Model selection method and apparatus, communication device and readable storage medium

Abstract

The present disclosure belongs to the field of wireless technology. Provided are a model selection method and apparatus, and a communication device and a readable storage medium. The model selection method in the embodiments of the present disclosure comprises: a terminal acquiring a first channel statistical parameter of a target scenario; sending the first channel statistical parameter to a base station; and receiving a target model fed back by the base station, wherein the target model is a model of the target scenario that is selected on the basis of the first channel statistical parameter.

Description

Model selection method, device, communication equipment and readable storage medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This disclosure claims priority to Chinese Patent Application No. 202410398565.3 filed in China on April 3, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure belongs to the field of wireless technology, and particularly relates to a model selection method, apparatus, communication equipment, and readable storage medium. Background Art

[0004] Related technologies typically require measuring channel data for various communication scenarios, such as indirect, direct, complex, and simple scattering. These data sets are then constructed to train corresponding models. Selecting the required model typically requires the terminal to feed back a large amount of downlink channel data to the base station, which then performs model selection. This results in significant feedback overhead. Summary of the Invention

[0005] The purpose of the embodiments of the present disclosure is to provide a model selection method, apparatus, communication device, and readable storage medium to solve the problem of high feedback overhead when selecting a model in related technologies.

[0006] In order to solve the above technical problems, the present disclosure is implemented as follows:

[0007] In a first aspect, a model selection method is provided, which is applied to a terminal and includes:

[0008] The terminal obtains a first channel statistical parameter of the target scene;

[0009] The terminal sends the first channel statistical parameter to the base station;

[0010] The terminal receives the target model fed back by the base station, wherein the target model is a model of the target scenario selected according to the first channel statistical parameter.

[0011] Optionally, the method further includes:

[0012] The terminal receives the configuration information sent by the base station;

[0013] The configuration information is used to configure at least one of the following:

[0014] Channel statistical parameters that need to be fed back;

[0015] Measurement parameters;

[0016] Feedback period of channel statistical parameters;

[0017] Trigger threshold value of feedback channel statistical parameters;

[0018] Constructed channel statistics parameters for multiple scenarios.

[0019] Optionally, obtaining the first channel statistical parameter of the target scene includes:

[0020] Perform channel measurement in the target scenario according to the configured measurement parameters to obtain channel data;

[0021] Channel parameters are extracted from the channel data, and statistical analysis is performed on the extracted channel parameters to obtain the first channel statistical parameters.

[0022] Optionally, the first channel statistical parameter includes at least one of the following:

[0023] Rice K factor;

[0024] Multipath delay spread value;

[0025] A multipath angle extension value, wherein the multipath angle extension value includes at least one of the following: a transmission horizontal angle extension value, a transmission vertical angle extension value, a reception horizontal angle extension value, and a reception vertical angle extension value;

[0026] Intra-cluster multipath delay spread value;

[0027] Intra-cluster multipath angle spread value, the intra-cluster multipath angle spread value comprising at least one of the following: a transmission horizontal angle spread value, a transmission vertical angle spread value, a reception horizontal angle spread value, and a reception vertical angle spread value;

[0028] Doppler shift.

[0029] Optionally, the terminal sending the first channel statistical parameter to the base station includes at least one of the following:

[0030] The terminal periodically sends the first channel statistical parameter to the base station;

[0031] When a similarity value between the first channel statistical parameter and a network configuration or a predefined channel statistical parameter corresponding to the target scenario is greater than or equal to a trigger threshold, the terminal sends the first channel statistical parameter to a base station.

[0032] Optionally, the method further includes:

[0033] The terminal sends channel statistical parameters of multiple scenarios to the base station;

[0034] The channel statistical parameters of the multiple scenarios are used to be divided into N types of channel statistical parameters, and N data sets are obtained according to the N types of channel statistical parameters, and corresponding models are trained according to each of the N data sets.

[0035] In a second aspect, a model selection method is provided, which is applied to a base station and includes:

[0036] The base station receives a first channel statistical parameter of a target scene sent by the terminal;

[0037] The base station selects, from a pre-constructed data set, a second channel statistical parameter that is most similar to the first channel statistical parameter;

[0038] The base station selects a model corresponding to the second channel statistical parameter as a target model of the target scene;

[0039] The base station sends the target model to the terminal.

[0040] Optionally, the method further includes:

[0041] The base station sends configuration information to the terminal;

[0042] The configuration information is used to configure at least one of the following:

[0043] Channel statistical parameters that need to be fed back;

[0044] Measurement parameters;

[0045] Feedback period of channel statistical parameters;

[0046] Trigger threshold value of feedback channel statistical parameters;

[0047] Constructed channel statistics parameters for multiple scenarios.

[0048] Optionally, the first channel statistical parameter includes at least one of the following:

[0049] Rice K factor;

[0050] Multipath delay spread value;

[0051] A multipath angle extension value, wherein the multipath angle extension value includes at least one of the following: a transmission horizontal angle extension value, a transmission vertical angle extension value, a reception horizontal angle extension value, and a reception vertical angle extension value;

[0052] Intra-cluster multipath delay spread value;

[0053] Intra-cluster multipath angle spread value, the intra-cluster multipath angle spread value comprising at least one of the following: a transmission horizontal angle spread value, a transmission vertical angle spread value, a reception horizontal angle spread value, and a reception vertical angle spread value;

[0054] Doppler shift.

[0055] Optionally, the method further includes:

[0056] receiving channel statistical parameters of multiple scenarios sent by at least one terminal;

[0057] Dividing the channel statistical parameters of the multiple scenarios into N types of channel statistical parameters according to the pre-divided number N of sub-scenarios, and obtaining N data sets based on the N types of channel statistical parameters, each data set corresponding to a sub-scenario, where N is an integer greater than 1;

[0058] According to each of the N data sets, a model of each of the N sub-scenes is obtained by training respectively.

[0059] Optionally, if the channel statistical parameters include K type parameters, the kth type parameter in the K type parameters is k sub-scenes, n k ≥1, 1≤k≤K, then:

[0060] Optionally, dividing the channel statistical parameters of the multiple scenarios into N types of channel statistical parameters according to the pre-divided number N of sub-scenarios includes:

[0061] According to the number N of pre-divided sub-scenarios, the channel statistical parameters of the multiple scenarios are classified to obtain the N types of channel statistical parameters.

[0062] In a third aspect, a model selection device is provided, which is applied to a terminal and includes:

[0063] An acquisition module, configured to acquire statistical parameters of a first channel of a target scene;

[0064] A first sending module, configured to send the first channel statistical parameter to a base station;

[0065] A first receiving module is configured to receive a target model fed back by the base station, wherein the target model is a model of the target scenario selected according to the first channel statistical parameter.

[0066] In a fourth aspect, a model selection device is provided, which is applied to a base station and includes:

[0067] A third receiving module, configured to receive a first channel statistical parameter of a target scenario sent by a terminal;

[0068] a first selection module, configured to select a second channel statistical parameter that is most similar to the first channel statistical parameter from a pre-constructed data set;

[0069] A second selection module, configured to select a model corresponding to the second channel statistical parameter as a target model of the target scene;

[0070] The third sending module is configured to send the target model to the terminal.

[0071] In a fifth aspect, a communication device is provided, comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method described in the first aspect or the steps of the method described in the second aspect.

[0072] In a sixth aspect, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method described in the first aspect or the steps of the method described in the second aspect are implemented.

[0073] In a seventh aspect, a computer program product is provided, comprising computer instructions, which, when executed by a processor, implement the steps of the method described in the first aspect or the steps of the method described in the second aspect.

[0074] Through the solution in the embodiments of the present disclosure, after entering a target scenario, a terminal can obtain first channel statistical parameters for the target scenario, send these first channel statistical parameters to a base station, and receive feedback from the base station about a target model for the target scenario selected based on the first channel statistical parameters. Thus, when selecting the desired model, the terminal need not feed back a large amount of downlink channel data to the base station; instead, it only needs to feed back the channel statistical parameters for the target scenario, thereby reducing feedback overhead. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] FIG1 is a flow chart of a model selection method provided by an embodiment of the present disclosure;

[0076] FIG2 is a schematic diagram of a model selection process in an embodiment of the present disclosure;

[0077] FIG3 is a flow chart of another model selection method provided by an embodiment of the present disclosure;

[0078] FIG4 is a schematic diagram of a model training process in an embodiment of the present disclosure;

[0079] FIG5A is a schematic diagram of model training and selection in an embodiment of the present disclosure;

[0080] FIG5B is a schematic diagram of data comparison in an embodiment of the present disclosure;

[0081] FIG6 is a schematic structural diagram of a model selection device provided by an embodiment of the present disclosure;

[0082] FIG7 is a schematic structural diagram of another model selection device provided by an embodiment of the present disclosure;

[0083] FIG8 is a schematic structural diagram of a communication device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0084] The following will clearly and completely describe the technical solutions in the embodiments of the present disclosure in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present disclosure without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0085] The terms "first", "second", etc. in the specification and claims of the present disclosure are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present disclosure can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects related to each other are in an "or" relationship.

[0086] The model selection method, apparatus, communication device, and readable storage medium provided by the embodiments of the present disclosure are described in detail below with reference to specific embodiments and their application scenarios in conjunction with the accompanying drawings.

[0087] Please refer to FIG1 , which is a flowchart of a model selection method provided by an embodiment of the present disclosure. The method is applied to a terminal. As shown in FIG1 , the method includes the following steps:

[0088] Step 11: The terminal obtains statistical parameters of the first channel of the target scene;

[0089] Step 12: The terminal sends the first channel statistical parameter to the base station;

[0090] Step 13: The terminal receives the target model fed back by the base station, where the target model is a model of the target scenario selected according to the first channel statistical parameters.

[0091] In the embodiment of the present disclosure, the target scene may be a scene where the terminal enters or is to be deployed, which may be, but is not limited to, an indoor communication scene, an outdoor communication scene, an indirect scene, a direct scene, a complex scattering scene, a simple scattering scene, etc.

[0092] The first channel statistical parameter may include multiple types of parameters, including but not limited to at least one of the following: Rice's K factor (abbreviated as KF), multipath delay spread value (such as (Delay Spread, DS)), multipath angle spread value (such as (Angular Spread, AS)), intra-cluster multipath delay spread value, intra-cluster multipath angle spread value, Doppler shift, etc. The multipath angle spread value may include but not limited to at least one of the following: transmit horizontal angle spread value, transmit vertical angle spread value, receive horizontal angle spread value, receive vertical angle spread value, etc. The intra-cluster multipath angle spread value may include but not limited to at least one of the following: transmit horizontal angle spread value, transmit vertical angle spread value, receive horizontal angle spread value, receive vertical angle spread value, etc.

[0093] The channel statistical parameters to be obtained and fed back to the base station can be determined based on predefined or preset rules, or based on network configuration, and there is no limitation on this. For example, the base station can configure the channel statistical parameters that need to be fed back for the terminal. The channel statistical parameters that need to be fed back include, but are not limited to, at least one of the following: Rice's K factor, multipath delay spread value, multipath angle spread value, intra-cluster multipath delay spread value, intra-cluster multipath angle spread value, Doppler shift, etc. For another example, the relevant configuration information can be sent down through the control channel.

[0094] The target model may be trained based on a pre-constructed data set, wherein relevant channel statistical parameters in the data set are reported to the base station by one or more terminals. The embodiments of the present disclosure do not limit the specific model type / function, as long as it is trained based on channel statistical parameters.

[0095] The target model may be an artificial intelligence (AI) model.

[0096] In an optional embodiment, after receiving the target model fed back by the base station, the terminal may deploy the target model on the terminal side for subsequent use.

[0097] In an optional embodiment, when receiving the target model fed back by the base station, the terminal may receive an identifier (such as a number, etc.) of the target model sent by the base station. For example, the identifier of the target model may be sent via a control channel.

[0098] Through the solution in the embodiments of the present disclosure, after entering a target scenario, a terminal can obtain first channel statistical parameters for the target scenario, send the first channel statistical parameters to a base station, and receive feedback from the base station on a target model for the target scenario selected based on the first channel statistical parameters. As a result, when selecting the desired model, the terminal need not feed back a large amount of downlink channel data to the base station; instead, it only needs to feed back the channel statistical parameters for the target scenario, thereby reducing feedback overhead.

[0099] In the embodiment of the present disclosure, the first channel statistical parameter can be obtained by channel measurement. The obtaining of the first channel statistical parameter of the target scene may include:

[0100] The terminal performs channel measurement in the target scenario according to the configured measurement parameters to obtain channel data;

[0101] The terminal extracts channel parameters from the channel data, and performs statistical analysis on the extracted channel parameters to obtain the first channel statistical parameters.

[0102] Optionally, the channel data mainly includes a channel matrix. When extracting channel parameters, it is mainly necessary to extract parameters such as multipath delay, multipath angle, and multipath cluster (such as intra-cluster multipath delay and intra-cluster multipath angle).

[0103] Exemplarily, the channel data may include a frequency domain channel matrix H(f) or a time domain channel matrix H(t), where H(f)=N r *N t *N sb , H(t)=N r *N t *N tau ; N r is the number of receiving antenna ports, N t is the number of transmitting antenna ports, N sb is the number of subbands, N tau is the delay number.

[0104] Signal processing can be performed on the frequency domain channel matrix H(f) or the time domain channel matrix H(t) to obtain the channel parameters of the measured channel data. Exemplarily, the channel impulse response (CIR) can be applied to extract angle information during channel modeling. In an embodiment of the present disclosure, multipath angle information can be obtained by transforming the port domain into the angle domain; in one example, a two-dimensional discrete Fourier transform (DFT) can be performed on the frequency domain channel matrix H(f) to obtain a matrix H(s). In this way, the frequency domain channel matrix can be transformed into the multipath angle domain of the receiving end and the transmitting end, and the matrix H(s) = UH *H(f)*V, where U is the size of N r *N r The spatial domain beam basis vector, V is the size of N t *N t The spatial domain beam basis vectors are , and the superscript H represents the conjugate transpose.

[0105] It can be seen that the size of H(s) is N r *N t *N sb If you need to get the receiving end angle, just stack H(s) to N r dimension; for example, we can first calculate H(s) in N sb Dimensions are summed, and then, for N sb The sum of the dimensions is N t By summing the dimensions, we can get the angle information of the receiving end.

[0106] For example, the multipath delay can be extracted based on the delay channel, and then the multipath power information can be extracted according to formula (1). n )=||h(τ n )|| 2 ,n=1,...,N (1)

[0107] Among them, h(τ n ) represents the time domain channel matrix, τ n represents the delay of the nth multipath, P(τ n ) is the power of the nth multipath, and N is the number of multipaths.

[0108] Illustratively, the method for extracting the intra-cluster multipath delay is similar to the method for extracting the multipath delay, and the method for extracting the intra-cluster multipath angle is similar to the method for extracting the multipath angle, which will not be described in detail here.

[0109] Exemplarily, after extracting the channel parameters, the channel statistical parameters can be obtained by performing statistical feature analysis on the channel parameters. In one example, the channel statistical parameters may include a multipath angle spread value, a multipath delay spread value, etc. Here, the multipath angle spread value may be a multipath transmission horizontal angle spread value, a multipath reception horizontal angle spread value, a multipath transmission vertical angle spread value, or a multipath reception vertical angle spread value; the multipath angle spread value σ can be calculated according to formula (2). AS .

[0110] Where L represents the number of channel multipaths, P l represents the multipath power of the lth channel, θ l,uRepresents the angle difference between the multipath angle of the lth channel and the angle mean; here, the angle can be selected as the transmitting horizontal angle, the receiving horizontal angle, the transmitting vertical angle or the receiving vertical angle.

[0111] In addition, the multipath delay spread value τ can be calculated according to formula (3): rms .

[0112] Among them, τ l represents the multipath delay of the lth channel, τ mean Indicates the mean multipath delay.

[0113] In the disclosed embodiments, since the channel statistical parameters fed back by the terminal are used for model detection / selection, there is no need to provide frequent feedback, as with Channel State Information (CSI). Therefore, the terminal can periodically feed back the channel statistical parameters, or trigger feedback of the channel statistical parameters.

[0114] Optionally, sending the first channel statistical parameter to the base station may include at least one of the following:

[0115] 1) periodically sending the first channel statistical parameter to a base station;

[0116] The feedback period in 1) can be set to a longer time, such as 1s, 2s, etc., to reduce signaling overhead. The feedback period can also be predefined and can be configured by the network for the terminal, such as the feedback period of the channel statistical parameters configured by the base station.

[0117] Optionally, in 1), the first channel statistical parameter may be sent to the base station via a control channel or a shared channel.

[0118] 2) When a similarity value between the first channel statistical parameter and a network configuration or a predefined channel statistical parameter corresponding to the target scenario is greater than or equal to a trigger threshold, the first channel statistical parameter is sent to a base station.

[0119] The trigger threshold value in 2) can be a threshold value obtained through a large number of experiments to ensure feedback accuracy in a specific scenario, or it can be pre-set or network configured, such as the trigger threshold value of the feedback channel statistical parameter configured by the base station.

[0120] For the network configuration or predefined channel statistical parameters, it can be a set of channel statistical parameters or a collection of channel statistical parameters, which includes multiple sets of channel statistical parameters. When calculating the similarity value between the first channel statistical parameter and the network configuration or predefined channel statistical parameters, it can be to perform a similarity operation with the network configuration or a predefined set of channel statistical parameters, and trigger the reporting of the channel statistical parameters when the obtained similarity value is greater than or equal to the trigger threshold value, otherwise the reporting of the channel statistical parameters is not triggered; it can also be to perform a similarity operation with a set of channel statistical parameters in the network configuration or predefined multiple sets of channel statistical parameters, and trigger the reporting of the channel statistical parameters when the obtained similarity value is greater than or equal to the trigger threshold value, otherwise the reporting of the channel statistical parameters is not triggered; it can also be to perform a similarity operation with each set of channel statistical parameters in the network configuration or predefined multiple sets of channel statistical parameters, and trigger the reporting of the channel statistical parameters when some / all values ​​in the obtained similarity value are greater than or equal to the trigger threshold value, otherwise the reporting of the channel statistical parameters is not triggered; the specific triggering conditions can be set based on actual needs.

[0121] For example, assuming that the first channel statistical parameter fed back by the terminal includes n channel statistical parameters, the similarity calculation formula below can be used to perform similarity calculation: β = α1*p1+...+α s *p s +…+α n *p n , 1≤s≤n

[0122] Among them, p s Represents the similarity metric of the s-th channel statistical parameter; for example, if the s-th channel statistical parameter is the multipath delay spread value DS, then Where DS′ is the network configuration or predefined multipath delay spread value. s represents the weight coefficient, and α1+…+α s …+α n = 1. The smaller the β value, the more similar the corresponding channel statistical parameters are.

[0123] Exemplarily, the channel statistical parameters configured by the network may be channel statistical parameters in a data set obtained from a base station for training a corresponding model, so that the terminal determines whether the corresponding similarity value is greater than or equal to a trigger threshold value.

[0124] Optionally, the model selection method in the embodiment of the present disclosure may further include:

[0125] The terminal receives configuration information sent by the base station, where the configuration information is used to configure at least one of the following: measurement parameters, channel statistical parameters to be fed back, channel parameter weight coefficients, feedback periods of channel statistical parameters, trigger thresholds for feedback channel statistical parameters, and channel statistical parameters constructed for multiple scenarios.

[0126] It should be noted that for measurement parameters, corresponding measurement parameters can be configured for different scenarios, or they can be configured uniformly regardless of the scenario. For channel statistical parameters that require feedback, corresponding channel statistical parameters that require feedback can be configured for different scenarios, or they can be configured uniformly regardless of the scenario. The channel statistical parameters constructed for multiple scenarios can be used to determine whether to trigger reporting of the channel statistical parameters.

[0127] Optionally, the model selection method in the embodiment of the present disclosure may further include:

[0128] The terminal sends channel statistical parameters of multiple scenarios to the base station; wherein the channel statistical parameters of the multiple scenarios are used to be divided into N types of channel statistical parameters, and N data sets are obtained according to the N types of channel statistical parameters, and a corresponding model is obtained by training each data set in the N data sets.

[0129] Optionally, if the channel statistical parameters of the multiple scenarios include K-type parameters, the k-th parameter in the K-type parameters is equal to n k sub-scenes, that is, based on the k-th parameter, the corresponding scene can be divided into n k Class (divided into n k sub-scenes), n k ≥1, 1≤k≤K, then: The number of sub-scenes that can be divided based on each type of parameter can be determined by actual measurements of multiple scenes and comprehensive consideration of factors such as acquisition overhead.

[0130] For example, as shown in Figure 2, the specific model selection process may include:

[0131] Step 21: The base station sends configuration information to the terminal. This configuration information is used to configure measurement parameters (or measurement quantities) and channel statistical parameters that require feedback for the terminal. This configuration information can be sent via a control channel. The channel statistical parameters that require feedback may include, but are not limited to, at least one of the following: Ricean K factor, multipath delay spread value, multipath angle spread value, intra-cluster multipath delay spread value, intra-cluster multipath angle spread value, Doppler shift, etc.

[0132] Step 22: The terminal performs channel data collection, channel parameter extraction, and statistical analysis of channel statistical parameters in sequence; the collected channel data mainly includes the channel matrix; the extracted channel parameters mainly include multipath delay, multipath angle, multipath cluster (such as intra-cluster multipath delay, intra-cluster multipath angle), etc.; the channel statistical parameters mainly include but are not limited to at least one of the following: Rice's K factor, multipath delay spread value, multipath angle spread value, intra-cluster multipath delay spread value, intra-cluster multipath angle spread value, Doppler frequency shift, etc.

[0133] Step 23: The terminal feeds back the channel statistical parameters of a specific scenario to the base station. For example, the specific scenario may be the cell environment where the terminal is to be deployed.

[0134] Optionally, the terminal may feed back the channel statistical parameters of a specific scenario to the base station periodically, or may feed back the channel statistical parameters of a specific scenario to the base station when a trigger condition is met.

[0135] Step 24: The base station selects a pre-constructed data set that is closest to / most similar to the received channel statistical parameters;

[0136] Step 25: The base station selects and deploys the model corresponding to the dataset.

[0137] Step 26: The base station sends the identifier (such as a number) of the selected model to the terminal, and the terminal deploys the model. For example, the identifier of the model can be sent via a control channel.

[0138] Therefore, after the terminal enters a certain scene, it can analyze the channel statistical parameters through channel measurement statistics and report the channel statistical parameters to the network side; the network side can compare the received channel statistical parameters with the channel statistical parameters in the pre-constructed data set to select the closest data set and the corresponding model, such as an AI model.

[0139] Please refer to FIG3 , which is a flow chart of a model selection method provided by an embodiment of the present disclosure. The method is applied to a base station. As shown in FIG3 , the method includes the following steps:

[0140] Step 31: The base station receives a first channel statistical parameter of a target scene sent by the terminal;

[0141] Step 32: The base station selects a second channel statistical parameter that is most similar to the first channel statistical parameter from the pre-constructed data set;

[0142] Step 33: The base station selects a model corresponding to the second channel statistical parameter as a target model for the target scene;

[0143] Step 34: The base station sends the target model to the terminal.

[0144] In the embodiment of the present disclosure, the target scene can be a scene that the terminal enters or is to be deployed, which can be selected from but not limited to indoor communication scenes, outdoor communication scenes, indirect scenes, direct scenes, complex scattering scenes, simple scattering scenes, dynamic scattering scenes, static scattering scenes, etc.

[0145] The first channel statistical parameter may include multiple types of parameters, including but not limited to at least one of the following: Rice's K factor (abbreviated as KF), multipath delay spread value (such as DS), multipath angle spread value (such as AS), intra-cluster multipath delay spread value, intra-cluster multipath angle spread value, Doppler shift, etc. The multipath angle spread value may include but not limited to at least one of the following: transmit horizontal angle spread value, transmit vertical angle spread value, receive horizontal angle spread value, receive vertical angle spread value, etc. The intra-cluster multipath angle spread value may include but not limited to at least one of the following: transmit horizontal angle spread value, transmit vertical angle spread value, receive horizontal angle spread value, receive vertical angle spread value, etc.

[0146] The second channel statistical parameters are specifically channel statistical parameters in a pre-constructed data set. The second channel statistical parameters may include multiple types of parameters, including but not limited to at least one of the following: Rice's K factor (abbreviated as KF), multipath delay spread value (such as DS), multipath angle spread value (such as AS), intra-cluster multipath delay spread value, intra-cluster multipath angle spread value, Doppler shift, etc. The multipath angle spread value may include but not limited to at least one of the following: transmission horizontal angle spread value, transmission vertical angle spread value, reception horizontal angle spread value, reception vertical angle spread value, etc. The intra-cluster multipath angle spread value may include but not limited to at least one of the following: transmission horizontal angle spread value, transmission vertical angle spread value, reception horizontal angle spread value, reception vertical angle spread value, etc.

[0147] The channel statistical parameters to be obtained and fed back to the base station can be determined based on predefined or preset rules, or based on network configuration, and there is no limitation on this. For example, the base station can configure the channel statistical parameters that need to be fed back for the terminal. The channel statistical parameters that need to be fed back include, but are not limited to, at least one of the following: Rice's K factor, multipath delay spread value, multipath angle spread value, intra-cluster multipath delay spread value, intra-cluster multipath angle spread value, Doppler shift, etc. For another example, the relevant configuration information can be sent down through the control channel.

[0148] The target model may be trained based on a pre-constructed data set, wherein relevant channel statistical parameters in the data set are reported to the base station by one or more terminals. The embodiments of the present disclosure do not limit the specific model type / function, as long as it is trained based on channel statistical parameters.

[0149] The target model may be an artificial intelligence (AI) model.

[0150] In an optional embodiment, assuming that the first channel statistical parameter fed back by the terminal includes n channel statistical parameters, the following similarity calculation formula can be used to select the second channel statistical parameter most similar to the first channel statistical parameter: β = α1*p1+…+α s *p s +…+α n *p n , 1≤s≤n

[0151] Among them, p s Represents the similarity metric of the s-th channel statistical parameter; for example, if the s-th channel statistical parameter is the multipath delay spread value DS, then Among them DS ′ is the multipath delay spread value in the pre-constructed data set. s represents the weight coefficient, and α1+…+α s …+α n = 1. The smaller the β value, the more similar the corresponding channel statistical parameters are. In this way, with the help of the similarity parameter β, model selection can be achieved in a "quantized" manner.

[0152] It should be noted that the weight coefficients mentioned above can be defined as the same, or they can be set differently based on different scenarios. For example, indoor scenarios are more concerned with scattering complexity, so the AS / DS weight is higher; while outdoor scenarios are more concerned with the KF factor and cluster variance parameters, so the KF and cluster variance parameters are given higher weights.

[0153] In an optional embodiment, when sending the target model to the terminal, the base station may send an identifier (such as a number, etc.) of the target model to the terminal. For example, the identifier of the target model may be sent via a control channel.

[0154] Through the solution in the embodiments of the present disclosure, a first channel statistical parameter for a target scenario sent by a terminal can be received, a second channel statistical parameter that is most similar to the first channel statistical parameter can be selected from a pre-constructed data set, and a model corresponding to the second channel statistical parameter can be selected as the target model for the target scenario, which can then be sent to the terminal. As a result, when selecting the required model, the terminal does not need to feed back a large amount of downlink channel data to the base station; instead, it only needs to feed back the channel statistical parameters for the target scenario, thereby reducing feedback overhead.

[0155] Optionally, the model selection method in the embodiment of the present disclosure may further include:

[0156] The base station sends configuration information to the terminal, where the configuration information is used to configure at least one of the following: measurement parameters, channel statistical parameters to be fed back, channel parameter weight coefficients, feedback cycles of channel statistical parameters, and trigger thresholds for feedback of channel statistical parameters.

[0157] It should be noted that for measurement parameters, corresponding measurement parameters can be configured for different scenarios separately, or they can be configured uniformly regardless of the scenario. For channel statistical parameters that require feedback, corresponding channel statistical parameters that require feedback can be configured for different scenarios separately, or they can be configured uniformly regardless of the scenario.

[0158] In the embodiment of the present disclosure, in order to achieve model selection, relevant models can be pre-trained. The model selection method may also include:

[0159] The base station receives channel statistical parameters of multiple scenarios sent by at least one terminal;

[0160] The base station divides the channel statistical parameters of the multiple scenarios into N categories of channel statistical parameters according to the pre-divided number N of sub-scenarios, and obtains N data sets based on the N categories of channel statistical parameters, each data set corresponding to a sub-scenario, where N is an integer greater than 1; for example, after the N categories of channel statistical parameters are obtained by division, the average value, cluster center parameter, or center parameter of the N categories of channel statistical parameters can be reported to the simulation platform to update the channel statistical parameters in the simulation platform based on the average value, cluster center parameter, or center parameter, and then use the new simulation platform with the updated parameters to generate the corresponding category data set to obtain N data sets, and perform model training based on the N data sets;

[0161] The base station trains a model for each of the N sub-scenes based on each of the N data sets, that is, trains a model corresponding to each data set.

[0162] Optionally, the base station can configure channel statistical parameters that need to be fed back / sent for the terminal as data set classification criteria. The relevant configuration information can be sent down through the control channel. The channel statistical parameters that need to be fed back / sent may include, but are not limited to, at least one of the following: Rice's K factor (abbreviated as KF), multipath delay spread value (such as DS), multipath angle spread value (such as AS), intra-cluster multipath delay spread value, intra-cluster multipath angle spread value, Doppler frequency shift, etc.

[0163] Optionally, when the channel statistical parameters of multiple scenarios are fed back to the base station, multiple terminals can be used to perform channel measurements separately in multiple scenarios, and the multiple terminals can feed back corresponding channel statistical parameters; or a single terminal can be used to perform channel measurements in multiple scenarios, and the single terminal can feed back corresponding channel statistical parameters.

[0164] It should be noted that the above dataset can be understood as a "standardized" dataset, which is a simulated dataset based on channel model parameters obtained by statistically analyzing the distribution of measured data from a large number of scenarios. The training process of the above model can be selected based on actual conditions, such as deep learning, and this embodiment is not limited to this.

[0165] In this way, multi-scene data sets can be classified with the help of channel statistical parameters, thereby reducing data acquisition overhead and reducing the overhead of data set construction.

[0166] Optionally, if the channel statistical parameters of the multiple scenarios include K-type parameters, the k-th parameter in the K-type parameters is equal to n k sub-scenes, that is, based on the k-th parameter, the corresponding scene can be divided into n k Class (divided into n k sub-scenes), n k ≥1, 1≤k≤K, then: The number of sub-scenes that can be divided based on each type of parameter can be determined by actual measurements of multiple scenes and comprehensive consideration of factors such as acquisition overhead.

[0167] For example, taking the urban macrocell (UMa) scenario (this UMa scenario can be defined as an urban macrocell scenario, where the user density indoors and outdoors is high, and the base station is higher than the height of the surrounding buildings), the KF and extension value (such as DS / AS) are used to divide the UMa scenario into m and n categories respectively: (1) KF is used to identify non-line-of-sight (NLos) scenarios and line-of-sight (Los) scenarios. If m = 2, the UMa scenario can be divided into NLos scenarios and Los scenarios; (2) Extension value (such as DS / AS) is used to identify the complexity of the scene environment scattering. If n = 2, the UMa scenario can be divided into complex scattering scenarios and simple scattering scenarios. Therefore, the UMa scenario can be divided into N = m*n sub-UMa scenarios (i.e., sub-UMa scenarios). The specific values ​​of m and n can be determined by actual measurements of multiple UMa scenarios and taking into account factors such as acquisition overhead.

[0168] Optionally, the process of dividing the channel statistical parameters of the multiple scenarios into N categories of channel statistical parameters based on the pre-divided number N of sub-scenarios may include: classifying the channel statistical parameters of the multiple scenarios based on the pre-divided number N of sub-scenarios to obtain N categories of channel statistical parameters. This classification may be, for example, classification or clustering based on parameter features.

[0169] For example, if the channel statistical parameters of the multiple scenarios include K types of parameters (such as the KF, DS / AS parameters mentioned above) and the number of pre-divided sub-scenarios is N, then the received channel statistical parameters can be clustered using unsupervised learning k-means (K-means) clustering to form N types of channel statistical parameters. In addition, when classifying the received channel statistical parameters, different weight coefficients can be assigned to different classification parameters based on the scenario classification requirements to improve classification accuracy.

[0170] For example, as shown in Figure 4, the specific dataset construction process may include:

[0171] Step 41: The base station configures the terminal with channel statistical parameters that need to be measured / feedbacked as data set classification criteria. This configuration can be sent down via the control channel. The required / feedback channel statistical parameters may include, but are not limited to, at least one of the following: Rice's K factor, multipath delay spread value, multipath angle spread value, intra-cluster multipath delay spread value, intra-cluster multipath angle spread value, Doppler shift, etc. The multipath angle spread value may include, but is not limited to, at least one of the following: transmit horizontal angle spread value, transmit vertical angle spread value, receive horizontal angle spread value, receive vertical angle spread value, etc. The intra-cluster multipath angle spread value may include, but is not limited to, at least one of the following: transmit horizontal angle spread value, transmit vertical angle spread value, receive horizontal angle spread value, receive vertical angle spread value, etc.

[0172] Step 42: The terminal sequentially performs channel data collection, channel parameter extraction, and statistical analysis of channel statistical parameters. The collected channel data primarily includes the channel matrix; the channel parameters primarily include multipath delay, multipath angle, and multipath clustering (e.g., intra-cluster multipath delay and intra-cluster multipath angle); and the channel statistical parameters primarily include, but are not limited to, at least one of the following: Rice's K factor, multipath delay spread, multipath angle spread, intra-cluster multipath delay spread, intra-cluster multipath angle spread, and Doppler shift. The multipath angle spread may include, but is not limited to, at least one of the following: transmit horizontal angle spread, transmit vertical angle spread, receive horizontal angle spread, and receive vertical angle spread. The intra-cluster multipath angle spread may include, but is not limited to, at least one of the following: transmit horizontal angle spread, transmit vertical angle spread, receive horizontal angle spread, and receive vertical angle spread.

[0173] Step 43: The terminal feeds back the channel statistical parameters of the multiple scenarios to the base station. Here, multiple terminals can be used to perform channel measurement in multiple scenarios, or a single terminal can be used to perform channel measurement in multiple scenarios.

[0174] Step 44: The base station divides the received channel statistical parameters of the multiple scenarios into N sub-scenarios, and divides the channel statistical parameters of the multiple scenarios into N types of channel statistical parameters, and obtains N data sets based on the N types of channel statistical parameters, each data set corresponding to a sub-scenario.

[0175] Step 45: The base station trains a model for each of the N sub-scenes based on each of the N data sets, that is, trains a model corresponding to each data set (such as an AI model).

[0176] The present disclosure is described below with reference to specific examples.

[0177] In the specific example of the present disclosure, taking the UMa scene as an example, KF is applied to divide the UMa scene into m categories, namely A1, A2, ..., A m ; DS / AS is used to divide the UMa scene into n categories, namely B1, B2,…, B n ; Apply the variance of cluster number to divide the UMa scene into q categories, namely C1, C2, ..., C q ; Thus, the UMa scene can be divided into N = m*n*q sub-UMa scenes. Afterwards, the channel statistical parameters of the UMa scene reported by the terminal can be constructed into m*n*q data sets through clustering processing, and the corresponding AI models can be trained separately, namely M 111 ,…,M 11q ,M 1n1 ,…,M 1nq ,M m11 ,…,M m1q ,M mn1 ,…,M mnq , as shown in Figure 5A.

[0178] After the terminal enters a certain scene, as shown in Figure 5A, it can perform channel measurement (such as measuring the channel matrix) according to the configured measurement parameters (such as the Channel State Information Reference Signal (CSI-RS)), and perform statistical analysis based on the extracted channel parameters to obtain channel statistical parameters, and feed the channel statistical parameters back to the base station; after receiving the channel statistical parameters, the base station selects a pre-constructed data set that is closest / most similar to the channel statistical parameters, and selects the AI ​​model corresponding to the data set, and then sends the label of the AI ​​model to the terminal, and the terminal selects the corresponding AI model for deployment.

[0179] Through the simulation process, it can be proved that the solution in the present disclosure is effective. For example, in the table shown in Figure 5B, the data sets (Dataset) corresponding to scenes A, B and C are data sets pre-constructed through measured data, and the corresponding AI models have been trained offline; when the terminal enters a new scene D, the channel statistical parameters obtained through channel measurement are shown in the last column of the table. By comparison, the channel statistical parameters of scene D are closest to the pre-constructed data set B, so the AI ​​model of application scene B works under scene D. Through simulation verification, the accuracy of the AI ​​model of scene B is 0.9127, which is very close to the accuracy of 0.9133 actually trained in scene D, and the accuracy is the highest. It can be verified that the solution in the present disclosure is effective.

[0180] It should be noted that the model selection method provided in the embodiments of the present disclosure can be executed by a model selection device or a control module in the model selection device for executing the model selection method. The model selection device provided in the embodiments of the present disclosure is described by taking the execution of the model selection method by the model selection device as an example.

[0181] Please refer to FIG6 , which is a schematic diagram of the structure of a model selection device provided by an embodiment of the present disclosure. The device is applied to a terminal. As shown in FIG6 , the model selection device 60 includes:

[0182] An acquisition module 61 is configured to acquire first channel statistical parameters of a target scene;

[0183] A first sending module 62, configured to send the first channel statistical parameter to a base station;

[0184] The first receiving module 63 is configured to receive the target model fed back by the base station, wherein the target model is a model of the target scenario selected according to the first channel statistical parameter.

[0185] Through the solution in the embodiments of the present disclosure, after entering a target scenario, a terminal can obtain first channel statistical parameters for the target scenario, send these first channel statistical parameters to a base station, and receive feedback from the base station about a target model for the target scenario selected based on the first channel statistical parameters. Thus, when selecting the desired model, the terminal need not feed back a large amount of downlink channel data to the base station; instead, it only needs to feed back the channel statistical parameters for the target scenario, thereby reducing feedback overhead.

[0186] Optionally, the model selection device 60 includes:

[0187] A second receiving module is configured to receive configuration information sent by the base station, wherein the configuration information is used to configure at least one of the following:

[0188] Channel statistical parameters that need to be fed back;

[0189] Measurement parameters;

[0190] Feedback period of channel statistical parameters;

[0191] Trigger threshold value of feedback channel statistical parameters;

[0192] Constructed channel statistics parameters for multiple scenarios.

[0193] Optionally, the acquisition module 61 is specifically used to: perform channel measurement in the target scenario according to the configured measurement parameters to obtain channel data; extract channel parameters from the channel data, and perform statistical analysis on the extracted channel parameters to obtain the first channel statistical parameters.

[0194] Optionally, the first channel statistical parameter includes at least one of the following:

[0195] Rice K factor;

[0196] Multipath delay spread value;

[0197] A multipath angle extension value, wherein the multipath angle extension value includes at least one of the following: a transmission horizontal angle extension value, a transmission vertical angle extension value, a reception horizontal angle extension value, and a reception vertical angle extension value;

[0198] Intra-cluster multipath delay spread value;

[0199] Intra-cluster multipath angle spread value, the intra-cluster multipath angle spread value comprising at least one of the following: a transmission horizontal angle spread value, a transmission vertical angle spread value, a reception horizontal angle spread value, and a reception vertical angle spread value;

[0200] Doppler shift.

[0201] Optionally, the first sending module 62 is configured to perform at least one of the following:

[0202] Periodically sending the first channel statistical parameter to a base station;

[0203] When a similarity value between the first channel statistical parameter and a network configuration or a predefined channel statistical parameter corresponding to the target scenario is greater than or equal to a trigger threshold value, the first channel statistical parameter is sent to a base station.

[0204] Optionally, the model selection device 60 includes:

[0205] The second sending module is used to send channel statistical parameters of multiple scenarios to the base station; wherein the channel statistical parameters of the multiple scenarios are used to be divided into N types of channel statistical parameters, and N data sets are obtained according to the N types of channel statistical parameters, and corresponding models are obtained by training each data set in the N data sets.

[0206] The model selection device 60 of the embodiment of the present disclosure can implement each process of the method embodiment shown in Figure 1 above and can achieve the same technical effect. To avoid repetition, it will not be described here.

[0207] Please refer to FIG. 7 , which is a schematic diagram of the structure of a model selection device provided by an embodiment of the present disclosure. The device is applied to a base station. As shown in FIG. 7 , the model selection device 70 includes:

[0208] The third receiving module 71 is configured to receive a first channel statistical parameter of a target scenario sent by a terminal;

[0209] A first selection module 72 is configured to select a second channel statistical parameter that is most similar to the first channel statistical parameter from a pre-constructed data set;

[0210] A second selection module 73 is configured to select a model corresponding to the second channel statistical parameter as a target model of the target scene;

[0211] The third sending module 74 is configured to send the target model to the terminal.

[0212] Optionally, the model selection device 70 further includes:

[0213] The fourth sending module is used to send configuration information to the terminal; wherein the configuration information is used to configure at least one of the following: measurement parameters, channel statistical parameters to be fed back, feedback period of channel statistical parameters, trigger threshold value of feedback channel statistical parameters; channel statistical parameters constructed for multiple scenarios.

[0214] Optionally, the first channel statistical parameter includes at least one of the following:

[0215] Rice K factor;

[0216] Multipath delay spread value;

[0217] A multipath angle extension value, wherein the multipath angle extension value includes at least one of the following: a transmission horizontal angle extension value, a transmission vertical angle extension value, a reception horizontal angle extension value, and a reception vertical angle extension value;

[0218] Intra-cluster multipath delay spread value;

[0219] Intra-cluster multipath angle spread value, the intra-cluster multipath angle spread value comprising at least one of the following: a transmission horizontal angle spread value, a transmission vertical angle spread value, a reception horizontal angle spread value, and a reception vertical angle spread value;

[0220] Doppler shift.

[0221] Optionally, the model selection device 70 further includes:

[0222] A third receiving module, configured to receive channel statistical parameters of multiple scenarios sent by at least one terminal;

[0223] a division module, configured to divide the channel statistical parameters of the multiple scenarios into N types of channel statistical parameters according to the pre-divided number N of sub-scenarios, and obtain N data sets based on the N types of channel statistical parameters, each data set corresponding to a sub-scenario, where N is an integer greater than 1;

[0224] The training module is used to train a model of each sub-scene in the N sub-scenes based on each data set in the N data sets.

[0225] Optionally, if the channel statistical parameters include K type parameters, the kth type parameter in the K type parameters is k sub-scenes, n k ≥1, 1≤k≤K, then:

[0226] Optionally, the division module is specifically used to:

[0227] According to the number N of pre-divided sub-scenarios, the channel statistical parameters of the multiple scenarios are classified to obtain the N types of channel statistical parameters.

[0228] The model selection device 70 of the embodiment of the present disclosure can implement each process of the method embodiment shown in Figure 3 above and can achieve the same technical effect. To avoid repetition, it will not be described here.

[0229] Optionally, as shown in FIG8 , an embodiment of the present disclosure further provides a communication device 80, including a processor 81, a memory 82, and a program or instruction stored in the memory 82 and executable on the processor 81. For example, when the communication device 80 is a terminal, the program or instruction, when executed by the processor 81, implements the various processes of the method embodiment shown in FIG1 above, and can achieve the same technical effect. When the communication device 80 is a base station, the program or instruction, when executed by the processor 81, implements the various processes of the method embodiment shown in FIG3 above, and can achieve the same technical effect. To avoid repetition, they are not described here.

[0230] The embodiments of the present disclosure also provide a computer program product, including computer instructions. When the computer instructions are executed by a processor, the various processes of the method embodiments shown in Figures 1 or 3 above can be implemented and the same technical effects can be achieved. To avoid repetition, they will not be described here.

[0231] The embodiment of the present disclosure also provides a readable storage medium on which a program or instruction is stored. When the program or instruction is executed by a processor, the various processes of the above-mentioned model selection method embodiment can be implemented and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0232] Computer-readable media include both permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology for information storage. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. According to the definition in this article, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0233] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0234] The serial numbers of the above-mentioned embodiments of the present disclosure are for description only and do not represent the advantages or disadvantages of the embodiments.

[0235] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present disclosure is essentially or the part that contributes to the existing technology can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a number of instructions for enabling a service classification device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present disclosure.

[0236] The above is only a preferred embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present disclosure. These improvements and modifications should also be regarded as within the scope of protection of the present disclosure.

Claims

1. A model selection method comprising: The terminal obtains a first channel statistical parameter of the target scene; The terminal sends the first channel statistical parameter to the base station; The terminal receives the target model fed back by the base station, wherein the target model is a model of the target scenario selected according to the first channel statistical parameter.

2. The method according to claim 1, further comprising: The terminal receives the configuration information sent by the base station; The configuration information is used to configure at least one of the following: Channel statistical parameters that need to be fed back; Measurement parameters; Feedback period of channel statistical parameters; Trigger threshold value of feedback channel statistical parameters; Constructed channel statistics parameters for multiple scenarios.

3. The method according to claim 1 or 2, wherein: The terminal obtains a first channel statistical parameter of a target scene, including: The terminal performs channel measurement in the target scenario according to the configured measurement parameters to obtain channel data; The terminal extracts channel parameters from the channel data, and performs statistical analysis on the extracted channel parameters to obtain the first channel statistical parameters.

4. The method according to claim 1, wherein The first channel statistical parameter includes at least one of the following: Rice K factor; Multipath delay spread value; A multipath angle extension value, wherein the multipath angle extension value includes at least one of the following: a transmission horizontal angle extension value, a transmission vertical angle extension value, a reception horizontal angle extension value, and a reception vertical angle extension value; Intra-cluster multipath delay spread value; Intra-cluster multipath angle spread value, the intra-cluster multipath angle spread value comprising at least one of the following: a transmission horizontal angle spread value, a transmission vertical angle spread value, a reception horizontal angle spread value, and a reception vertical angle spread value; Doppler shift.

5. The method according to claim 1, wherein The terminal sending the first channel statistical parameter to the base station includes at least one of the following: The terminal periodically sends the first channel statistical parameter to the base station; When a similarity value between the first channel statistical parameter and a network configuration or a predefined channel statistical parameter corresponding to the target scenario is greater than or equal to a trigger threshold, the terminal sends the first channel statistical parameter to a base station.

6. The method according to claim 1, further comprising: The terminal sends channel statistical parameters of multiple scenarios to the base station; The channel statistical parameters of the multiple scenarios are used to be divided into N types of channel statistical parameters, and N data sets are obtained according to the N types of channel statistical parameters, and corresponding models are trained according to each of the N data sets.

7. A model selection method comprising: The base station receives a first channel statistical parameter of a target scene sent by the terminal; The base station selects, from a pre-constructed data set, a second channel statistical parameter that is most similar to the first channel statistical parameter; The base station selects a model corresponding to the second channel statistical parameter as a target model of the target scene; The base station sends the target model to the terminal.

8. The method according to claim 7, further comprising: The base station sends configuration information to the terminal; The configuration information is used to configure at least one of the following: Channel statistical parameters that need to be fed back; Measurement parameters; Feedback period of channel statistical parameters; Trigger threshold value of feedback channel statistical parameters; Constructed channel statistics parameters for multiple scenarios.

9. The method according to claim 7 or 8, wherein The first channel statistical parameter includes at least one of the following: Rice K factor; Multipath delay spread value; A multipath angle extension value, wherein the multipath angle extension value includes at least one of the following: a transmission horizontal angle extension value, a transmission vertical angle extension value, a reception horizontal angle extension value, and a reception vertical angle extension value; Intra-cluster multipath delay spread value; Intra-cluster multipath angle spread value, the intra-cluster multipath angle spread value comprising at least one of the following: a transmission horizontal angle spread value, a transmission vertical angle spread value, a reception horizontal angle spread value, and a reception vertical angle spread value; Doppler shift.

10. The method according to claim 7, further comprising: The base station receives channel statistical parameters of multiple scenarios sent by at least one terminal; The base station divides the channel statistical parameters of the multiple scenarios into N types of channel statistical parameters according to the pre-divided number N of sub-scenarios, and obtains N data sets according to the N types of channel statistical parameters; Each data set corresponds to a sub-scene, and N is an integer greater than 1; The base station trains a model of each of the N sub-scenes based on each of the N data sets.

11. The method according to claim 10, wherein: If the channel statistical parameters include K type parameters, the kth type parameter in the K type parameters is equal to n k sub-scenes, n k ≥1, 1≤k≤K, then:

12. The method according to claim 10, wherein: The dividing the channel statistical parameters of the multiple scenarios into N types of channel statistical parameters according to the number N of pre-divided sub-scenarios includes: The base station classifies the channel statistical parameters of the multiple scenarios according to the number N of pre-divided sub-scenarios to obtain the N types of channel statistical parameters.

13. A model selection device comprising: An acquisition module, configured to acquire statistical parameters of a first channel of a target scene; A first sending module, configured to send the first channel statistical parameter to a base station; A first receiving module is configured to receive a target model fed back by the base station, wherein the target model is a model of the target scenario selected according to the first channel statistical parameter.

14. A model selection device comprising: A third receiving module, configured to receive a first channel statistical parameter of a target scenario sent by a terminal; a first selection module, configured to select a second channel statistical parameter that is most similar to the first channel statistical parameter from a pre-constructed data set; A second selection module, configured to select a model corresponding to the second channel statistical parameter as a target model of the target scene; The third sending module is configured to send the target model to the terminal.

15. A communication device comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method according to any one of claims 1 to 6, or the steps of the method according to any one of claims 7 to 12.

16. A readable storage medium storing a program or instruction, wherein the program or instruction, when executed by a processor, implements the steps of the method according to any one of claims 1 to 6, or the steps of the method according to any one of claims 7 to 12.

17. A computer program product comprising computer instructions, wherein when the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 or the steps of the method according to any one of claims 7 to 12 are implemented.

Images