the second frequency domain resource (referring to the above multiple groups of second frequency domain resources).

[0268] When each group of second frequency domain resources corresponds to a third model, the first CSI is obtained through inference or prediction on the corresponding second frequency domain resource by each third model (i.e., each model performs one inference or prediction), and the CSI feedback information fed back to the network device comprises the first CSI.

[0269] It should be noted that if a set comprises two groups of second frequency domain resources, the third model corresponding to the set performs inference or prediction on each group of second frequency domain resources in the set, i.e., the third model corresponding to the set performs two inferences or predictions. If a set comprises one group of second frequency domain resources, the third model corresponding to the set performs inference or prediction on the group of second frequency domain resources, i.e., the third model corresponding to the set performs one inference or prediction.

[0270] In this case, the CSI feedback information can further comprise information indicating model-based feedback (if necessary, the specific model can be specified by a model ID, a data set ID trained by the model, etc., i.e., the CSI feedback information can further comprise identification information for determining the model used by the terminal) and information indicating the frequency domain location of the second frequency domain resource corresponding to the model used by the terminal.

[0271] Further, the method of the present disclosure further comprises:

[0272] The frequency domain position range of the first frequency domain resource is combined from the frequency domain position ranges of multiple groups of second frequency domain resources, and in the case that the third preset condition is not met, the target feedback mode is determined as the sixth mode.

[0273] It should be understood that the third preset condition not being met means that the terminal itself does not meet the third preset condition.

[0274] Referring to FIG. 5, the implementation process of the method of the present disclosure is described through the following embodiment four when the frequency domain resource indicated by the network device is the combination of multiple groups of frequency domain resources on the terminal side.

[0275] Embodiment four

[0276] Step 501, model information interaction between the network device and the terminal.

[0277] That is, before the network device indicates the subband / PRB of the CSI feedback of the terminal, the two parties can first obtain a part of the key information of the model of the other party (the first information and the second information described above).

[0278] Here, the time of obtaining the key information and which information the key information is contained in can be referred to the description of the first information and the second information part above, which will not be repeated here.

[0279] Then, on the network device side: based on the model information obtained from the terminal side, confirm which models on the terminal side match the models on the network device side.

[0280] On the terminal side: based on the model information obtained from the network device side, confirm which models on the terminal side match the models on the network device side.

[0281] Step 502, the network device determines the first indication information according to the model information on the terminal side and its own needs (such as the CSI of which subband is needed, what kind of frequency domain configuration the input of the model on the network device side needs), and sends it to the terminal.

[0282] The specific content included in the first indication information can be referred to the description of the corresponding part above, which will not be repeated here.

[0283] Step 503, the terminal finds that any one group of second frequency domain resources is not completely the same as the first frequency domain resource, but the first frequency domain resource can be combined from multiple groups of second frequency domain resources. The terminal performs CSI feedback according to the following rules.

[0284] Step 503-1, the terminal selects multiple sets of second frequency domain resources and one or more models corresponding thereto, and satisfies the condition b, then after performing multiple model inferences or predictions, feeds back the output of the model on each set of second frequency domain resources, and informs the network device that the feedback this time is based on the model (if necessary, the specific model can be specified through a model ID, a data set ID for training the model, etc.). The terminal also reports information indicating the frequency domain positions corresponding to each set of second frequency domain resources, for the network device to synchronize the terminal's division of the first frequency domain resources. The condition b includes:

[0285] The performance indicator of the third model on the terminal side is good (there is no problem such as mismatch between the data set and inference time);

[0286] Other conditions related to the terminal capability, such as:

[0287] The delay of the terminal in completing multiple inferences meets the preset delay requirement (may involve whether multiple models can be used in parallel, whether a single model can be inferred at the same time);

[0288] The current computing capability of the terminal reaches the complexity of the selected model for completing inference or prediction.

[0289] It should be noted that the CSI feedback information obtained in step 503-1 can include the first CSI, information indicating feedback based on the model (if necessary, the specific model can be specified through a model ID, a data set ID for training the model, etc., that is, the CSI feedback information can also include identification information for determining the model used by the terminal), and information indicating the frequency domain positions corresponding to the second frequency domain resources (referring to the multiple sets of second frequency domain resources).

[0290] Step 503-2, when the conditions in step 503-1 are not met, the terminal reverts to a non-AI / ML model (such as a codebook feedback-based mode) to complete CSI reporting, and informs the network device that this reporting does not use a model.

[0291] It should be noted that the CSI feedback information obtained in step 503-3 can include the feedback CSI information and information indicating codebook-based feedback.

[0292] Step 504, the network device recovers the CSI according to the feedback mode reported by the terminal. For step 503-1, the model on the network device side infers on each second frequency domain resource, and then according to the second frequency domain resource combination (i.e., the multiple sets of second frequency domain resources) reported by the terminal, maps the recovered CSI to the first frequency domain resource. For step 503-2, the CSI is obtained according to the codebook.

[0293] It should be noted that in this embodiment four, the first frequency domain resource and the second frequency domain resource do not necessarily represent continuous subbands / PRBs.

[0294] The above is a description of the content related to mode four and mode six. Next, the content related to mode five and mode six is described.

[0295] In some embodiments, the step 102 of determining the target feedback mode of channel state information (CSI) feedback based on the frequency domain location relationship between the first frequency domain resource and the at least one group of second frequency domain resources comprises:

[0296] In the case that the frequency domain location of the first frequency domain resource does not match the frequency domain location of all the second frequency domain resources in the at least one group of second frequency domain resources, and a fourth preset condition is met, the target feedback mode is determined as the mode five.

[0297] It should be noted that the frequency domain location of the first frequency domain resource does not match the frequency domain location of all the second frequency domain resources in the at least one group of second frequency domain resources, that is, the frequency domain location of the first frequency domain resource does not match the frequency domain location of each group of second frequency domain resources, nor does it match the frequency domain location corresponding to any combination of multiple groups of second frequency domain resources. That is, the frequency domain location relationship between the first frequency domain resource and the at least one group of second frequency domain resources does not meet the above-mentioned embodiments two, three and four.

[0298] The mode five corresponds to the case that the frequency domain resource indicated by the network device side does not match the frequency domain resource corresponding to the terminal side model, and through some additional calculation processing, the terminal can complete the model-based CSI feedback. In this way, the terminal can use the model as much as possible to perform CSI feedback, improve the CSI feedback accuracy, and thus improve the system performance.

[0299] The fourth preset condition comprises at least one of the following:

[0300] The performance index of the fourth model is good, and the fourth model is a model corresponding to the second frequency domain resource used by the terminal. The standard for measuring the performance index of the terminal side fourth model is determined by the terminal itself. Here, if the terminal side fourth model is one, the performance index of the fourth model is good; if the terminal side fourth model is multiple, the performance index of each fourth model is good. It should be noted that the fourth model can be multiple, that is, multiple fourth models are used for inference or prediction; or one fourth model can be used, that is, the fourth model uses generalization / extensibility to perform inference or prediction on multiple groups of second frequency domain resources.

[0301] The input of the fourth model is processed by an algorithm to obtain the CSI on the second frequency domain resource corresponding to the fourth model. Here, the input of the fourth model is the first CSI on the first frequency domain resource, and the algorithm processing of the input of the fourth model is used to transform the first CSI into the CSI on the second frequency domain resource corresponding to the fourth model.

[0302] The output of the fourth model is processed by an algorithm to obtain the third CSI; here, the processing of the output of the fourth model is used to transform the output of the fourth model into the model output on the frequency domain resource containing the first frequency domain resource. That is, the obtained third CSI includes the first CSI.

[0303] The computing capability of the terminal reaches the complexity of the fourth model;

[0304] The processing delay of the fourth model meets the preset delay requirement; here, the processing delay of the fourth model includes the delay of the algorithm processing of the input and output of the fourth model, and the delay of the inference or prediction of the fourth model.

[0305] The performance loss value of the fourth model is less than the preset performance loss threshold; here, it should be understood that the algorithm processing of the input and output of the fourth model cannot cause obvious performance loss to the fourth model (for example, the model input after algorithm processing is accurate enough, and the model output after algorithm processing cannot be mapped to PRBs far away corresponding to a certain sub-band / PRB).

[0306] The part corresponding to the first frequency domain resource can be obtained from the output of the fourth model. Here, the output of the fourth model is the CSI on the first frequency domain resource and other redundant frequency domain resources (sub-bands / PRBs). The terminal can distinguish the part corresponding to the first frequency domain resource from the output of the fourth model, of course, if the terminal does not have the capability to distinguish the part corresponding to the first frequency domain resource from the output of the fourth model, the optional item does not exist.

[0307] Correspondingly, the step 103 of determining the CSI feedback information fed back to the network device based on the target feedback mode includes:

[0308] Selecting one or more fourth models, for each fourth model, performing first preset algorithm processing on the input of the fourth model to obtain processed input, and performing inference or prediction on the processed input by the corresponding fourth model to obtain the output of the model, and performing second preset algorithm processing on all model outputs to obtain the third CSI, and determining that the CSI feedback information fed back to the network device includes the third CSI.

[0309] Here, the input of the fourth model at the beginning is the CSI on the first frequency domain resource, and the first preset algorithm processing includes interpolation, repetition, etc. The first preset algorithm processing is used to transform the first CSI on the first frequency domain resource into the CSI on the second frequency domain resource corresponding to the fourth model.

[0310] The input of each fourth model is processed by the first preset algorithm to obtain CSI on a second frequency domain resource (which can be a group of second frequency domain resources or multiple groups of second frequency domain resources, i.e., the fourth model can correspond to one group or multiple groups of second frequency domain resources) corresponding to the fourth model.

[0311] The second preset algorithm processing includes interpolation, repetition, sampling, mapping between frequency domain resources (subbands / RRBs), and the like. The second preset algorithm processing is used to transform the CSI output by the model into third CSI including the first CSI, i.e., into CSI information including the first CSI of the first frequency domain resource.

[0312] The CSI feedback information can further include information indicating model-based feedback, identification information for determining the model used by the terminal, frequency domain location information indicating the second frequency domain resource corresponding to the model used by the terminal, and the like.

[0313] Further, the method of the present disclosure further includes:

[0314] When the frequency domain location of the first frequency domain resource does not match the frequency domain location of all second frequency domain resources in the at least one group of second frequency domain resources and does not satisfy the fourth preset condition, it is determined that the target feedback mode is the sixth mode.

[0315] It should be understood that not satisfying the fourth preset condition means that the terminal itself does not satisfy the fourth preset condition.

[0316] Referring to FIG. 6, the implementation process of the method of the present disclosure is described by the following embodiment five when the frequency domain resource indicated by the network device does not match all second frequency domain resources on the terminal side (i.e., does not satisfy embodiment two, embodiment three, and embodiment four).

[0317] Embodiment five

[0318] Step 601, model information interaction between the network device and the terminal.

[0319] That is, before the network device indicates the subband / PRB of the CSI feedback of the terminal, the two parties can first obtain a part of the key information of the model of the other party (the first information and the second information described above).

[0320] Here, the time of obtaining the key information and which information the key information is contained in can be referred to the description of the first information and the second information above, which will not be described here.

[0321] After that, on the network device side: based on the model information obtained from the terminal side, confirm which models on the terminal side match the models on the network device side.

[0322] Terminal side: based on the model information obtained from the network device side, confirm which models of the terminal side match the models of the network device side.

[0323] Step 602, the network device determines the first indication information according to the model information of the terminal side and its own needs (such as the CSI of which sub-band is needed, and what kind of frequency domain configuration the input of the network device side model needs), and sends it to the terminal.

[0324] Among them, the specific content included in the first indication information can refer to the elaboration of the corresponding part described above, which will not be repeated here.

[0325] For this embodiment five, even if the network device indicates the first frequency domain resource, it is known that the first frequency domain resource does not match all the second frequency domain resources of the terminal side, and it is difficult to compare the model feedback CSI, but the network device allows the terminal to report the CSI based on the model. If the network device completely excludes the terminal from feeding back based on the model, it is equivalent to completely returning to the existing standard from step 602, that is, not to start the characteristics of the model-based feedback, which is not within the scope of discussion of this disclosure.

[0326] Step 603, the terminal performs CSI feedback according to the following rules:

[0327] Step 603-1, the terminal selects one or more models, and after meeting the c condition, obtains the CSI feedback of the corresponding first frequency domain resource through additional processing (i.e. algorithm processing) of the model input and output. The model output after feedback processing, and report the network device this time feedback based on the model (if necessary, the specific model can be specified through model ID, data set ID trained by the model, etc.). The same as embodiment four, the terminal also reports the information indicating the frequency domain position of each group of second frequency domain resources (the second frequency domain resources involved in the model).

[0328] The additional processing step is:

[0329] Step 1: The terminal first performs interpolation, repetition, etc. on the CSI corresponding to the first frequency domain resource to obtain the CSI information on one or more second frequency domain resources.

[0330] Step 2: The terminal side performs model inference to obtain the model output corresponding to the second frequency domain resource.

[0331] Step 3: Through interpolation, repetition, sampling, mapping between sub-bands / PRBs, etc., the CSI of the model output is transformed to the CSI of the frequency domain including the first frequency domain resource.

[0332] Among them, the c condition includes:

[0333] The performance index of the model used by the terminal side is good (multiple models can be used for inference or prediction, or one model can be used for inference or prediction on multiple sets of second frequency domain resources using generalization / expansion).

[0334] The model input used by the terminal side is processed by the Step1 algorithm to obtain the CSI corresponding to the second frequency domain resource of the model; the model output used by the terminal side is processed by the Step3 algorithm to obtain the CSI including the first frequency domain resource. That is, the terminal implements the algorithms of Step1 and Step3, especially the algorithm of transforming the model output in step3.

[0335] The algorithm processing of Step1 and Step3 cannot cause significant performance loss to the model. (For example, the model input obtained in step1 is accurate enough, and the model output corresponding to a certain PRB / subband in step3 cannot be mapped to PRBs that are too far apart.)

[0336] It should be noted that the CSI feedback information obtained in step 603-1 can include third CSI, model-based feedback information, and if necessary, the specific model can be explicitly specified by model ID, dataset ID for model training, etc., that is, the CSI feedback information can also include identification information for determining the model used by the terminal and information indicating the frequency domain position corresponding to the second frequency domain resource used.

[0337] In step 603-2, the terminal selects one or more sets of second frequency domain resources, which can cover the first frequency domain resource and satisfy the d condition, and then obtains the model input corresponding to these second frequency domain resources by performing additional processing on the input of the model (the model corresponding to the second frequency domain resource selected by the terminal). The terminal transforms the model output to obtain compressed CSI including the first frequency domain resource. The terminal feeds back the transformed model output to inform the network device that this feedback is based on the model (if necessary, the specific model can be explicitly specified by model ID, dataset ID for model training, etc.). The terminal also needs to report the frequency domain position of the redundant frequency domain resource outside the first frequency domain resource indicated by the network device, and the frequency domain position of each second frequency domain resource used.

[0338] The additional processing step is:

[0339] Step1: The terminal first interpolates, repeats, etc. the CSI corresponding to the first frequency domain resource to obtain the CSI information corresponding to one or more second frequency domain resources.

[0340] Step2: The terminal side performs model inference to obtain the model output corresponding to the second frequency domain resource.

[0341] Step3: Transform the CSI output by the model to the frequency domain, including the CSI of the first frequency domain resource, through interpolation, repetition, sampling, mapping between subbands / PRBs, and the like.

[0342] wherein the d condition comprises:

[0343] The performance indicator of the model used by the terminal is good (multiple models can be used for inference or prediction, or one model can be used for inference or prediction on multiple sets of second frequency domain resources using generalization / expansion).

[0344] The algorithmic processing of steps 1 and 3 does not cause significant performance loss to the model. (For example, the model input obtained in step 1 is accurate enough, and the model output corresponding to a certain PRB / subband in step 3 cannot be mapped to PRBs that are too far apart.)

[0345] The computing power of the terminal reaches the complexity of the model used; wherein the computing power of the terminal includes the computing power of additional processing of model input and output, and the computing power related to model inference or prediction.

[0346] The sum of the latency of the algorithmic processing of the input and output of the used model and the latency of the inference or prediction of the used model meets the preset latency requirement.

[0347] It should be noted that the CSI feedback information obtained in step 603-2 can include third CSI, model feedback-based information, and if necessary, the specific model can be explicitly specified through model ID, dataset ID for model training, and the like, i.e., the CSI feedback information can also include identification information for determining the model used by the terminal, frequency domain location information indicating the frequency domain locations corresponding to the additional frequency domain resources other than the first frequency domain resource, and information indicating the frequency domain location corresponding to the used second frequency domain resource.

[0348] Step 603-3: The implementation process is similar to step 603-2, but the terminal can distinguish the part corresponding to the first frequency domain resource from the model output, and report the CSI corresponding to the first frequency domain resource indicated by the network device. Step 603-3 does not have additional subbands / PRBs to report to the network device. At this time, the network device can normally recover the CSI through the feedback of the terminal. In this case, the terminal needs to meet the d condition and be able to distinguish the part corresponding to the first frequency domain resource from the model output.

[0349] It should be noted that the CSI feedback information obtained in step 603-3 can include the CSI of the first frequency domain resource and model feedback-based information, and if necessary, the specific model can be explicitly specified through model ID, dataset ID for model training, and the like, i.e., the CSI feedback information can also include identification information for determining the model used by the terminal.

[0350] Step 603-4, when steps 603-1, 603-2 and 603-3 are not met, the terminal falls back to codebook-based CSI feedback, and informs the network device that the current transmission does not use the model.

[0351] Step 604, the network device recovers the CSI according to the feedback mode reported by the terminal. Optionally, the network device recovers the CSI on the first frequency domain resource according to the terminal-reported CSI on each group of second frequency domain resources (corresponding to steps 603-1, 603-2 and 603-3), or the network device obtains the CSI according to the codebook (corresponding to step 603-4).

[0352] It should be noted that in this embodiment, the first frequency domain resource and the second frequency domain resource do not necessarily represent consecutive subbands / PRBs. It is impossible to combine the second frequency domain resources to form the first frequency domain resource, or to cover the configuration of the first frequency domain resource through one inference or prediction. At this time, corresponding to this embodiment, the terminal can use step 603-1 to infer or predict on the second frequency domain resource 1, the second frequency domain resource 2 and the second frequency domain resource 3. Alternatively, the terminal can use step 603-2 or step 603-3 to infer or predict on the second frequency domain resource 1 and the second frequency domain resource 4.

[0353] The CSI feedback method of the embodiments of the present disclosure first receives the first indication information sent by the network device, wherein the first indication information is used to indicate the first frequency domain resource; then determines the target feedback mode of channel state information (CSI) feedback based on the frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, wherein the second frequency domain resource is the frequency domain resource corresponding to the terminal-side model, and at least one group of the second frequency domain resources corresponds to one terminal-side model; then determines the CSI feedback information fed back to the network device based on the target feedback mode; and finally sends the CSI feedback information to the network device. In this way, the terminal determines the actual feedback mode to be executed based on the indication of the network device and the implementation of its own model, and the feedback mode is flexible, which can improve the flexibility of CSI feedback.

[0354] As shown in FIG. 7, it is a flowchart of the CSI recovery method provided by the embodiments of the present disclosure. The method is applied to a network device, i.e., executed by the network device. In some embodiments, the network device is a base station. Specifically, the method comprises the following steps:

[0355] Step 701, sending first indication information to a terminal, wherein the first indication information is used to indicate a first frequency domain resource, and the first frequency domain resource is used for channel state information (CSI) feedback of the terminal;

[0356] It should be noted that before the first indication information is sent to the terminal, the network device and the terminal can perform information interaction of the two-party model. On the one hand, it provides a basis for the network device to determine the first indication information; on the other hand, it provides a basis for the terminal to use which model or which models for CSI feedback. The implementation process of the information interaction of the two-party model between the terminal and the network device can be illustrated by the following embodiments.

[0357] In some embodiments, before step 701, the method of the present disclosure further includes:

[0358] sending, to the terminal, first information related to the model on the network device side, the first information being used to determine a model on the terminal side that matches the model on the network device side.

[0359] Here, the related explanation and description of the first information can be found in the description of the terminal-side method, which will not be repeated here.

[0360] In this embodiment, after the network device sends the first information to the terminal, the terminal determines the terminal-side model that matches the model on the network device side according to the first information sent by the network device, that is, determines which model or which models match the model on the network device side, so as to subsequently perform CSI feedback through the model, so as to improve the CSI feedback accuracy.

[0361] In some embodiments, before step 701, the method of the present disclosure further includes:

[0362] receiving second information related to the model on the terminal side and sent by the terminal;

[0363] determining, according to the second information, a model on the terminal side that matches the model on the network device side;

[0364] determining the first indication information based on the model on the terminal side that matches the model on the network device side.

[0365] Here, the related explanation and description of the second information can be found in the description of the terminal-side method, which will not be repeated here.

[0366] Here, the matching of the model on the terminal side and the model on the network device side means that the model on the network device side can infer and recover the CSI information based on the output of the model on the terminal side.

[0367] In this embodiment, the network device determines which model or models on the terminal side match the model on the network device side according to the second information sent by the terminal, and can also learn the frequency domain resources corresponding to the model on the terminal side that matches the model on the network device side. This enables the network device to determine how to configure the frequency domain resources (such as subbands / PRBs) that need to be fed back, that is, to determine the first indication information, based on the information obtained above and in combination with its own needs (such as the CSI of which subbands is needed, and what kind of frequency domain configuration the input of the model on the network device side needs), so as to enable the terminal to more easily perform model-based inference or prediction and improve the accuracy of CSI feedback.

[0368] The model described in the embodiments of the present disclosure can be an AI model, an AI / ML model, an encoder (on the network device side), a decoder (on the terminal side), a CSI generation part (on the terminal side), a CSI reconstruction part (on the network device side), and the like. These can be collectively referred to as a model.

[0369] It should be understood that different frequency domain location relationships correspond to different target feedback modes of CSI feedback. In some embodiments, the target feedback mode includes one of the following modes:

[0370] Mode one: performing a model inference or prediction once to obtain the first CSI on the first frequency domain resource, and feeding back the first CSI;

[0371] Here, mode one corresponds to the case where the terminal performs a model inference or prediction once without additional processing. That is, the terminal can obtain the first CSI on the first frequency domain resource (i.e., the frequency domain resource required by the network device) by performing a model inference or prediction once. It should be understood that in this case, the model on the terminal that performs the inference or prediction corresponds to the second frequency domain resource that is exactly the same as the first frequency domain resource (i.e., the two sets of frequency domain resources contain the same subbands / PRBs).

[0372] Mode two: performing a model inference or prediction once to obtain the second CSI on the second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the second CSI and the frequency domain location corresponding to the third frequency domain resource, or feeding back the second CSI and the frequency domain location corresponding to the second frequency domain resource;

[0373] Mode three: performing a model inference or prediction once to obtain the second CSI on the second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the first CSI;

[0374] Here, the second and third manners correspond to the terminal performing one-time model inference or prediction, but need to solve the case that the CSI output by the model corresponds to a third frequency domain resource other than the first frequency domain resource indicated by the network device side.

[0375] Specifically, if the second manner is adopted, the network device side needs to handle the extra processing, that is, after the terminal feeds back the second CSI and the frequency domain position corresponding to the third frequency domain resource (i.e., the redundant frequency domain resource), the network device side deletes the CSI on the third frequency domain resource after recovering the CSI (i.e., deletes the CSI on the third frequency domain resource), and finally obtains the first CSI on the first frequency domain resource.

[0376] If the second manner is adopted, the terminal side needs to handle the extra processing, that is, the terminal distinguishes the first CSI on the first frequency domain resource (the first CSI corresponding to the first frequency domain resource) and the CSI on the third frequency domain resource (the CSI corresponding to the third frequency domain resource) from the second CSI, and then only feeds back the first CSI, that is, the network device side does not need other processing.

[0377] The fourth manner: performing multiple model inferences or predictions to obtain the first CSI, and feeding back the first CSI.

[0378] Here, the fourth manner corresponds to the case that the terminal needs to perform multiple model inferences or predictions to cover the frequency domain resources indicated by the network device side.

[0379] The fifth manner: after the model input is processed by an algorithm, the model performs inference or prediction on one or more groups of the second frequency domain resources, and after the obtained model output is processed by an algorithm, the third CSI is obtained and fed back; wherein the third CSI includes the first CSI.

[0380] Here, the fifth manner corresponds to the case that the frequency domain resources indicated by the network device side do not match the frequency domain resources corresponding to the terminal side model, and some extra calculation processing enables the terminal to complete the model-based CSI feedback.

[0381] The sixth manner: performing codebook-based CSI feedback, or reporting that the terminal side model inference or prediction cannot be completed, or skipping this time of CSI feedback.

[0382] Here, the sixth manner corresponds to the case that the terminal cannot perform model inference or prediction. In this case, it can fall back to codebook-based CSI feedback, or report the case that the terminal side model inference or prediction cannot be completed to the network device side, or skip this time of CSI feedback.

[0383] It should be noted that the terminal determines the actual feedback mode (including one of the above six modes) based on the indication of the network device and the model of the terminal, the feedback mode is flexible, the flexibility of the CSI feedback can be improved, and the mode six is used only when the modes one to five cannot be implemented, that is, the terminal can use the model to perform CSI feedback as much as possible, the CSI feedback accuracy is improved, and the system performance is improved.

[0384] At step 702, the CSI feedback information sent by the terminal is received, wherein the CSI feedback information is obtained by the terminal based on the target feedback mode; the target feedback mode is determined by the terminal based on the frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resource is the frequency domain resource corresponding to the model on the terminal side, and at least one group of the second frequency domain resources corresponds to one model on the terminal side.

[0385] In some embodiments, the CSI feedback information includes: the feedback CSI information; or the feedback CSI information and at least one of the following:

[0386] The identification information of the model used by the terminal is used to determine the model used by the terminal.

[0387] The information for indicating the model-based feedback or the codebook-based feedback is used to indicate the model-based feedback or the codebook-based feedback.

[0388] The information for indicating the second frequency domain resource corresponding to the model used by the terminal is used to indicate the second frequency domain resource corresponding to the model used by the terminal.

[0389] The information for indicating the redundant frequency domain resource in addition to the first frequency domain resource is used to indicate the redundant frequency domain resource in addition to the first frequency domain resource.

[0390] The information for indicating the entire frequency domain resource corresponding to the feedback CSI information is used to indicate the entire frequency domain resource corresponding to the feedback CSI information.

[0391] It should be noted that according to the protocol agreement, if the CSI feedback information only includes the feedback CSI information, it is defaulted that the terminal performs the codebook-based CSI feedback. Alternatively, when the target feedback mode is mode one, according to the protocol agreement, if the CSI feedback information only includes the feedback CSI information, it is defaulted that the terminal performs the model-based CSI feedback.

[0392] The information for indicating the model-based feedback or the codebook-based feedback as an optional item can be included in the CSI feedback information when the terminal performs the model-based CSI feedback, or can be included in the CSI feedback information when the terminal performs the codebook-based CSI feedback; and the remaining information other than the feedback CSI information and the information for indicating the model-based feedback or the codebook-based feedback as an optional item can be included in the CSI feedback information when the terminal performs the model-based CSI feedback.

[0393] At step 703, the CSI is recovered according to the CSI feedback information.

[0394] It should be noted that the CSI recovery method on the network device side corresponds to the CSI feedback method on the terminal side, and the explanation of some terms can be found in the terminal side, which will not be repeated here.

[0395] In some embodiments, the first indication information includes:

[0396] information indicating a size of a subband;

[0397] information indicating whether CSI needs to be fed back for each subband;

[0398] information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or indicating a sequence number of a starting subband in the BWP;

[0399] information indicating a number of PRBs between two adjacent subbands;

[0400] information indicating a number of PRBs or subbands;

[0401] information indicating a position of a starting PRB or subband in the BWP;

[0402] information indicating a gap between two adjacent PRBs or subbands;

[0403] information indicating a position of a starting PRB of each segment of continuous PRBs in the BWP;

[0404] information indicating a position of a first PRB of each segment of starting subbands in the BWP;

[0405] information indicating a number of each segment of continuous PRBs or subbands;

[0406] information indicating an output form of a terminal-side model.

[0407] The various information included in the first indication information can be represented by one parameter, each information can be represented by one parameter, or each information can be grouped and represented by one parameter, and the specific form can be determined as appropriate and is not limited.

[0408] It should be understood that the first indication information can include one or more combinations of the above information. Specifically, it can be divided into the following cases.

[0409] Case 1: The first indication information includes information indicating a size of a subband and information indicating whether CSI needs to be fed back for each subband.

[0410] Example one, the first indication information includes subbandSize and csi-ReportingBand, wherein, subbandSize indicates the subband size, and csi-ReportingBand indicates the information whether each subband needs to feed back the CSI. Wherein, csi-ReportingBand is a 3-19 bit string, and each bit is used to indicate whether the corresponding subband needs to feed back the CSI. The above case one is the existing indication mode compatible with the protocol.

[0411] Case two, 1) the first indication information includes: information indicating the sequence number of the starting physical resource block (PRB) in the bandwidth part (BWP) or indicating the sequence number of the starting subband in the BWP, information indicating the subband size, information indicating the number of PRBs between adjacent two subbands, and information indicating the number of subbands.

[0412] Example two, the first indication information includes p1, p2, p3 and p4, wherein, p1 is used to indicate the sequence number of the starting physical resource block (PRB) in the bandwidth part (BWP) (such as frequency sorting, the network device and the terminal know the sorting rule), p2 is used to indicate the subband size, p2 is greater than or equal to 1, p3 is used to indicate the number of PRBs between adjacent two subbands, which can be 0, and p4 is used to indicate the number of subbands, that is, how many subbands in total.

[0413] Wherein, when p2 = 1, this example two indicates to feed back on equally spaced PRBs. When p2 > 1, this example two indicates to feed back on several equally spaced subbands in frequency, and the first subband can be at any position.

[0414] Here, if the sequence number of the starting PRB in the BWP is an integer multiple of p2, at this time, p1 is used to indicate the sequence number of the starting subband in the BWP.

[0415] 2) The first indication information includes: information indicating the sequence number of the starting PRB in the BWP or indicating the sequence number of the starting subband in the BWP, information indicating the subband size, and information indicating the number of subbands.

[0416] If the subbands indicated by the network device side are continuous, the information included in the first indication information is the case 2) described above, that is, the information indicating the number of PRBs between adjacent two subbands can be omitted.

[0417] The above case two can indicate discontinuous but equally spaced subbands.

[0418] Case three, the first indication information includes: information indicating the position of the starting PRB in the BWP, information indicating the interval between adjacent two PRBs, and information indicating the number of PRBs. Or, the first indication information includes: information indicating the position of the starting sub-band in the BWP, information indicating the interval between adjacent two sub-bands, information indicating the size of the sub-band, and information indicating the number of sub-bands.

[0419] Example three, the first indication information includes: an array prb_interval including start_prb, len, and len-1 (indicating length); wherein start_prb indicates the position of the starting PRB in the BWP (frequency low to high sorting), len indicates the number of PRBs (i.e., how many PRBs are reported in total), and len-1 indicates the interval between adjacent two PRBs.

[0420] Example four, the first indication information includes: an array sb_interval including start_sb, len, sbsize, and len-1 (indicating length); wherein start_sb indicates the position of the starting sub-band in the BWP (frequency low to high sorting), len indicates the number of sub-bands (i.e., how many sub-bands are reported in total), sbsize indicates the size of the sub-band, and len-1 indicates the interval between adjacent two sub-bands.

[0421] The above case three is used to indicate the case where there is no simple rule between sub-bands / PRBs.

[0422] Case four, the first indication information includes information indicating the position of the starting PRB of each segment of continuous PRBs in the BWP and information indicating the number of each segment of continuous PRBs. Or, the first indication information includes information indicating the position of the first PRB of each segment of starting sub-bands in the BWP, information indicating the number of each segment of continuous sub-bands, and information indicating the size of the sub-band.

[0423] Example five, the first indication information includes: an array start_prb indicating the position of the starting PRB of each segment of continuous PRBs in the BWP, and an array len indicating the number of each segment of continuous PRBs. The length of the array start_prb is the same as that of the array len.

[0424] Example six, the first indication information includes: an array start_sb indicating the position of the first PRB of each segment of starting sub-bands in the BWP, an array len indicating the number of each segment of continuous sub-bands, and sbsize indicating the size of the sub-band. The length of the array start_sb is the same as that of the array len.

[0425] Case four is used to indicate the case that most PRBs are continuous but there are a few interruptions in the middle.

[0426] Case five, the first indication information includes: information indicating the output form of the terminal side model and information indicating the position of the starting PRB in the BWP.

[0427] Example seven, the first indication information includes patternID and p1, wherein patternID indicates the output form of the terminal side model, and p1 indicates the position of the starting PRB in the BWP.

[0428] If the network device knows in advance which output the terminal side model supports (such as determining the model ID or the dataset ID for model training of the terminal side), the network device can configure the parameter patternID to indicate the output form of a unique terminal side model, and the parameter p1 to indicate the position of the starting PRB in the BWP. Here, the model output form refers to the relative position relationship between the subbands / PRBs corresponding to the terminal side compressed CSI. According to the position of the starting PRB in the BWP indicated by p1 and the relative relationship between all PRBs, the terminal can completely determine all the subbands / PRBs that need to be fed back, that is, determine the first frequency domain resource.

[0429] Case five is used to achieve a simpler indication method in the case that the output form of the terminal side model can be determined by standardizing the dataset format or model ID, etc.

[0430] The above-mentioned indication method of indicating which subbands / PRBs the UE reports shows that the network device side can more flexibly indicate the frequency domain configuration (first frequency domain resource) of the CSI that needs to be reported, so as to better adapt to the model supported by the network device side and the terminal side, thereby solving the problem that when the network device expects to recover the CSI information reported by the terminal based on the model, the indication of the network device to the terminal may not match the input and output of the model of the terminal side, resulting in that the two-end model cannot work.

[0431] In addition, it should be noted that the terminal side may also involve the case that the frequency domain position of the frequency domain resource needs to be reported to the network device side. As can be seen from the specific content included in the above-mentioned CSI feedback information, if one or more of the information indicating the second frequency domain resource corresponding to the model used by the terminal, the information indicating the redundant frequency domain resource in addition to the first frequency domain resource, and the information indicating the entire frequency domain resource corresponding to the feedback CSI information are included in the CSI feedback information, that is, the terminal needs to send second indication information to the network device to indicate the frequency domain position of the frequency domain resource. The specific content included in the second indication information sent by the terminal to the network device is referred to the above-mentioned first indication information, which will not be described here.

[0432] In some embodiments, the step 703 of recovering the CSI according to the CSI feedback information comprises:

[0433] In the case that the CSI feedback information comprises information for indicating model-based feedback, the CSI is recovered according to the CSI feedback information by a model at the network device side;

[0434] In the case that the CSI feedback information comprises information for indicating codebook-based feedback, the CSI is recovered according to the CSI feedback information by a codebook.

[0435] As can be known from the method at the terminal side, the terminal can feed back the CSI based on a model or based on a codebook. If the terminal feeds back the CSI based on a model, the network device also recovers the CSI from the fed back CSI information by a model. If the terminal feeds back the CSI based on a codebook, the network device recovers the CSI from the fed back CSI information by a codebook.

[0436] Further, the CSI is recovered according to the CSI feedback information by a model at the network device side, comprising:

[0437] In the case that the fed back CSI information is the first CSI on the first frequency domain resource, the CSI is recovered by the model at the network device side.

[0438] In this case, the first CSI can be determined by the terminal based on the first mode, the third mode or the fifth mode (embodiment five corresponds to 603-3). At this time, the network device recovers the CSI from the fed back first CSI information by the model at the network device side.

[0439] In the case that the fed back CSI information is the first CSI, the CSI is recovered by the model at the network device side based on the information for indicating the second frequency domain resource corresponding to the model used by the terminal.

[0440] In this case, the first CSI can be determined by the terminal based on the fourth mode, i.e., the terminal performs multiple model inferences or predictions. At this time, the model at the network device side inferences the first CSI on each second frequency domain resource based on the information for indicating the second frequency domain resource corresponding to the model used by the terminal, and then maps the recovered CSI to the first frequency domain resource.

[0441] In a case where the fed back CSI information is CSI on the first frequency domain resource and a frequency domain resource in addition to the first frequency domain resource, the CSI is recovered by the model of the network side, and CSI on the frequency domain resource in addition to the first frequency domain resource is deleted from the recovered CSI based on the information indicating the frequency domain resource in addition to the first frequency domain resource or the information indicating all frequency domain resources corresponding to the fed back CSI information.

[0442] In this case, the CSI can be determined by the terminal based on the second mode, and after the recovered CSI is recovered by the model of the network device side, CSI not on the first frequency domain resource (i.e., CSI on the frequency domain resource in addition to the first frequency domain resource) is deleted based on the information indicating the frequency domain resource in addition to the first frequency domain resource.

[0443] Alternatively, after the recovered CSI is recovered by the model of the network device side, the frequency domain resource in addition to the first frequency domain resource is determined based on the information indicating all frequency domain resources corresponding to the fed back CSI information and the known first frequency domain resource, and then CSI on the frequency domain resource in addition to the first frequency domain resource is deleted from the recovered CSI.

[0444] In a case where the fed back CSI is CSI on the first frequency domain resource and a frequency domain resource in addition to the first frequency domain resource, the CSI is recovered by the model of the network device side based on the information indicating the second frequency domain resource corresponding to the model used by the terminal, and CSI on the frequency domain resource in addition to the first frequency domain resource is deleted from the recovered CSI based on the information indicating the frequency domain resource in addition to the first frequency domain resource or the information indicating all frequency domain resources corresponding to the fed back CSI information.

[0445] The CSI recovery method of the embodiments of the present disclosure recovers CSI by sending first indication information to a terminal, the first indication information being used to indicate a first frequency domain resource, the first frequency domain resource being used for channel state information (CSI) feedback of the terminal; receiving CSI feedback information sent by the terminal, wherein the CSI feedback information is obtained by the terminal based on a target feedback mode, the target feedback mode being determined by the terminal based on a frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resource being a frequency domain resource corresponding to a model of the terminal side, and at least one group of the second frequency domain resources corresponding to one model of the terminal side; and recovering CSI according to the CSI feedback information, so that the CSI feedback information determined by the terminal side using a flexible feedback mode is received, the CSI feedback information can fully adapt to the network device, the network device recovers the CSI, and the system performance is improved.

[0446] As shown in FIG. 8, the embodiment of the present disclosure further provides a terminal, comprising: a transceiver 800, a memory 820, a processor 810, and a computer program stored in the memory 820 and capable of running on the processor 810; the transceiver 800 is used for receiving and sending data under the control of the processor 810; the transceiver 800 is used for receiving first indication information sent by a network device, wherein the first indication information is used for indicating a first frequency domain resource;

[0447] The processor 810 is used for reading the program in the memory 820 and performing the following process:

[0448] Based on the frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, a target feedback mode of channel state information (CSI) feedback is determined, wherein the second frequency domain resource is a frequency domain resource corresponding to a terminal-side model, and at least one group of the second frequency domain resources corresponds to one terminal-side model;

[0449] Based on the target feedback mode, CSI feedback information fed back to the network device is determined.

[0450] The transceiver 800 is further used for sending the CSI feedback information to the network device.

[0451] In FIG. 8, the bus architecture can include any number of interconnected buses and bridges, which are specifically linked together by various circuits of the processor 810 representing one or more processors and the memory 820 representing the memory. The bus architecture can also link various other circuits such as peripheral devices, voltage stabilizers, and power management circuits, which are well known in the art, and therefore, further description thereof will not be given herein. The bus interface provides an interface. The transceiver 800 can be a plurality of elements, i.e., including a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which can be a wireless channel, a wired channel, an optical cable, etc. The user interface 830 can also be an interface capable of connecting the required devices externally or internally for different user devices, including but not limited to a keypad, a display, a speaker, a microphone, a joystick, etc.

[0452] The processor 810 is responsible for managing the bus architecture and general processing, and the memory 820 can store data used by the processor 810 in performing operations.

[0453] In some embodiments, the processor 810 can be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or a complex programmable logic device (CPLD), and can also adopt a multi-core architecture.

[0454] The processor 810 is configured to execute any of the methods provided by the embodiments of the present disclosure by invoking stored program instructions of the memory. The processor 810 and the memory 820 can also be physically arranged separately.

[0455] In some embodiments, the target feedback mode includes one of the following modes:

[0456] Mode one: performing a model inference or prediction to obtain a first CSI on the first frequency domain resource, and feeding back the first CSI;

[0457] Mode two: performing a model inference or prediction to obtain a second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the second CSI and a frequency domain location corresponding to the third frequency domain resource, or feeding back the second CSI and a frequency domain location corresponding to the second frequency domain resource;

[0458] Mode three: performing a model inference or prediction to obtain a second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the first CSI;

[0459] Mode four: performing multiple model inferences or predictions to obtain the first CSI, and feeding back the first CSI;

[0460] Mode five: after algorithm processing of model input, performing inference or prediction of the model on one or more groups of the second frequency domain resource, after algorithm processing of the obtained model output, obtaining a third CSI, and feeding back the third CSI; wherein the third CSI includes the first CSI;

[0461] Mode six: performing codebook-based CSI feedback, or reporting that model inference or prediction at the terminal side cannot be completed, or skipping this time of CSI feedback.

[0462] In some embodiments, the transceiver 800 is further configured to receive first information related to the model at the network device side and sent by the network device.

[0463] The processor 810 is further configured to determine a terminal-side model matched with the network device-side model according to the first information.

[0464] In some embodiments, the transceiver 800 is further configured to send second information related to the terminal-side model to the network device, where the second information is used to determine a terminal-side model matched with the network device-side model.

[0465] In some embodiments, the first indication information includes:

[0466] information indicating a subband size;

[0467] information indicating whether CSI needs to be fed back for each subband;

[0468] information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or indicating a sequence number of a starting subband in the BWP;

[0469] information indicating a number of PRBs between two adjacent subbands;

[0470] information indicating a number of PRBs or subbands;

[0471] information indicating a position of a starting PRB or subband in the BWP;

[0472] information indicating a gap between two adjacent PRBs or subbands;

[0473] information indicating a position of a starting PRB of each segment of continuous PRBs in the BWP;

[0474] information indicating a position of a first PRB of each segment of starting subbands in the BWP;

[0475] information indicating a number of each segment of continuous PRBs or subbands;

[0476] information indicating an output form of the terminal-side model.

[0477] In some embodiments, the processor 810 is further configured to determine that the target feedback mode is the mode one when a frequency domain position of the first frequency domain resource is completely same as a frequency domain position of a second frequency domain resource in the at least one group of second frequency domain resources, and a first preset condition is met.

[0478] The first preset condition includes that a performance index of a terminal-side first model is good, and / or a computing capability of the terminal reaches a complexity of the first model, and the first model is a model corresponding to a second frequency domain resource whose frequency domain position is completely same as the frequency domain position of the first frequency domain resource.

[0479] In some embodiments, the processor 810 is further configured to determine that the target feedback manner is the sixth manner in a case where the frequency domain position of the first frequency domain resource is completely same as the frequency domain position of a group of second frequency domain resources in the at least one group of second frequency domain resources, and the first preset condition is not met.

[0480] In some embodiments, the processor 810 is further configured to determine that the target feedback manner is the second manner or the third manner in a case where the frequency domain position range of the first frequency domain resource is contained in the frequency domain position range of a target frequency domain resource, and a second preset condition is met, the target frequency domain resource being a group of second frequency domain resources in the at least one group of second frequency domain resources.

[0481] The second preset condition includes at least one of the following:

[0482] The performance index of a terminal-side second model is good, and the second model is a model corresponding to the target frequency domain resource.

[0483] A part corresponding to the first frequency domain resource can be obtained from the output of the second model.

[0484] A network device-side model can input the part corresponding to the first frequency domain resource output by the second model, and restore the first CSI.

[0485] A first specified rule related to terminal capability.

[0486] In some embodiments, the processor 810 is further configured to determine that the target feedback manner is the sixth manner in a case where the frequency domain position range of the first frequency domain resource is contained in the frequency domain position range of the target frequency domain resource, and the second preset condition is not met.

[0487] In some embodiments, the processor 810 is further configured to determine that the target feedback manner is the fourth manner in a case where the frequency domain position range of the first frequency domain resource is combined from the frequency domain position ranges of multiple groups of second frequency domain resources, and a third preset condition is met.

[0488] The third preset condition includes that a performance index of a terminal-side third model is good, and / or a second specified rule related to terminal capability, the third model being a model related to the multiple groups of second frequency domain resources.

[0489] In some embodiments, the processor 810 is further configured to determine that the target feedback manner is the sixth manner in a case where the frequency domain position range of the first frequency domain resource is combined from the frequency domain position ranges of multiple groups of second frequency domain resources, and the third preset condition is not met.

[0490] In some embodiments, the processor 810 is further configured to determine that the target feedback mode is the fifth mode when the frequency domain position of the first frequency domain resource does not match the frequency domain position of all second frequency domain resources in the at least one group of second frequency domain resources, and a fourth preset condition is met.

[0491] The fourth preset condition includes at least one of the following:

[0492] The fourth model is a model corresponding to a second frequency domain resource used by the terminal, and the performance index of the fourth model is good.

[0493] The input of the fourth model is processed by an algorithm to obtain the CSI on the second frequency domain resource corresponding to the fourth model.

[0494] The output of the fourth model is processed by an algorithm to obtain the third CSI.

[0495] The computing capability of the terminal reaches the complexity of the fourth model.

[0496] The processing delay of the fourth model meets the preset delay requirement.

[0497] The performance loss value of the fourth model is less than a preset performance loss threshold.

[0498] Part of the first frequency domain resource can be obtained from the output of the fourth model.

[0499] In some embodiments, the processor 810 is further configured to determine that the target feedback mode is the sixth mode when the frequency domain position of the first frequency domain resource does not match the frequency domain position of all second frequency domain resources in the at least one group of second frequency domain resources, and the fourth preset condition is not met.

[0500] In some embodiments, the CSI feedback information includes: feedback CSI information; or, the feedback CSI information and at least one of the following:

[0501] Identification information of a model used by the terminal;

[0502] Information indicating model-based feedback or codebook-based feedback;

[0503] Information indicating a second frequency domain resource corresponding to a model used by the terminal;

[0504] Information indicating a redundant frequency domain resource in addition to the first frequency domain resource;

[0505] Information indicating all frequency domain resources corresponding to the feedback CSI information.

[0506] It should be noted that the above device provided by the embodiments of the present disclosure can realize all the method steps achieved by the method embodiments and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be described in detail.

[0507] As shown in FIG. 9, the embodiments of the present disclosure further provide a terminal, comprising:

[0508] A first receiving unit 901 is configured to receive first indication information sent by a network device, wherein the first indication information is used to indicate a first frequency domain resource;

[0509] A first processing unit 902 is configured to determine a target feedback mode of channel state information (CSI) feedback based on a frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, wherein the second frequency domain resources are frequency domain resources corresponding to a terminal-side model, and at least one group of the second frequency domain resources corresponds to one terminal-side model.

[0510] A second processing unit 903 is configured to determine CSI feedback information to be fed back to the network device based on the target feedback mode.

[0511] A first sending unit 904 is configured to send the CSI feedback information to the network device.

[0512] In some embodiments, the target feedback mode comprises one of the following modes:

[0513] Mode one: performing one model inference or prediction to obtain a first CSI on the first frequency domain resource, and feeding back the first CSI;

[0514] Mode two: performing one model inference or prediction to obtain a second CSI on a second frequency domain resource, wherein the second frequency domain resource comprises the first frequency domain resource and a third frequency domain resource, and feeding back the second CSI and a frequency domain location corresponding to the third frequency domain resource, or feeding back the second CSI and a frequency domain location corresponding to the second frequency domain resource;

[0515] Mode three: performing one model inference or prediction to obtain a second CSI on a second frequency domain resource, wherein the second frequency domain resource comprises the first frequency domain resource and a third frequency domain resource, and feeding back the first CSI;

[0516] Mode four: performing multiple model inferences or predictions to obtain the first CSI, and feeding back the first CSI;

[0517] Manner five: after the model input is processed by an algorithm, the model performs inference or prediction on one or more sets of the second frequency domain resources, and after the obtained model output is processed by an algorithm, a third CSI is obtained, and the third CSI is fed back; wherein the third CSI includes the first CSI;

[0518] Manner six: performing codebook-based CSI feedback, or the model inference or prediction on the terminal side cannot be completed, or skipping this time of CSI feedback.

[0519] In some embodiments, the apparatus of the embodiments of the present disclosure further includes:

[0520] A third receiving unit configured to receive first information related to the model on the network device side and sent by the network device;

[0521] A fourth processing unit configured to determine a model on the terminal side matched with the model on the network device side according to the first information.

[0522] In some embodiments, the apparatus of the embodiments of the present disclosure further includes:

[0523] A third sending unit configured to send second information related to the model on the terminal side to the network device, the second information being used to determine a model on the terminal side matched with the model on the network device side.

[0524] In some embodiments, the first indication information includes:

[0525] Information indicating a sub-band size;

[0526] Information indicating whether CSI needs to be fed back for each sub-band;

[0527] Information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or indicating a sequence number of a starting sub-band in the BWP;

[0528] Information indicating a number of PRBs between two adjacent sub-bands;

[0529] Information indicating a number of PRBs or sub-bands;

[0530] Information indicating a position of a starting PRB or sub-band in the BWP;

[0531] Information indicating a gap between two adjacent PRBs or sub-bands;

[0532] Information indicating a position of a starting PRB of each segment of continuous PRBs in the BWP;

[0533] Information indicating a position of a first PRB of each segment of starting sub-bands in the BWP;

[0534] information indicating a number of each continuous PRB or sub-band;

[0535] information indicating an output form of a terminal-side model.

[0536] In some embodiments, the first processing unit 902 is specifically configured to:

[0537] In a case where the frequency domain position of the first frequency domain resource is completely same as the frequency domain position of a group of second frequency domain resources in the at least one group of second frequency domain resources, and a first preset condition is met, the target feedback mode is determined as the mode one.

[0538] The first preset condition includes that a performance index of a terminal-side first model is good, and / or a computing capability of the terminal reaches a complexity of the first model. The first model is a model corresponding to a second frequency domain resource whose frequency domain position is completely same as the frequency domain position of the first frequency domain resource.

[0539] In some embodiments, the apparatus of the embodiments of the present disclosure further includes:

[0540] In a case where the frequency domain position of the first frequency domain resource is completely same as the frequency domain position of a group of second frequency domain resources in the at least one group of second frequency domain resources, and the first preset condition is not met, the fifth processing unit is configured to determine the target feedback mode as the mode six.

[0541] In some embodiments, the first processing unit 902 is specifically configured to:

[0542] In a case where the frequency domain position range of the first frequency domain resource is contained in a frequency domain position range of a target frequency domain resource, and a second preset condition is met, the target feedback mode is determined as the mode two or the mode three. The target frequency domain resource is a group of second frequency domain resources in the at least one group of second frequency domain resources.

[0543] The second preset condition includes at least one of the following:

[0544] A performance index of a terminal-side second model is good. The second model is a model corresponding to the target frequency domain resource.

[0545] A part corresponding to the first frequency domain resource can be obtained from an output of the second model;

[0546] A network device-side model can input the part corresponding to the first frequency domain resource output by the second model, and restore the first CSI.

[0547] A first specified rule related to a terminal capability.

[0548] In some embodiments, the apparatus of the embodiments of the present disclosure further includes:

[0549] The sixth processing unit is configured to determine that the target feedback manner is the sixth manner when the frequency domain location range of the first frequency domain resource is contained in the frequency domain location range of the target frequency domain resource and the second preset condition is not met.

[0550] In some embodiments, the first processing unit 902 is specifically configured to:

[0551] The frequency domain location range of the first frequency domain resource is combined by the frequency domain location ranges of the plurality of groups of second frequency domain resources, and the third preset condition is met, and the target feedback manner is determined as the fourth manner.

[0552] The third preset condition includes that a performance index of a third model at a terminal side is good, and / or a second specified rule related to a terminal capability, and the third model is a model related to the plurality of groups of second frequency domain resources.

[0553] In some embodiments, the apparatuses of the embodiments of the present disclosure further include:

[0554] The seventh processing unit is configured to determine that the target feedback manner is the sixth manner when the frequency domain location range of the first frequency domain resource is combined by the frequency domain location ranges of the plurality of groups of second frequency domain resources and the third preset condition is not met.

[0555] In some embodiments, the first processing unit 902 is specifically configured to:

[0556] The frequency domain location of the first frequency domain resource does not match the frequency domain locations of all second frequency domain resources in the at least one group of second frequency domain resources, and the fourth preset condition is met, and the target feedback manner is determined as the fifth manner.

[0557] The fourth preset condition includes at least one of the following:

[0558] A performance index of a fourth model is good, and the fourth model is a model corresponding to a second frequency domain resource used by a terminal;

[0559] An input of the fourth model is processed by an algorithm to obtain a CSI on a second frequency domain resource corresponding to the fourth model;

[0560] An output of the fourth model is processed by an algorithm to obtain the third CSI;

[0561] A computing capability of the terminal reaches a complexity of the fourth model;

[0562] A processing time delay of the fourth model meets a preset time delay requirement;

[0563] The performance loss value of the fourth model is less than a preset performance loss threshold.

[0564] The output of the fourth model can obtain a part corresponding to the first frequency domain resource.

[0565] In some embodiments, the apparatus of the embodiments of the present disclosure further includes:

[0566] The eighth processing unit is configured to, in a case where the frequency domain position of the first frequency domain resource does not match the frequency domain position of all the second frequency domain resources in the at least one group of second frequency domain resources, and the fourth preset condition is not met, determine that the target feedback mode is the sixth mode.

[0567] In some embodiments, the CSI feedback information includes: feedback CSI information; or, feedback CSI information and at least one of the following:

[0568] The identification information of the model used by the terminal;

[0569] Information for indicating model-based feedback or codebook-based feedback;

[0570] Information for indicating the second frequency domain resource corresponding to the model used by the terminal;

[0571] Information for indicating the redundant frequency domain resource in addition to the first frequency domain resource;

[0572] Information for indicating the entire frequency domain resource corresponding to the feedback CSI information.

[0573] It should be noted that the division of units in the embodiments of the present disclosure is illustrative, and is only a logical functional division. In actual implementation, another division mode can be used. In addition, each functional unit in each embodiment of the present disclosure can be integrated in one processing unit, or each unit can be physically present alone, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0574] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a processor-readable storage medium. Based on such an understanding, the technical solutions of the present disclosure, essentially or in other words, the part that contributes to the related art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform all or part of the steps of the methods described in the various embodiments of the present disclosure. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0575] It should be noted that the above-mentioned device provided by the embodiments of the present disclosure can realize all the method steps realized by the method embodiments and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be described in detail.

[0576] As shown in FIG. 10, the embodiments of the present disclosure further provide a network device, comprising: a transceiver 1000, a memory 1020, a processor 1010, and a computer program stored in the memory 1020 and executable on the processor 1010; the transceiver 1000 is configured to receive and send data under the control of the processor 1010, and the transceiver 1000 is further configured to send first indication information to a terminal, the first indication information being used to indicate a first frequency domain resource, and the first frequency domain resource being used for channel state information (CSI) feedback of the terminal; and receiving CSI feedback information sent by the terminal;

[0577] The processor 1010 is configured to read the program in the memory 1020 and perform the following process:

[0578] According to the CSI feedback information, the CSI is recovered.

[0579] The CSI feedback information is obtained by the terminal based on a target feedback mode; the target feedback mode is determined by the terminal based on a frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resource being a frequency domain resource corresponding to a model on the terminal side, and at least one group of the second frequency domain resources corresponding to one model on the terminal side.

[0580] In FIG. 10, the bus architecture can include any number of interconnected buses and bridges, specifically, various circuitry of one or more processors, represented by the processor 1010, and memory, represented by the memory 1020, linked together by a bus architecture. The bus architecture can also link various other circuitry, such as peripheral devices, voltage regulators, and power management circuitry, which are well known in the art, and thus, are not further described herein. The bus interface provides an interface. The transceiver 1000 can be a plurality of elements, i.e., including a transmitter and a receiver, providing a means for communicating with various other apparatuses over transmission media, including wireless channels, wired channels, optical cables, and the like. The processor 1010 is responsible for managing the bus architecture and general processing, and the memory 1020 can store data used by the processor 1010 in executing operations.

[0581] The processor 1010 can be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD), and can also adopt a multi-core architecture.

[0582] The processor 1010 is configured to execute any of the methods provided by the embodiments of the present disclosure by invoking computer program stored in the memory 1020 according to obtained executable instructions. The processor 1010 and the memory 1020 can also be physically arranged separately.

[0583] In some embodiments, the target feedback manner includes one of the following manners:

[0584] Manner one: performing one model or prediction inference to obtain a first CSI on the first frequency domain resource, and feeding back the first CSI;

[0585] Manner two: performing one model inference or prediction to obtain a second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the second CSI and a frequency domain location corresponding to the third frequency domain resource, or feeding back the second CSI and a frequency domain location corresponding to the second frequency domain resource;

[0586] Manner three: performing one model inference or prediction to obtain a second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the first CSI;

[0587] Manner four: performing multiple model inferences or predictions to obtain the first CSI, and feeding back the first CSI;

[0588] Manner five: performing model inference or prediction on one or more groups of the second frequency domain resources after algorithm processing of the model input, obtaining third CSI after algorithm processing of the obtained model output, and feeding back the third CSI; wherein the third CSI includes the first CSI;

[0589] Manner six: performing codebook-based CSI feedback, or skipping this time of CSI feedback, or reporting that the terminal-side model inference or prediction cannot be completed.

[0590] In some embodiments, the transceiver 1000 is further configured to send first information related to the network device-side model to the terminal, wherein the first information is used to determine a terminal-side model matched with the network device-side model.

[0591] In some embodiments, the transceiver 1000 is further configured to receive second information related to the terminal-side model sent by the terminal;

[0592] The processor 1010 is further configured to determine a terminal-side model matched with the network device-side model according to the second information.

[0593] Determine the first indication information based on the terminal-side model matched with the network device-side model.

[0594] In some embodiments, the first indication information includes:

[0595] Information indicating a sub-band size;

[0596] Information indicating whether CSI needs to be fed back for each sub-band;

[0597] Information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or indicating a sequence number of a starting sub-band in the BWP;

[0598] Information indicating a number of PRBs between two adjacent sub-bands;

[0599] Information indicating a number of PRBs or sub-bands;

[0600] Information indicating a position of a starting PRB or sub-band in the BWP;

[0601] Information indicating an interval between two adjacent PRBs or sub-bands;

[0602] Information indicating a position of a starting PRB of each segment of continuous PRBs in the BWP;

[0603] information indicating a position of a first PRB of each segment sub-band in a BWP;

[0604] information indicating a number of consecutive PRBs or sub-bands of each segment;

[0605] information for indicating an output form of a terminal-side model.

[0606] In some embodiments, the CSI feedback information comprises: feedback CSI information; or, the feedback CSI information and at least one of the following:

[0607] identification information for determining a model used by a terminal;

[0608] information for indicating model-based feedback or codebook-based feedback;

[0609] information for indicating a second frequency domain resource corresponding to a model used by a terminal;

[0610] information for indicating a redundant frequency domain resource in addition to the first frequency domain resource;

[0611] information for indicating all frequency domain resources corresponding to the feedback CSI information.

[0612] In some embodiments, the processor 1010 is further configured to, in a case where the CSI feedback information comprises information for indicating model-based feedback, recover CSI through a network device-side model according to the CSI feedback information;

[0613] In a case where the CSI feedback information comprises information for indicating codebook-based feedback, recover CSI through a codebook according to the CSI feedback information.

[0614] In some embodiments, the processor 1010 is further configured to, in a case where the feedback CSI information is first CSI on the first frequency domain resource, recover CSI through the network device-side model;

[0615] In a case where the feedback CSI information is the first CSI, recover CSI through the network device-side model based on the information for indicating a second frequency domain resource corresponding to a model used by a terminal;

[0616] In a case where the feedback CSI information is CSI on the first frequency domain resource and a redundant frequency domain resource in addition to the first frequency domain resource, recover CSI through the network-side model, and delete CSI on the redundant frequency domain resource from the recovered CSI based on the information for indicating the redundant frequency domain resource in addition to the first frequency domain resource, or the information for indicating all frequency domain resources corresponding to the feedback CSI information;

[0617] In a case where the fed back CSI is CSI on the first frequency domain resource and the frequency domain resource in addition to the first frequency domain resource, the network device recovers the CSI by the model on the network device side based on the information indicating the second frequency domain resource corresponding to the model used by the terminal, and deletes the CSI on the frequency domain resource in addition to the first frequency domain resource from the recovered CSI based on the information indicating the frequency domain resource in addition to the first frequency domain resource or the information indicating all the frequency domain resources corresponding to the fed back CSI information.

[0618] It should be noted that the above device provided by the embodiments of the present disclosure can realize all the method steps realized by the above method embodiments, and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be described in detail.

[0619] As shown in FIG. 11, the embodiments of the present disclosure further provide a network device, comprising:

[0620] A second sending unit 1101 is configured to send first indication information to a terminal, the first indication information being used for indicating a first frequency domain resource, and the first frequency domain resource being used for channel state information (CSI) feedback of the terminal.

[0621] A second receiving unit 1102 is configured to receive CSI feedback information sent by the terminal.

[0622] A third processing unit 1103 is configured to recover CSI according to the CSI feedback information.

[0623] The CSI feedback information is obtained by the terminal based on a target feedback mode, and the target feedback mode is determined by the terminal based on a frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resource being a frequency domain resource corresponding to a model on the terminal side, and at least one group of the second frequency domain resources corresponding to one model on the terminal side.

[0624] In some embodiments, the target feedback mode includes one of the following modes:

[0625] Mode one: performing one model or prediction inference to obtain first CSI on the first frequency domain resource, and feeding back the first CSI;

[0626] Mode two: performing one model inference or prediction to obtain second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the second CSI and a frequency domain position corresponding to the third frequency domain resource, or feeding back the second CSI and a frequency domain position corresponding to the second frequency domain resource;

[0627] The third mode: performing model inference or prediction once to obtain a second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the first CSI;

[0628] The fourth mode: performing model inference or prediction multiple times to obtain the first CSI, and feeding back the first CSI;

[0629] The fifth mode: after algorithm processing of model input, performing model inference or prediction on one or more groups of the second frequency domain resource, obtaining a third CSI after algorithm processing of the obtained model output, and feeding back the third CSI; wherein the third CSI includes the first CSI;

[0630] The sixth mode: performing codebook-based CSI feedback, or skipping this time of CSI feedback, or reporting that model inference or prediction on the terminal side cannot be completed.

[0631] In some embodiments, the apparatus of the embodiments of the present disclosure further includes:

[0632] A fourth sending unit, configured to send first information related to the model on the network device side to the terminal, the first information being used to determine a model on the terminal side that matches the model on the network device side.

[0633] In some embodiments, the apparatus of the embodiments of the present disclosure further includes:

[0634] A fourth receiving unit, configured to receive second information related to the model on the terminal side sent by the terminal;

[0635] A ninth processing unit, configured to determine a model on the terminal side that matches the model on the network device side according to the second information.

[0636] A tenth processing unit, configured to determine the first indication information based on the model on the terminal side that matches the model on the network device side.

[0637] In some embodiments, the first indication information includes:

[0638] Information indicating a subband size;

[0639] Information indicating whether CSI needs to be fed back for each subband;

[0640] Information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or indicating a sequence number of a starting subband in the BWP;

[0641] Information indicating a number of PRBs between two adjacent subbands;

[0642] Information indicating a number of PRBs or subbands;

[0643] information indicating a position of a starting PRB or sub-band in a BWP;

[0644] information indicating a gap between two adjacent PRBs or sub-bands;

[0645] information indicating a position of a starting PRB of each segment of consecutive PRBs in a BWP;

[0646] information indicating a position of a first PRB of each segment of starting sub-bands in a BWP;

[0647] information indicating a number of each segment of consecutive PRBs or sub-bands;

[0648] information indicating an output form of a terminal-side model.

[0649] In some embodiments, the CSI feedback information comprises: feedback CSI information; or, the feedback CSI information and at least one of the following:

[0650] identification information for determining a model used by a terminal;

[0651] information for indicating model-based feedback or codebook-based feedback;

[0652] information for indicating a second frequency domain resource corresponding to a model used by a terminal;

[0653] information for indicating an extra frequency domain resource in addition to the first frequency domain resource;

[0654] information for indicating all frequency domain resources corresponding to feedback CSI information.

[0655] In some embodiments, the third processing unit 1103 is specifically configured to:

[0656] when the CSI feedback information comprises information for indicating model-based feedback, recover CSI through a network device-side model according to the CSI feedback information;

[0657] when the CSI feedback information comprises information for indicating codebook-based feedback, recover CSI through a codebook according to the CSI feedback information.

[0658] In some embodiments, the third processing unit 1103 is specifically configured to:

[0659] when the feedback CSI information is first CSI on the first frequency domain resource, recover CSI through the network device-side model;

[0660] In a case where the fed back CSI information is the first CSI, the CSI is recovered by the model on the network device side based on the information indicating the second frequency domain resource corresponding to the model used by the terminal;

[0661] In a case where the fed back CSI information is CSI on the first frequency domain resource and an extra frequency domain resource in addition to the first frequency domain resource, the CSI is recovered by the model on the network side, and CSI on the extra frequency domain resource is deleted from the recovered CSI based on the information indicating the extra frequency domain resource in addition to the first frequency domain resource or the information indicating all the frequency domain resources corresponding to the fed back CSI information;

[0662] In a case where the fed back CSI is CSI on the first frequency domain resource and an extra frequency domain resource in addition to the first frequency domain resource, the CSI is recovered by the model on the network device side based on the information indicating the second frequency domain resource corresponding to the model used by the terminal, and CSI on the extra frequency domain resource is deleted from the recovered CSI based on the information indicating the extra frequency domain resource in addition to the first frequency domain resource or the information indicating all the frequency domain resources corresponding to the fed back CSI information.

[0663] It should be noted that the division of units in the embodiments of the present disclosure is illustrative, and is only a logical function division. In actual implementation, another division manner can be used. In addition, each functional unit in each embodiment of the present disclosure can be integrated in one processing unit, or each unit can be physically present alone, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0664] When the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, the integrated unit can be stored in a processor readable storage medium. Based on this understanding, the technical solutions of the present disclosure essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present disclosure. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0665] It should be noted that the above device provided by the embodiments of the present disclosure can realize all the method steps achieved by the method embodiments and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be repeated here.

[0666] In some embodiments of the present disclosure, a non-transitory readable storage medium is also provided, which stores a program for performing the CSI feedback method or the CSI recovery method described above.

[0667] The non-transitory readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to a magnetic storage (such as a floppy disk, a hard disk, a magnetic tape, a magneto optical disk (MO), etc.), an optical storage (such as a compact disk (CD), a digital video disk (DVD), a Blu-ray disk (BD), a high-definition versatile disk (HVD), etc.), and a semiconductor memory (such as a ROM, an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a non-volatile memory (NAND FLASH), a solid state disk (SSD), etc.).

[0668] The program is executed by the processor to realize all the implementation manners in the method applied to the terminal side as shown in FIG. 1 or the network device side embodiment as shown in FIG. 7, and thus the details are not repeated here.

[0669] The embodiments of the present disclosure also provide a computer program product, which includes computer instructions, and the computer instructions are executed by the processor to realize each process of the method embodiments shown in FIG. 1 or FIG. 7, and achieve the same technical effects. Therefore, the details are not repeated here.

[0670] The technical solutions provided by the embodiments of the present disclosure can be applied to various systems. For example, the applicable systems can be a global system of mobile communication (GSM) system, a code division multiple access (CDMA) system, a wideband code division multiple access (WCDMA) general packet radio service (GPRS) system, a long term evolution (LTE) system, an LTE frequency division duplex (FDD) system, an LTE time division duplex (TDD) system, a long term evolution advanced (LTE-A) system, a universal mobile system (UMTS), a worldwide interoperability for microwave access (WiMAX) system, a 5th generation mobile communication technology (5G) new radio (NR) system and an evolved communication system thereof, a 6th generation mobile communication technology (6G) system, and the like. The various systems all include terminal devices and network devices. The system can also include a core network part, for example, an evolved packet system (EPS), a 5G system (5GS), and the like.

[0671] The terminal device involved in the embodiments of the present disclosure can refer to a device that provides voice and / or data connectivity for a user, a handheld device with wireless connection function, or other processing devices connected to a wireless modem, etc. In different systems, the name of the terminal device can also be different, for example, in the 5G system or the 6G system, the terminal device can be called user equipment (User Equipment, UE). The wireless terminal device can be a USB storage device, other personal computer memory devices and a dongle, and can also communicate with one or more core networks (Core Network, CN) through a radio access network (Radio Access Network, RAN). The wireless terminal device can be a mobile terminal device, such as a mobile phone (or called "cellular" phone) and a computer with a mobile terminal device, for example, it can be a portable, pocket, handheld, computer built-in or vehicle-mounted mobile device, which exchanges language and / or data with the radio access network. For example, personal communication service (Personal Communication Service, PCS) phones, cordless phones, session initiation protocol (Session Initiated Protocol, SIP) phones, wireless local loop (Wireless Local Loop, WLL) stations, personal digital assistants (Personal Digital Assistant, PDA), personal computers, tablet computers, machine type communication (Machine-type Communication, MTC) terminal devices, etc. The wireless terminal device can also be called a system, a subscriber unit, a subscriber station, a mobile station, a mobile, a remote station, an access point, a remote terminal, an access terminal, a user terminal, a user agent, a user device, and a wireless access device and a router / modem that meet the limitations of the present definition, etc. The embodiments of the present disclosure are not limited.

[0672] The network device related to the embodiments of the present disclosure can be a base station, which can include a plurality of cells serving terminals. According to different application scenarios, the base station can also be referred to as an access point, or can be a device in an access network that communicates with wireless terminal devices through one or more sectors over an air interface, or other names. The network device can be used to exchange received air frames and Internet Protocol (IP) packets as a router between the wireless terminal device and the rest of the access network, which can include an Internet Protocol (IP) communication network. The network device can also coordinate the management of the properties of the air interface. For example, the network device related to the embodiments of the present disclosure can be a network device (Base Transceiver Station, BTS) in the Global System for Mobile Communications (GSM) or Code Division Multiple Access (CDMA), or a network device (NodeB) in Wide-band Code Division Multiple Access (WCDMA), or an evolved network device (evolutional Node B, eNB or e-NodeB) in a long term evolution (LTE) system, or a 5G base station (the next Generation Node B, gNB) in a next generation system, or a Home evolved Node B (HeNB), a relay node, a femto, a pico, a network test device, etc., which are not limited in the embodiments of the present disclosure. In some network structures, the network device can include a centralized unit (CU) node and a distributed unit (DU) node, and the centralized unit and the distributed unit can also be geographically separated.

[0673] The network device and the terminal device can each use one or more antennas for multi-input multi-output (MIMO) transmission, which can be single-user MIMO (SU-MIMO) or multiple-user MIMO (MU-MIMO). According to the form and number of root antenna combinations, the MIMO transmission can be two-dimensional MIMO (2D-MIMO), three-dimensional MIMO (3D-MIMO), full-dimensional MIMO (FD-MIMO), or massive-MIMO, and can also be diversity transmission or precoding transmission or beamforming transmission, etc.

[0674] Those skilled in the art will understand that the embodiments of the present disclosure can be provided as a method, a system, or a computer program product. Therefore, the present disclosure can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present disclosure can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage, etc.) containing computer-usable program code.

[0675] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer executable instructions. These computer executable instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a means for implementing the functions specified in one or more flows of the flowcharts and / or one or more blocks of the block diagrams.

[0676] These processor executable instructions can also be stored in a processor readable memory that can direct the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the processor readable memory produce a product including instruction means, which implements the functions specified in one or more flows of the flowcharts and / or one or more blocks of the block diagrams.

[0677] These processor-executable instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operational steps are performed on the computer or other programmable device to generate a computer-implemented process such that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more flows in the flowchart and / or one or more blocks in the block diagram.

[0678] In addition, it should be noted that in the apparatus and method of the present disclosure, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions of the present disclosure. Moreover, the steps of performing the above series of processes can naturally be executed in time sequence according to the order of description, but do not necessarily have to be executed in time sequence, and certain steps can be executed in parallel or independently of each other. It can be understood by those skilled in the art that all or any steps or components of the method and apparatus of the present disclosure can be implemented in hardware, firmware, software or a combination thereof in any computing device (including processors, storage media, etc.) or network of computing devices, which can be implemented by those skilled in the art using their basic programming skills after reading the description of the present disclosure.

[0679] It should be noted that it should be understood that the division of each module above is only a logical division of functions, and in actual implementation, it can be integrated into a physical entity in whole or in part, or physically separated. And these modules can all be implemented in the form of software called by a processing element; they can all be implemented in the form of hardware; or some modules can be implemented in the form of software called by a processing element, and some modules can be implemented in the form of hardware. For example, a certain module can be a separately established processing element, or it can be implemented in a certain chip of the above device, and in addition, it can be stored in the form of program code in the memory of the above device, and called and executed by a certain processing element of the above device to determine the function of the above module. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together or independently implemented. The processing element described herein can be an integrated circuit with signal processing capability. In the implementation process, each step of the above method or each of the above modules can be completed by the integrated logic circuit of hardware or the instruction of software in the processing element.

[0680] For example, each module, unit, subunit, or submodule can be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together to implement a system-on-a-chip (SOC).

[0681] The terms “first,” “second,” etc., used in this disclosure and in the claims are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this disclosure described herein may be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. Additionally, the use of “and / or” in the specification and claims indicates at least one of the connected objects, such as A and / or B and / or C, indicating seven possibilities: A alone, B alone, C alone, and both A and B, both B and C, both A and C, and A, B, and C. Similarly, the use of “at least one of A and B” in this specification and claims should be understood as “A alone, B alone, or both A and B.”

[0682] Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims of this disclosure and their equivalents, this disclosure is also intended to include such modifications and variations.

Claims

1. A CSI feedback method applied to a terminal, the method comprising: receiving first indication information sent by a network device, the first indication information being used to indicate first frequency domain resources; determining a target feedback mode of channel state information (CSI) feedback based on a frequency domain location relationship between the first frequency domain resources and at least one group of second frequency domain resources, the second frequency domain resources being frequency domain resources corresponding to a model at a terminal side, and at least one group of the second frequency domain resources corresponding to one model at the terminal side; determining CSI feedback information to be fed back to the network device based on the target feedback mode; and sending the CSI feedback information to the network device. The target feedback mode comprises one of the following modes: Mode 1: performing one model inference or prediction to obtain first CSI on the first frequency domain resources, and feeding back the first CSI; Mode 2: performing one model inference or prediction to obtain second CSI on second frequency domain resources, the second frequency domain resources comprising the first frequency domain resources and third frequency domain resources, and feeding back the second CSI and a frequency domain location corresponding to the third frequency domain resources, or feeding back the second CSI and a frequency domain location corresponding to the second frequency domain resources; Mode 3: performing one model inference or prediction to obtain second CSI on second frequency domain resources, the second frequency domain resources comprising the first frequency domain resources and third frequency domain resources, and feeding back the first CSI; Mode 4: performing multiple model inferences or predictions to obtain the first CSI, and feeding back the first CSI; Mode 5: performing model inference or prediction on one or more groups of the second frequency domain resources after algorithm processing of model input, obtaining third CSI after algorithm processing of the obtained model output, and feeding back the third CSI, wherein the third CSI comprises the first CSI; and Mode 6: performing codebook-based CSI feedback, or reporting that model inference or prediction at the terminal side cannot be completed, or skipping this time of CSI feedback. 3.The method of claim 1, further comprising: receiving first information related to a model at a network device side sent by the network device; and determining a model at a terminal side matched with the model at the network device side according to the first information. 4.The method of claim 1, further comprising: sending second information related to the model at the terminal side to the network device, the second information being used to determine a model at a terminal side matched with the model at the network device side. The first indication information comprises: information indicating a subband size; information indicating whether CSI needs to be fed back for each subband; information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or information representing a sequence number of a starting subband in the BWP; information indicating a number of PRBs between two adjacent subbands; information indicating a number of PRBs or subbands; information indicating a location of a starting PRB or subband in the BWP; information indicating an interval between two adjacent PRBs or subbands; information indicating a location of a starting PRB of each segment of continuous PRBs in the BWP; and information indicating a location of a starting subband of each segment of continuous subbands in the BWP. ​ 2. The method of claim 1, wherein, ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ 5. The method of claim 1, wherein, ​ ​ ​ ​ ​ ​ ​ ​ ​ information indicating a position of a first PRB of each segment sub-band in the BWP; information indicating a number of each segment continuous PRB or sub-band; information indicating an output form of a terminal side model.

6. The method of claim 2, wherein, determining a target feedback mode of channel state information, CSI, feedback based on a frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, comprises: in a case that a frequency domain position of the first frequency domain resource is completely same as a frequency domain position of a group of second frequency domain resources in the at least one group of second frequency domain resources, and a first preset condition is met, determining that the target feedback mode is the mode one. wherein the first preset condition comprises that a performance index of a first model at a terminal side is good, and / or a calculation capability of the terminal reaches a complexity of the first model, and the first model is a model corresponding to the group of second frequency domain resources whose frequency domain position is completely same as the frequency domain position of the first frequency domain resource.

7. The method of claim 6, further comprising: in a case that the frequency domain position of the first frequency domain resource is completely same as the frequency domain position of the group of second frequency domain resources in the at least one group of second frequency domain resources, and the first preset condition is not met, determining that the target feedback mode is the mode six.

8. The method of claim 2, wherein, determining a target feedback mode of channel state information, CSI, feedback based on a frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, comprises: in a case that a frequency domain position range of the first frequency domain resource is contained in a frequency domain position range of a target frequency domain resource, and a second preset condition is met, determining that the target feedback mode is the mode two or the mode three, and the target frequency domain resource is a group of second frequency domain resources in the at least one group of second frequency domain resources; wherein the second preset condition comprises at least one of the following: a performance index of a second model at a terminal side is good, and the second model is a model corresponding to the target frequency domain resource; a part corresponding to the first frequency domain resource can be obtained from an output of the second model; a network device side model can input the part corresponding to the first frequency domain resource output by the second model, and restore the first CSI; a first specified rule related to a terminal capability.

9. The method of claim 8, further comprising: in a case that the frequency domain position range of the first frequency domain resource is contained in the frequency domain position range of the target frequency domain resource, and the second preset condition is not met, determining that the target feedback mode is the mode six.

10. The method of claim 2, wherein, determining a target feedback mode of channel state information, CSI, feedback based on a frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, comprises: in a case that a frequency domain position range of the first frequency domain resource is combined by frequency domain position ranges of multiple groups of second frequency domain resources, and a third preset condition is met, determining that the target feedback mode is the mode four. wherein the third preset condition comprises that a performance index of a third model at a terminal side is good, and / or a second specified rule related to a terminal capability, and the third model is a model related to the multiple groups of second frequency domain resources.

11. The method of claim 10, further comprising: the frequency domain position range of the first frequency domain resource is combined by the frequency domain position ranges of multiple groups of second frequency domain resources, and in the case that the third preset condition is not met, determining that the target feedback mode is the mode six.

12. The method of claim 2, wherein, the target feedback mode of channel state information (CSI) feedback is determined based on the frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources. in the case that the frequency domain position of the first frequency domain resource does not match the frequency domain positions of all second frequency domain resources in the at least one group of second frequency domain resources and the fourth preset condition is met, determining that the target feedback mode is the mode five; wherein the fourth preset condition comprises at least one of the following: the performance index of the fourth model is good, and the fourth model is a model corresponding to a second frequency domain resource used by the terminal; the input of the fourth model is processed by an algorithm to obtain the CSI on the second frequency domain resource corresponding to the fourth model; the output of the fourth model is processed by an algorithm to obtain the third CSI; the computing capability of the terminal reaches the complexity of the fourth model; the processing delay of the fourth model meets the preset delay requirement; the performance loss value of the fourth model is less than the preset performance loss threshold; part of the first frequency domain resource can be obtained from the output of the fourth model.

13. The method of claim 12, further comprising: in the case that the frequency domain position of the first frequency domain resource does not match the frequency domain positions of all second frequency domain resources in the at least one group of second frequency domain resources and the fourth preset condition is not met, determining that the target feedback mode is the mode six.

14. The method of claim 1, wherein, the CSI feedback information comprises: the feedback CSI information; or the feedback CSI information and at least one of the following: identification information for determining the model used by the terminal; information for indicating model-based feedback or codebook-based feedback; information for indicating the second frequency domain resource corresponding to the model used by the terminal; information for indicating the redundant frequency domain resource in addition to the first frequency domain resource; information for indicating all the frequency domain resources corresponding to the feedback CSI information.

15. A CSI recovery method applied to a network device, the method comprising: sending first indication information to a terminal, the first indication information being used to indicate a first frequency domain resource, the first frequency domain resource being used for channel state information (CSI) feedback of the terminal; receiving CSI feedback information sent by the terminal; recovering the CSI according to the CSI feedback information; wherein the CSI feedback information is obtained by the terminal based on a target feedback mode; the target feedback mode is determined by the terminal based on the frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resource being a frequency domain resource corresponding to a model on the terminal side, and at least one group of the second frequency domain resources corresponding to one model on the terminal side.

16. The method of claim 15, wherein, the target feedback mode comprises one of the following modes: Manner one: performing a model or prediction inference to obtain a first CSI on the first frequency domain resource, and feeding back the first CSI; Manner two: performing a model inference or prediction to obtain a second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the second CSI and a frequency domain location corresponding to the third frequency domain resource, or feeding back the second CSI and a frequency domain location corresponding to the second frequency domain resource; Manner three: performing a model inference or prediction to obtain a second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the first CSI; Manner four: performing multiple model inferences or predictions to obtain the first CSI, and feeding back the first CSI; Manner five: after algorithm processing of model input, performing inference or prediction of a model on one or more groups of the second frequency domain resource, after algorithm processing of obtained model output, obtaining a third CSI, and feeding back the third CSI; wherein the third CSI includes the first CSI; Manner six: performing codebook-based CSI feedback, or skipping this time of CSI feedback because model inference or prediction at a terminal side cannot be completed.

17. The method of claim 15, further comprising: sending, to the terminal, first information related to the model at the network device side, the first information being used to determine a model at the terminal side that matches the model at the network device side.

18. The method of claim 15, further comprising: receiving second information related to the model at the terminal side and sent by the terminal; determining, according to the second information, a model at the terminal side that matches the model at the network device side; determining the first indication information based on the model at the terminal side that matches the model at the network device side.

19. The method of claim 15, wherein, The first indication information includes: information indicating a subband size; information indicating whether CSI needs to be fed back for each subband; information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or information indicating a sequence number of a starting subband in the BWP; information indicating a number of PRBs between two adjacent subbands; information indicating a number of PRBs or subbands; information indicating a position of a starting PRB or subband in the BWP; information indicating a gap between two adjacent PRBs or subbands; information indicating a position of a starting PRB of each segment of continuous PRBs in the BWP; information indicating a position of a first PRB of each segment of starting subbands in the BWP; information indicating a number of each segment of continuous PRBs or subbands; information indicating an output form of the model at the terminal side.

20. The method of claim 15, wherein, The CSI feedback information includes: fed-back CSI information; or, fed-back CSI information and at least one of the following: identification information used to determine a model used by the terminal; information used to indicate model-based feedback or codebook-based feedback; information used to indicate a second frequency domain resource corresponding to the model used by the terminal; information used to indicate a frequency domain resource in addition to the first frequency domain resource. information for indicating all frequency domain resources corresponding to the fed back CSI information.

21. The method of claim 20, wherein, The CSI is recovered according to the CSI feedback information, including: In a case where the CSI feedback information includes information for indicating model-based feedback, the CSI is recovered according to the CSI feedback information through a model at a network device side; In a case where the CSI feedback information includes information for indicating codebook-based feedback, the CSI is recovered according to the CSI feedback information through a codebook.

22. The method of claim 21, wherein, The CSI is recovered according to the CSI feedback information through a model at a network device side, including: In a case where the fed back CSI information is first CSI on the first frequency domain resource, the CSI is recovered through the model at the network device side; In a case where the fed back CSI information is the first CSI, the CSI is recovered through the model at the network device side based on the information for indicating the second frequency domain resource corresponding to the model used by the terminal; In a case where the fed back CSI information is CSI on the first frequency domain resource and on an extra frequency domain resource in addition to the first frequency domain resource, the CSI is recovered through the model at the network side, and CSI on the extra frequency domain resource is deleted from the recovered CSI based on the information for indicating the extra frequency domain resource in addition to the first frequency domain resource or the information for indicating all frequency domain resources corresponding to the fed back CSI information; In a case where the fed back CSI is CSI on the first frequency domain resource and on an extra frequency domain resource in addition to the first frequency domain resource, the CSI is recovered through the model at the network device side based on the information for indicating the second frequency domain resource corresponding to the model used by the terminal, and CSI on the extra frequency domain resource is deleted from the recovered CSI based on the information for indicating the extra frequency domain resource in addition to the first frequency domain resource or the information for indicating all frequency domain resources corresponding to the fed back CSI information.

23. A terminal comprising: a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; wherein the transceiver is configured to receive first indication information sent by a network device, the first indication information being used for indicating a first frequency domain resource; the processor is configured to read a program in the memory and perform the following process: determine a target feedback mode of channel state information (CSI) feedback based on a frequency domain location relationship between the first frequency domain resource and at least one set of second frequency domain resources, the second frequency domain resources being frequency domain resources corresponding to a model at a terminal side, and at least one set of the second frequency domain resources corresponding to one model at the terminal side; determine CSI feedback information fed back to the network device based on the target feedback mode; the transceiver is further configured to send the CSI feedback information to the network device.

24. The terminal of claim 23, wherein, The target feedback mode includes one of the following modes: Mode one: perform a model inference or prediction once to obtain first CSI on the first frequency domain resource, and feed back the first CSI; Manner two: performing model inference or prediction once to obtain a second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the second CSI and a frequency domain location corresponding to the third frequency domain resource, or feeding back the second CSI and a frequency domain location corresponding to the second frequency domain resource; Manner three: performing model inference or prediction once to obtain a second CSI on a second frequency domain resource, the second frequency domain resource including the first frequency domain resource and a third frequency domain resource, and feeding back the first CSI; Manner four: performing model inference or prediction multiple times to obtain the first CSI, and feeding back the first CSI; Manner five: after algorithm processing of model input, performing inference or prediction of a model on one or more groups of the second frequency domain resource, after algorithm processing of obtained model output, obtaining a third CSI, and feeding back the third CSI; wherein the third CSI includes the first CSI; Manner six: performing codebook-based CSI feedback, or skipping this time of CSI feedback because model inference or prediction at the terminal side cannot be completed.

25. The terminal according to claim 23, wherein the transceiver is further configured to receive first information related to the model at the network device side and transmitted by the network device; and the processor is further configured to determine a model at the terminal side that matches the model at the network device side according to the first information.

26. The terminal according to claim 23, wherein the transceiver is further configured to transmit second information related to the model at the terminal side to the network device, the second information being used to determine a model at the terminal side that matches the model at the network device side. The first indication information includes: information indicating a subband size; information indicating whether CSI needs to be fed back for each subband; 27. The terminal of claim 23, wherein, information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or information indicating a sequence number of a starting subband in the BWP; information indicating a number of PRBs between two adjacent subbands; information indicating a number of PRBs or subbands; information indicating a position of a starting PRB or subband in the BWP; information indicating an interval between two adjacent PRBs or subbands; information indicating a position of a starting PRB of each segment of continuous PRBs in the BWP; information indicating a position of a first PRB of each segment of starting subbands in the BWP; information indicating a number of each segment of continuous PRBs or subbands; information indicating an output form of the model at the terminal side.

28. The terminal according to claim 24, wherein the processor is further configured to, in a case where a frequency domain location of the first frequency domain resource is completely identical to a frequency domain location of a group of second frequency domain resources in the at least one group of second frequency domain resources and a first preset condition is met, determine that the target feedback manner is the manner one. The first preset condition includes that a performance index of a first model at the terminal side is good, and / or a computing capability of the terminal reaches a complexity of the first model, and the first model is a model corresponding to a second frequency domain resource whose frequency domain location is completely identical to that of the first frequency domain resource. ​ ​ ​ ​ 29. The terminal of claim 28, wherein, when the frequency domain location of the first frequency domain resource is completely same as the frequency domain location of a group of second frequency domain resources in the at least one group of second frequency domain resources, and the first preset condition is not met, the processor is further configured to determine that the target feedback mode is the sixth mode.

30. The terminal of claim 24, wherein, when the frequency domain location range of the first frequency domain resource is contained in the frequency domain location range of a target frequency domain resource, and a second preset condition is met, the processor is further configured to determine that the target feedback mode is the second mode or the third mode, the target frequency domain resource being a group of second frequency domain resources in the at least one group of second frequency domain resources. The performance index of a second model on the terminal side is good, the second model being a model corresponding to the target frequency domain resource; A part corresponding to the first frequency domain resource can be obtained from the output of the second model; A network device side model can input the part corresponding to the first frequency domain resource output by the second model, and restore the first CSI; A first specified rule related to the terminal capability.

31. The terminal of claim 30, wherein, when the frequency domain location range of the first frequency domain resource is contained in the frequency domain location range of the target frequency domain resource, and the second preset condition is not met, the processor is further configured to determine that the target feedback mode is the sixth mode.

32. The terminal of claim 24, wherein, when the frequency domain location range of the first frequency domain resource is combined from the frequency domain location ranges of multiple groups of second frequency domain resources, and a third preset condition is met, the processor is further configured to determine that the target feedback mode is the fourth mode. The performance index of a third model related to the multiple groups of second frequency domain resources is good, and / or a second specified rule related to the terminal capability.

33. The terminal of claim 32, wherein, when the frequency domain location range of the first frequency domain resource is combined from the frequency domain location ranges of the multiple groups of second frequency domain resources, and the third preset condition is not met, the processor is further configured to determine that the target feedback mode is the sixth mode.

34. The terminal of claim 24, wherein, when the frequency domain location of the first frequency domain resource does not match the frequency domain location of all second frequency domain resources in the at least one group of second frequency domain resources, and a fourth preset condition is met, the processor is further configured to determine that the target feedback mode is the fifth mode. The performance index of a fourth model is good, the fourth model being a model corresponding to the second frequency domain resource used by the terminal; The input of the fourth model is processed by an algorithm to obtain the CSI on the second frequency domain resource corresponding to the fourth model; The output of the fourth model is processed by an algorithm to obtain the third CSI; The computing capability of the terminal reaches the complexity of the fourth model. ​ ​ ​ ​ ​ ​ ​ ​ The processing delay of the fourth model satisfies a preset delay requirement. The performance loss value of the fourth model is less than a preset performance loss threshold. Part of the first frequency domain resource can be obtained from the output of the fourth model. 35.The terminal of claim 34, wherein, The processor is further configured to determine that the target feedback mode is the sixth mode when the frequency domain position of the first frequency domain resource does not match the frequency domain position of all the second frequency domain resources in the at least one group of second frequency domain resources, and the fourth preset condition is not satisfied.

36. The terminal of claim 23, wherein, The CSI feedback information comprises: feedback CSI information; or feedback CSI information and at least one of the following: Identification information of a model used by the terminal; Information indicating model-based feedback or codebook-based feedback; Information indicating a second frequency domain resource corresponding to the model used by the terminal; Information indicating an extra frequency domain resource in addition to the first frequency domain resource; Information indicating all frequency domain resources corresponding to the feedback CSI information. 37.A terminal, comprising: A first receiving unit configured to receive first indication information transmitted by a network device, the first indication information being used to indicate a first frequency domain resource; A first processing unit configured to determine a target feedback mode of channel state information (CSI) feedback based on a frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resource being a frequency domain resource corresponding to a model on the terminal side, and the at least one group of second frequency domain resources corresponding to one model on the terminal side; A second processing unit configured to determine CSI feedback information to be fed back to the network device based on the target feedback mode; A first transmitting unit configured to transmit the CSI feedback information to the network device.

38. A network device comprising: A transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor; wherein The transceiver is configured to transmit first indication information to a terminal, the first indication information being used to indicate a first frequency domain resource, the first frequency domain resource being used for channel state information (CSI) feedback of the terminal; Receive the CSI feedback information transmitted by the terminal; The processor is configured to read the program in the memory and perform the following process: Restore the CSI according to the CSI feedback information; The CSI feedback information is obtained by the terminal based on a target feedback mode; the target feedback mode is determined by the terminal based on a frequency domain position relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resource being a frequency domain resource corresponding to a model on the terminal side, and the at least one group of second frequency domain resources corresponding to one model on the terminal side.

39. The apparatus of claim 38, wherein, The target feedback mode comprises one of the following modes: Mode one: perform a model or a prediction inference to obtain first CSI on the first frequency domain resource, and feed back the first CSI; Manner two: performing model inference or prediction once to obtain second CSI on second frequency domain resources, the second frequency domain resources including the first frequency domain resources and third frequency domain resources, and feeding back the second CSI and the frequency domain location corresponding to the third frequency domain resources, or feeding back the second CSI and the frequency domain location corresponding to the second frequency domain resources; Manner three: performing model inference or prediction once to obtain second CSI on second frequency domain resources, the second frequency domain resources including the first frequency domain resources and third frequency domain resources, and feeding back the first CSI; Manner four: performing model inference or prediction multiple times to obtain the first CSI, and feeding back the first CSI; Manner five: after algorithm processing of model input, performing model inference or prediction on one or more groups of the second frequency domain resources, after algorithm processing of the obtained model output, obtaining third CSI, and feeding back the third CSI; wherein the third CSI includes the first CSI; Manner six: performing codebook-based CSI feedback, or skipping this time of CSI feedback because model inference or prediction at the terminal side cannot be completed.

40. The device of claim 38, wherein the transceiver is further configured to send, to the terminal, first information related to the model at the network device side, the first information being used to determine a model at the terminal side that matches the model at the network device side.

41. The device of claim 38, wherein the transceiver is further configured to receive, from the terminal, second information related to the model at the terminal side. The processor is further configured to determine, based on the second information, a model at the terminal side that matches the model at the network device side. The first indication information is determined based on the model at the terminal side that matches the model at the network device side.

42. The apparatus of claim 38, wherein, The first indication information includes: information indicating a subband size; information indicating whether CSI needs to be fed back for each subband; information indicating a sequence number of a starting physical resource block (PRB) in a bandwidth part (BWP) or information indicating a sequence number of a starting subband in the BWP; information indicating a number of PRBs between two adjacent subbands; information indicating a number of PRBs or subbands; information indicating a location of a starting PRB or subband in the BWP; information indicating a gap between two adjacent PRBs or subbands; information indicating a location of a starting PRB of each segment of continuous PRBs in the BWP; information indicating a location of a first PRB of each segment of starting subbands in the BWP; information indicating a number of each segment of continuous PRBs or subbands; information indicating an output form of the model at the terminal side.

43. The apparatus of claim 38, wherein, The CSI feedback information includes: fed-back CSI information; or the fed-back CSI information and at least one of the following: identification information used to determine a model used by the terminal; information used to indicate model-based feedback or codebook-based feedback; information used to indicate second frequency domain resources corresponding to the model used by the terminal; information used to indicate frequency domain resources in addition to the first frequency domain resources; information used to indicate all frequency domain resources corresponding to the fed-back CSI information.

44. The apparatus of claim 43, wherein, The processor is further configured to recover, according to the CSI feedback information, the CSI by a model at a network device side in a case where the CSI feedback information comprises information indicating model-based feedback. The processor is further configured to recover, according to the CSI feedback information, the CSI by a codebook in a case where the CSI feedback information comprises information indicating codebook-based feedback.

45. The apparatus of claim 43, wherein, The processor is further configured to recover, according to the CSI feedback information, the CSI by a model at a network device side in a case where the CSI feedback information comprises information indicating model-based feedback. The processor is further configured to recover, according to the CSI feedback information, the CSI by a model at a network device side in a case where the CSI feedback information comprises information indicating model-based feedback. The processor is further configured to recover, according to the CSI feedback information, the CSI by a model at a network device side in a case where the CSI feedback information comprises information indicating model-based feedback. The processor is further configured to recover, according to the CSI feedback information, the CSI by a model at a network device side in a case where the CSI feedback information comprises information indicating model-based feedback. The processor is further configured to recover, according to the CSI feedback information, the CSI by a model at a network device side in a case where the CSI feedback information comprises information indicating model-based feedback. 46.A network device, comprising: a second sending unit configured to send, to a terminal, first indication information, the first indication information being used to indicate a first frequency domain resource, the first frequency domain resource being used for channel state information (CSI) feedback of the terminal; a second receiving unit configured to receive CSI feedback information sent by the terminal; a third processing unit configured to recover, according to the CSI feedback information, the CSI; wherein the CSI feedback information is obtained by the terminal based on a target feedback mode; 47. A non-transitory readable storage medium, wherein, the target feedback mode is determined by the terminal based on a frequency domain location relationship between the first frequency domain resource and at least one group of second frequency domain resources, the second frequency domain resource being a frequency domain resource corresponding to a model at a terminal side, and at least one group of the second frequency domain resources corresponding to one model at the terminal side. The non-transitory readable storage medium stores a program, and the program is used to execute the CSI feedback method in any one of claims 1 to 14, or execute the CSI recovery method in any one of claims 15 to 22.

Images