WO2025065321A1 - 波束预测模型切换方法、装置、终端、基站及存储介质 - Google Patents

波束预测模型切换方法、装置、终端、基站及存储介质 Download PDF

Info

Publication number
WO2025065321A1
WO2025065321A1 PCT/CN2023/121924 CN2023121924W WO2025065321A1 WO 2025065321 A1 WO2025065321 A1 WO 2025065321A1 CN 2023121924 W CN2023121924 W CN 2023121924W WO 2025065321 A1 WO2025065321 A1 WO 2025065321A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal quality
prediction model
prediction
error value
prediction error
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/CN2023/121924
Other languages
English (en)
French (fr)
Inventor
张俊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
New H3C Technologies Co Ltd
Original Assignee
New H3C Technologies Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by New H3C Technologies Co Ltd filed Critical New H3C Technologies Co Ltd
Priority to PCT/CN2023/121924 priority Critical patent/WO2025065321A1/zh
Priority to CN202380010925.1A priority patent/CN120113284A/zh
Publication of WO2025065321A1 publication Critical patent/WO2025065321A1/zh
Anticipated expiration legal-status Critical
Pending legal-status Critical Current

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W16/00Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
    • H04W16/24Cell structures
    • H04W16/28Cell structures using beam steering

Definitions

  • the present application relates to the field of communication technology, and in particular to a beam prediction model switching method, device, terminal, base station and storage medium.
  • the traditional beam prediction method is for the base station to send pilot signals in all transmission beam directions to predict the beam direction of the optimal transmission beam.
  • this method will generate huge pilot overhead.
  • the beam prediction model can be trained by machine learning, and then the beam prediction model can be used to predict the beam direction of the optimal transmission beam through the measurement results of fewer beams.
  • the purpose of the embodiments of the present application is to provide a beam prediction model switching method, device, terminal, base station and storage medium, so as to timely switch a beam prediction model that is more suitable for the current scenario.
  • the specific technical solution is as follows:
  • an embodiment of the present application provides a beam prediction model switching method, which is applied to a terminal, wherein the terminal includes a first beam prediction model and multiple second beam prediction models, and the effective beam prediction model of the terminal is the first beam prediction model.
  • the method includes:
  • a target beam prediction model is selected from the multiple second beam prediction models, and the effective beam prediction model is switched to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • the selecting a target beam prediction model from the multiple second beam prediction models includes:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • the method further includes:
  • the first beam prediction model is disabled and beam prediction is performed using the full beam set.
  • the calculating a first prediction error value of the first beam prediction model based on the first predicted signal quality and the measured signal quality includes:
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by the first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • the first prediction error value
  • is the filtering factor.
  • the method before measuring the pilot signal of each beam included in the first monitoring beam set, the method further includes:
  • the method further includes:
  • the UCI includes beam indexes and signal qualities of multiple beams, and the signal qualities of the multiple beams are all minimum signal qualities;
  • the method further includes:
  • the base station Sending a second RRC message to the base station, where the second RRC message includes an identifier of a third measurement beam set corresponding to the target beam prediction model and an identifier of a third monitoring beam set;
  • an embodiment of the present application provides a beam prediction model switching method, which is applied to a base station, wherein the base station includes a first beam prediction model and multiple second beam prediction models, and the effective beam prediction model of the base station is the first beam prediction model.
  • the method includes:
  • the signal quality of each beam included in the first measurement beam set is obtained from the signal quality corresponding to each beam included in the first monitoring beam set. quality, and inputting the signal quality of each beam included in the first measurement beam set into the first beam prediction model to obtain a first predicted signal quality of the predicted beam output by the first beam prediction model;
  • a target beam prediction model is selected from the multiple second beam prediction models, and the effective beam prediction model is switched to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • the selecting a target beam prediction model from the multiple second beam prediction models includes:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • the method further includes:
  • the first beam prediction model is disabled and the full beam set is used to perform beam prediction.
  • the calculating a first prediction error value of the first beam prediction model based on the first predicted signal quality and the measured signal quality includes:
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by the first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • the first prediction error value
  • is the filtering factor.
  • an embodiment of the present application provides a beam prediction model switching device, which is applied to a terminal, wherein the terminal includes a first beam prediction model and multiple second beam prediction models, and the effective beam prediction model of the terminal is the first beam prediction model, and the device includes:
  • a measurement module configured to measure the pilot signal of each beam included in the first monitoring beam set, obtain the signal quality corresponding to each beam included in the first monitoring beam set, and select the signal quality with the largest value from the signal qualities corresponding to each beam included in the first monitoring beam set as the measured signal quality;
  • a prediction module configured to obtain the signal quality of each beam included in the first measurement beam set from the signal quality corresponding to each beam included in the first monitoring beam set, and input the signal quality of each beam included in the first measurement beam set into the first beam prediction model to obtain a first predicted signal quality of the predicted beam output by the first beam prediction model;
  • a calculation module configured to calculate a first prediction error value of the first beam prediction model based on the first predicted signal quality and the measured signal quality
  • a switching module is used to select a target beam prediction model from the multiple second beam prediction models if the first prediction error value is greater than or equal to a specified threshold value, and switch the effective beam prediction model to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • the switching module is specifically configured to:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • the prediction module is also used to deactivate the first beam prediction model and use the full beam set for beam prediction if the first prediction error value and the second prediction error values of the multiple second beam prediction models are both greater than or equal to the specified threshold value.
  • the calculation module is specifically configured to:
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by the first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • the first prediction error value
  • is the filtering factor.
  • the device further includes:
  • a receiving module is used to receive configuration information of a full beam set, configuration information of at least one measurement beam set, and configuration information of at least one monitoring beam set sent by a base station; and to receive a first RRC message sent by the base station, wherein the first RRC message includes an identifier of the first measurement beam set, an identifier of the first monitoring beam set, the specified threshold value, and the filter factor, and the first RRC message is used to instruct the terminal to perform beam measurement using the first monitoring beam set and perform beam prediction using the first measurement beam set.
  • the device further includes a sending module
  • the sending module is configured to send UCI to the base station, where the UCI includes beam indexes and signal qualities of multiple beams, and the signal qualities of the multiple beams are all minimum signal qualities;
  • the receiving module is also used to receive a first indication message sent by the base station, where the first indication message includes an identifier of the full beam set, and the first indication message is used to instruct the terminal to use the configuration information of the full beam set to perform beam measurement and beam prediction.
  • the sending module is further used to send a second RRC message to the base station, where the second RRC message includes an identifier of a third measurement beam set corresponding to the target beam prediction model and an identifier of a third monitoring beam set;
  • the receiving module is further used to receive a second indication message sent by the base station, where the second indication message is used to instruct the terminal to perform beam measurement using the third monitoring beam set and to perform beam prediction using the third measurement beam set.
  • an embodiment of the present application provides a beam prediction model switching device, which is applied to a base station, wherein the base station includes a first beam prediction model and multiple second beam prediction models, the effective beam prediction model of the base station is the first beam prediction model, and the device includes include:
  • an acquisition module configured to acquire signal qualities corresponding to the beams included in the first monitoring beam set reported by the terminal, and select a signal quality with the largest value from the signal qualities corresponding to the beams included in the first monitoring beam set as the measured signal quality;
  • a prediction module configured to obtain the signal quality of each beam included in the first measurement beam set from the signal quality corresponding to each beam included in the first monitoring beam set, and input the signal quality of each beam included in the first measurement beam set into the first beam prediction model to obtain a first predicted signal quality of the predicted beam output by the first beam prediction model;
  • a calculation module configured to calculate a first prediction error value of the first beam prediction model based on the first predicted signal quality and the measured signal quality
  • a switching module is used to select a target beam prediction model from the multiple second beam prediction models if the first prediction error value is greater than or equal to a specified threshold value, and switch the effective beam prediction model to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • the switching module is specifically configured to:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • the prediction module is further used to deactivate the first beam prediction model and use the full beam set to perform beam prediction if the first prediction error value and the prediction error values of the multiple second beam prediction models are both greater than or equal to the specified threshold value.
  • the calculation module is specifically configured to:
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by the first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • the first prediction error value
  • is the filtering factor.
  • an embodiment of the present application provides a terminal, the terminal including a first beam prediction model and multiple second beam prediction models, the effective beam prediction model of the terminal is the first beam prediction model, and the terminal includes:
  • a machine-readable storage medium storing machine-executable instructions that can be executed by the processor; the machine-executable instructions prompt the processor to perform the following steps:
  • a target beam prediction model is selected from the multiple second beam prediction models, and the effective beam prediction model is switched to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • the machine executable instruction further causes the processor to perform the following steps:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • the machine executable instruction further causes the processor to perform the following steps:
  • the first beam prediction model is disabled and beam prediction is performed using the full beam set.
  • the machine executable instruction further causes the processor to perform the following steps:
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by the first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • the first prediction error value
  • is the filtering factor.
  • the machine executable instruction further causes the processor to perform the following steps:
  • the machine executable instruction further causes the processor to perform the following steps:
  • the UCI includes beam indexes and signal qualities of multiple beams, and the signal qualities of the multiple beams are all minimum signal qualities;
  • the machine executable instruction further causes the processor to perform the following steps:
  • the base station Sending a second RRC message to the base station, where the second RRC message includes an identifier of a third measurement beam set corresponding to the target beam prediction model and an identifier of a third monitoring beam set;
  • an embodiment of the present application provides a base station, the base station comprising a first beam prediction model and multiple second beam prediction models, the effective beam prediction model of the base station is the first beam prediction model, and the base station comprises:
  • a machine-readable storage medium storing machine-executable instructions that can be executed by the processor; the machine-executable instructions prompt the processor to perform the following steps:
  • a target beam prediction model is selected from the multiple second beam prediction models, and the effective beam prediction model is switched to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • the machine executable instruction further causes the processor to perform the following steps:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • the machine executable instruction further causes the processor to perform the following steps:
  • the first beam prediction model is disabled and the full beam set is used to perform beam prediction.
  • the machine executable instruction further causes the processor to perform the following steps:
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by the first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • the first prediction error value
  • is the filtering factor.
  • an embodiment of the present application provides a machine-readable storage medium storing machine-executable instructions, which are called and executed by a processor. When executed, the machine executable instructions cause the processor to implement the method described in the first aspect or the second aspect.
  • an embodiment of the present application provides a computer program product, which prompts the processor to implement the method described in the first aspect or the second aspect.
  • the terminal measures the signal quality corresponding to each beam in the first monitoring beam set, and takes the signal quality with the largest value among the signal qualities corresponding to each beam included in the first monitoring beam set as the measured signal quality. Then, the signal quality of each beam included in the first measurement beam set is obtained from the signal quality corresponding to the first monitoring beam set, and the signal quantity of each beam included in the first measurement beam set is input into the first beam prediction model to obtain the first predicted signal quality of the predicted beam output by the first beam prediction model. Since the measured signal quality is the optimal signal quality actually measured by the terminal, the first predicted signal quality is the optimal signal quality predicted by the first beam prediction model. Therefore, the first prediction error value of the first beam prediction model can be obtained by the measured signal quality and the first predicted signal quality.
  • the terminal can select the target beam prediction model from multiple second beam prediction models in a timely manner, and switch the effective beam prediction model to the target beam prediction model. Since the prediction error value is less than the specified threshold value, the second prediction model is more suitable for the current communication scenario and can improve the communication quality between the base station and the terminal.
  • FIG1 is a flow chart of a beam prediction model switching method provided in an embodiment of the present application.
  • FIG2 is a flow chart of another beam prediction model switching method provided in an embodiment of the present application.
  • FIG3 is a flow chart of another beam prediction model switching method provided in an embodiment of the present application.
  • FIG4 is a flow chart of another beam prediction model switching method provided in an embodiment of the present application.
  • FIG5 is a flowchart of another beam prediction model switching method provided in an embodiment of the present application.
  • FIG6 is a schematic diagram of the structure of a beam prediction model device provided in an embodiment of the present application.
  • FIG7 is a schematic diagram of the structure of another beam prediction model device provided in an embodiment of the present application.
  • FIG8 is a schematic diagram of the structure of a terminal provided in an embodiment of the present application.
  • FIG. 9 is a schematic diagram of the structure of a terminal provided in an embodiment of the present application.
  • the beam prediction model involved in the embodiments of the present application is a pre-trained artificial intelligence (AI) model, and the beam prediction model can be deployed in a terminal or a base station.
  • the beam prediction model is used to predict the optimal beam in the full amount of beams using the signal quality of a small number of beams.
  • the input of the beam prediction model can be the beam index and corresponding signal quality of a small number of beams, and the output can be the predicted beam index and signal quality of the optimal beam, or the output can be the predicted beam index and signal quality of the top N beams in terms of signal quality.
  • the terminal uses the beam measurement result to monitor the performance of the currently effective beam prediction model to determine whether the currently effective beam prediction model needs to be switched.
  • the terminal When multiple beam prediction models are deployed in the base station, the terminal reports the beam measurement results to the base station, and the base station generates a beam prediction model based on the beam measurement results. The performance of the currently effective beam prediction model is monitored to determine whether it is necessary to switch the currently effective beam prediction model.
  • An embodiment of the present application provides a beam prediction model switching method, which is applied to a terminal, the terminal including a first beam prediction model and multiple second beam prediction models, and the effective beam prediction model of the terminal is the first beam prediction model.
  • the method includes:
  • the measured signal quality is the signal quality with the largest value among the signal qualities corresponding to the beams included in the first monitoring beam set, that is, the optimal signal quality obtained by actual measurement.
  • the base station sends a pilot signal in the beam direction of each beam included in the first monitoring beam set according to a pre-configured measurement period.
  • the terminal measures the pilot signal in each beam direction to obtain the signal quality corresponding to each beam, and determines the signal quality with the largest value from the measured signal qualities, and uses the signal quality with the largest value as the measured signal quality.
  • the signal quality can be represented by Layer 1 Reference Signal Receiving Power (L1-RSRP).
  • L1-RSRP can be used to represent the strength of the physical layer pilot signal. The larger the L1-RSRP, the better the signal quality represented.
  • the input parameter of the first beam prediction model is the signal quality of each beam included in the first measurement beam set.
  • the first measurement beam set is a subset of the first monitoring beam set, that is, the first monitoring beam set includes more beams than the first measurement beam set, so as to achieve the purpose of performance monitoring of the first beam prediction model.
  • the terminal can obtain the signal quality corresponding to each beam in the first measurement beam set.
  • the first beam prediction model can use the signal quality corresponding to each beam in the first measurement beam set to predict the optimal beam between the current base station and the terminal, that is, the predicted beam.
  • the first beam prediction model can also predict the signal quality of the predicted beam, that is, the first predicted signal quality.
  • the first predicted signal quality is the signal quality of the optimal predicted beam predicted by the first beam prediction model
  • the measured signal quality is the signal quality of the optimal beam among the beams in the first monitoring beam set obtained by actual measurement. Therefore, the first prediction error value can represent the error between the signal quality of the optimal beam predicted by the first beam prediction model and the measured signal quality of the optimal beam obtained by actual measurement.
  • a target beam prediction model is selected from multiple second beam prediction models, and the effective beam prediction model is switched to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • the terminal can select a target beam prediction model with a prediction error value less than the specified threshold value from multiple second beam prediction models to avoid deterioration of the communication quality between the base station and the terminal.
  • the terminal measures the signal quality corresponding to each beam in the first monitoring beam set, and takes the signal quality with the largest value among the signal qualities corresponding to each beam included in the first monitoring beam set as the measured signal quality. Then, the signal quality of each beam included in the first measurement beam set is obtained from the signal quality corresponding to the first monitoring beam set, and the signal quality of each beam included in the first measurement beam set is input to the measurement beam set. Input the first beam prediction model to obtain the first predicted signal quality of the predicted beam output by the first beam prediction model. Since the measured signal quality is the optimal signal quality actually measured by the terminal, the first predicted signal quality is the optimal signal quality predicted by the first beam prediction model.
  • the first prediction error value of the first beam prediction model can be obtained through the measured signal quality and the first predicted signal quality. If the first prediction error value is greater than or equal to the specified threshold value, it means that the error of the first beam prediction model is large, that is, the first beam prediction model is not suitable for the current communication scenario between the base station and the terminal. Then, the terminal can select the target beam prediction model from multiple second beam prediction models in time, and switch the effective beam prediction model to the target beam prediction model. Since the prediction error value is less than the specified threshold value, the second prediction model is more suitable for the current communication scenario and can improve the communication quality between the base station and the terminal.
  • the base station Before executing the method shown in FIG1 , the base station needs to configure the terminal, that is, before S101 , the following steps need to be performed:
  • the terminal receives configuration information of a full beam set, configuration information of at least one measurement beam set, and configuration information of at least one monitoring beam set sent by a base station.
  • a first RRC message sent by the base station is received, the first RRC message including an identifier of a first measurement beam set, an identifier of a first monitoring beam set, a specified threshold value, and a filter factor, and the first RRC message is used to instruct the terminal to perform beam measurement using the first monitoring beam set and perform beam prediction using the first measurement beam set.
  • At least one measurement beam set is all measurement beam sets supported by the base station, and at least one monitoring beam set is all monitoring beam sets supported by the base station.
  • At least one measurement beam set is a measurement beam set corresponding to each beam prediction model in the terminal
  • at least one monitoring beam set is a monitoring beam set corresponding to each beam prediction model in the terminal.
  • the input of the beam prediction model is the signal quality of each beam included in the measurement beam set.
  • the inputs of different beam prediction models in the terminal can be the same or different. When the inputs are different, the corresponding measurement beam sets are also different.
  • the monitoring beam sets corresponding to different measurement beam sets may be the same or different.
  • the monitoring beam set may be pre-set, for example, the measurement beam set and the monitoring beam set may be in a one-to-one relationship, and some beams may be added to the measurement beam set to obtain the monitoring beam set.
  • the measurement beam set and the monitoring beam set may be in a many-to-one relationship, that is, multiple measurement beam sets are subsets of the same monitoring beam set.
  • Each beam supported by the base station corresponds to a beam index.
  • the configuration information of each beam set includes: beam set identifier, beam index, measurement period and number of beams.
  • the terminal can store the configuration information of the full beam set, the configuration information of at least one measurement beam set, and the configuration information of at least one monitoring beam set. Then, when the first RRC message is received, the configuration information of the first measurement beam set is searched using the identifier of the first measurement beam set, and the configuration information of the first monitoring beam set is searched using the identifier of the first monitoring beam set. Then, beam measurement is performed on each beam included in the first monitoring beam set according to the configuration information of the first monitoring beam set, and beam prediction is performed using the signal quality of each beam included in the first measurement beam set. It can be understood that the base station can activate the first measurement beam set and the first monitoring beam set through the first RRC message, so that the terminal starts to use the first measurement beam set and the first monitoring beam set.
  • the input parameters corresponding to each beam prediction model in the terminal have a certain flexibility. If the first measurement beam set does not completely match the current input parameters of the first beam prediction model, the input parameters of the first beam prediction model can be adjusted so that the input parameters of the first beam prediction model are adapted to the first measurement beam set.
  • the base station can pre-send the configuration information of the full beam set, the configuration information of at least one measurement beam set, and the configuration information of at least one monitoring beam set to the terminal. Subsequently, the measurement beam set and the monitoring beam set required by the terminal can be configured for the terminal through the first RRC message. There is no need to re-send the configuration information of the measurement beam set and the configuration information of the monitoring beam set each time the terminal switches the beam prediction model, which can improve the model switching efficiency.
  • Method 1 taking the difference between the measured signal quality and the first predicted signal quality as the first prediction error value.
  • the terminal may determine the actual signal quality and the first predicted signal quality in each measurement period of the first monitoring beam set, and further calculate the first prediction error value in each measurement period.
  • Method 2 Calculate the difference between the measured signal quality and the first predicted signal quality to obtain the current prediction error value
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by a first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • the first prediction error value
  • the current prediction error value can be expressed as L1_RSRP_Gap_Current
  • the historical prediction error value can be expressed as L1_RSRP_Gap_Prev
  • the first prediction error value Gap
  • the current prediction error value calculated in the first measurement cycle can be used as the first prediction error value of the first measurement cycle.
  • the current prediction error value of the first measurement cycle can be used as the historical prediction error value
  • the first prediction error value of the second measurement cycle can be calculated using the above formula.
  • the current prediction error value of the second measurement cycle can be used as the historical prediction error value.
  • the above-mentioned first prediction error value can reflect the difference between the actual measured signal quality and the first predicted signal quality, and the difference can reflect the performance of the first beam prediction model in the current communication scenario between the terminal and the base station.
  • the first prediction error value is obtained by filtering the current prediction error value and the historical prediction error value, which can accurately reflect the performance of the first beam prediction model in two consecutive measurement cycles, so that the model switching can be performed more accurately based on the first prediction error value and the specified threshold value in the future.
  • the looser the switching conditions the more frequent the model switching, and the correspondingly greater the air interface overhead caused by the model switching.
  • the stricter the switching conditions the less frequent the model switching, which may lead to untimely model switching.
  • the embodiment of the present application can make the switching conditions more reasonable, and can balance the real-time nature of the switching and the air interface overhead, so as to maximize the overall performance.
  • the terminal may traverse multiple second beam prediction models and select a target beam prediction model from them.
  • the selection process includes the following steps:
  • Step 1 for the second beam prediction model, input the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain the second predicted signal quality of the predicted beam output by the second beam prediction model.
  • the measurement beam set corresponding to the second beam prediction model corresponds to the input parameter of the second beam measurement model.
  • the terminal can obtain the configuration information of the measurement beam set from the configuration information of at least one measurement beam set sent by the base station, thereby obtaining the beam index included in the measurement beam set from the configuration information of the measurement beam set.
  • the terminal obtains the signal quality corresponding to each beam included in the measurement beam set from the signal quality corresponding to each beam included in the first monitoring beam set according to the beam index included in the measurement beam set.
  • the signal quality corresponding to each beam included in the measurement beam set can be directly obtained; if the first monitoring beam set does not include all the beams of the measurement beam set, the terminal can obtain the signal quality corresponding to some beams not included in the measurement beam set from the first monitoring beam set, thereby obtaining the same signal quality as the number of beams included in the measurement beam set, and then adaptively adjust the input parameters of the second beam prediction model, and input the obtained signal quality corresponding to each beam into the second beam prediction model.
  • the first RRC message may include identifiers of multiple monitoring beam sets, and then the terminal will measure the beams included in the multiple monitoring beam sets to obtain the signal quality of each beam included in each monitoring beam set.
  • the terminal can obtain the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model from the signal quality of each beam included in the monitoring beam set corresponding to the second beam prediction model. Then, the signal quality of each beam included in the measurement beam set can be input into the second beam prediction model.
  • Step 2 Using the difference between the measured signal quality and the second predicted signal quality as the second prediction error value of the second beam prediction model.
  • the measured signal quality may be the signal quality with the largest value among the signal qualities of the beams included in the first monitoring beam set.
  • the measured signal quality may be the signal quality with the largest value among the signal qualities of the beams included in the monitoring beam set corresponding to the second beam prediction model.
  • Step 3 If the second prediction error value is less than the specified threshold value, the second beam prediction model is used as the target beam prediction model and the traversal operation is terminated.
  • the terminal switches the effective beam prediction model to the target beam prediction model.
  • Step 4 If the second prediction error value is greater than or equal to the specified threshold value, continue the traversal operation.
  • the traversal operation ends when a second prediction model whose second prediction error value is less than a specified threshold value is found.
  • the traversal operation is stopped.
  • the terminal can determine that the first prediction error value and the prediction error values of multiple second beam prediction models are greater than or equal to the specified threshold value, that is, there is no beam prediction model suitable for the current scene, then the first beam prediction model is disabled, and the full beam set is used for beam prediction.
  • the terminal can fall back to non-AI mode, that is, use the full beam set for beam prediction. This can achieve rapid fallback to non-AI mode when the performance of the beam prediction model deteriorates or when the generalization performance of the beam prediction model is poor in the early stage of beam prediction model deployment, so as to avoid affecting the communication quality between the base station and the terminal.
  • the terminal uses the full beam set to perform beam prediction, which means that the base station sends pilot signals in all beam directions respectively, and the terminal measures the pilot signals in each beam direction, so as to predict the beam corresponding to the optimal signal quality.
  • the terminal and the base station can subsequently communicate using the beam corresponding to the optimal signal quality.
  • the terminal may send uplink control information (UCI) to the base station to notify the base station that each beam prediction model is no longer applicable, so as to initiate a model fallback request to the base station.
  • UCI uplink control information
  • Step A Send UCI to the base station.
  • the UCI measurement reporting format can be reused, and the UCI includes the beam index and signal quality of multiple beams, and the signal quality of multiple beams is the minimum signal quality.
  • the minimum signal quality is -140dBm (milliwatt decibel).
  • the number of beam indexes included in the UCI is pre-configured by the base station for the terminal.
  • Table 1 exemplarily shows a UCI format.
  • Step B Receive a first indication message sent by the base station, where the first indication message includes an identifier of the full beam set.
  • the first indication message is used to instruct the terminal to use the configuration information of the full beam set to perform beam measurement and beam prediction. That is, the first indication message activates the full beam set.
  • the terminal After receiving the first indication message, the terminal can deactivate the first beam prediction model and fall back to the non-AI mode.
  • the first indication message can be a transmission configuration indication information (Transmission Configuration Indicator, TCI), a media access control control unit (Media Access Control-Control Element, MAC-CE) message or an RRC message.
  • TCI Transmission Configuration Indicator
  • MAC-CE Media Access Control-Control Element
  • RRC Radio Resource Control
  • the terminal sets the signal quality in the UCI to the minimum value, thereby notifying the base station that there is no applicable beam prediction model at present. Then, the base station can instruct the terminal to use the full beam set for beam prediction through the first indication message, and can quickly fall back to the non-AI mode.
  • the existing message types can be used between the terminal and the base station to implement model fallback, which has little impact on the current standard protocol.
  • the terminal after the terminal selects the target beam prediction model from multiple second beam prediction models, the terminal also needs to request the base station for a subsequent measurement beam set and a monitoring beam set applicable to the second beam prediction model.
  • the terminal sends a second RRC message to the base station, the second RRC message including an identifier of a third beam set corresponding to the target beam prediction model and an identifier of a third monitoring beam set. Then, the terminal receives a second indication message sent by the base station, the second indication message is used to instruct the terminal to perform beam measurement using the third monitoring beam set and to perform beam prediction using the third measurement beam set.
  • the second indication message includes an identifier of the third measurement beam set and an identifier of the third monitoring beam set. It can be understood that the base station can activate the third measurement beam set and the third monitoring beam set through the second indication message, and deactivate the first measurement beam set and the first monitoring beam set.
  • the second indication message may be a TCI message, a MAC-CE message or an RRC message.
  • the base station Since the base station has sent the configuration information of each measurement beam set and each monitoring beam set to the terminal in advance, the base station can start the terminal's measurement of the third measurement beam set and the third monitoring beam set through the second indication message, thereby realizing rapid switching of the beam prediction model.
  • the method includes:
  • L1_RSRPs corresponding to the monitoring beam set Set_monitor are traversed in sequence to obtain the L1_RSRP of the beam with the best performance, which is recorded as L1_RSRP_Best.
  • the best performance beam is the beam corresponding to the measured signal quality.
  • the terminal searches for the beam with the largest L1_RSRP from the beams included in Set_monitor, and records the L1_RSRP as L1_RSRP_Best.
  • the black square on the left side of S201 in Figure 2 is the best performance beam.
  • S202 Extract the L1_RSRP of each beam included in the measurement beam set Set_measure from the L1_RSRP of each beam included in Set_monitor. L1_RSRP, as the input of the current beam prediction model.
  • the black squares on the left side of S202 in FIG. 2 are beams included in Set_measure.
  • the shaded squares on the left side of S203 in Figure 2 are the optimal beams predicted by the beam prediction model, and the black squares are the beams with the best performance in S201.
  • S204 Calculate the difference between L1_RSRP_Best and L1_RSRP_Predict and perform filtering to obtain a prediction error value.
  • the filtering method may refer to the formula for calculating the first prediction error value using the filtering factor in the above embodiment.
  • the model switching request may be the second RRC message in the above embodiment.
  • model selection failure means that the prediction error values corresponding to the beam prediction models deployed in the terminal are all greater than or equal to the specified threshold value.
  • S211 select the next beam prediction model, and return to S201 for the next beam prediction model, so that each beam prediction model can be traversed.
  • the filter factor and specifying the threshold value By adopting the above method, by setting the filter factor and specifying the threshold value, the air interface measurement overhead and beam prediction performance can be dynamically adjusted, and the appropriate beam prediction model can be switched in time to meet the needs of different scenarios. In addition, when the performance of each beam prediction model deteriorates, it can quickly fall back to the non-AI mode to avoid terminal disconnection.
  • FIG3 shows an interaction process between the terminal and the base station, and the method includes:
  • the base station sends configuration information of the full beam set, configuration information of the measurement beam set, and configuration information of the monitoring beam set to the terminal.
  • the base station activates the measurement beam set and the monitoring beam set, and sends down the filter factor and the designated threshold value.
  • This step can be implemented through the first RRC message introduced in the above embodiment, which will not be repeated here.
  • the terminal measures the pilot signal of each beam included in the monitoring beam set, obtains the L1_RSRP corresponding to each beam, and defines the beam with the largest L1_RSRP as the measured optimal beam.
  • the measured optimal beam is the beam corresponding to the above-mentioned measured signal quality.
  • S304 The terminal performs beam prediction using the current beam prediction model, and analyzes the performance difference between the predicted optimal beam and the measured optimal beam.
  • the performance difference is the prediction error value introduced in the above embodiment.
  • the method of calculating the prediction error value can refer to the relevant description in the above embodiment, which will not be repeated here.
  • S305 The terminal decides to switch the model or return to the non-AI mode.
  • the method for the terminal to determine whether to switch the model and whether to fall back to the AI mode based on the specified threshold value can refer to the above embodiment. The relevant description will not be repeated here.
  • the terminal sends a model switching request or a request to fall back to a non-AI mode to the base station.
  • the base station activates/deactivates the monitoring beam set and the measurement beam set for the terminal.
  • This step may be implemented through the first indication message or the second indication message in the above embodiment.
  • the embodiment of the present application also provides a beam prediction model switching method, which is applied to a base station, the base station includes a first beam prediction model and multiple second beam prediction models, and the effective beam prediction model of the base station is the first beam prediction model.
  • the method includes:
  • the base station pre-configures the measurement period and reporting method of the first monitoring beam set for the terminal, and then the base station sends a pilot signal in the beam direction of each beam included in the first monitoring beam set according to the pre-configured measurement period.
  • the terminal measures the pilot signal in each beam direction, obtains the signal quality corresponding to each beam, and reports the signal quality corresponding to each beam to the base station.
  • the first measurement beam set is a subset of the first monitoring beam set.
  • the base station obtains the signal quality corresponding to each beam in the first monitoring beam set reported by the terminal, and takes the signal quality with the largest value among the signal qualities corresponding to each beam included in the first monitoring beam set as the measured signal quality. Then, the signal quality of each beam included in the first measurement beam set is obtained from the signal quality corresponding to the first monitoring beam set, and the signal quantity of each beam included in the first measurement beam set is input into the first beam prediction model to obtain the first predicted signal quality of the predicted beam output by the first beam prediction model. Since the measured signal quality is the optimal signal quality actually measured by the terminal, the first predicted signal quality is the optimal signal quality predicted by the first beam prediction model.
  • the first prediction error value of the first beam prediction model can be obtained by the measured signal quality and the first predicted signal quality. If the first prediction error value is greater than or equal to the specified threshold value, it means that the error of the first beam prediction model is large, that is, the first beam prediction model is not suitable for the current communication scenario between the base station and the terminal. Then, the base station can select the target beam prediction model from multiple second beam prediction models in a timely manner, and switch the effective beam prediction model to the target beam prediction model. Since the prediction error value is less than the specified threshold value, the second prediction model is more suitable for the current communication scenario and can improve the communication quality between the base station and the terminal.
  • the first beam prediction model can also be disabled and fall back to the non-AI mode to use the full beam set for beam prediction.
  • the base station Before falling back to non-AI mode, the base station can configure the measurement period and reporting method of the full beam set for the terminal. After falling back to non-AI mode, the base station sends a pilot signal in the beam direction of each beam included in the full beam set, and receives the measurement results reported by the terminal, and then selects the beam with the best signal quality as the beam for subsequent communication with the terminal based on the measurement results.
  • the method includes:
  • a base station sends a pilot signal to a terminal in a beam direction of each beam included in a first monitoring beam set.
  • the terminal reports the L1_RSRP of each beam in the first monitoring beam set.
  • the base station defines the beam with the largest L1_RSRP as the measured optimal beam.
  • the base station performs beam prediction using the current beam prediction model, and analyzes the performance difference between the predicted optimal beam and the measured optimal beam.
  • the performance difference is the prediction error value introduced in the above embodiment.
  • the method of calculating the prediction error value can refer to the relevant description in the above embodiment, which will not be repeated here.
  • S505 The base station decides to switch the model or fall back to the non-AI mode.
  • the method for the base station to determine whether to switch models and whether to fall back to the AI mode based on the specified threshold value can be referred to the relevant description in the above-mentioned terminal side embodiment and will not be repeated here.
  • the base station sends configuration information of a new measurement beam set to the terminal.
  • the base station decides to switch the model, it sends the configuration information of the monitoring beam set and the measurement beam set corresponding to the switched beam prediction model to the terminal.
  • the configuration information of the full beam set will be sent to the terminal.
  • the air interface measurement overhead and the beam prediction performance can be dynamically adjusted, and the appropriate beam prediction model can be switched in time to meet the needs of different scenarios.
  • the performance of each beam prediction model deteriorates, it can quickly fall back to the non-AI mode to avoid the terminal disconnection.
  • an embodiment of the present application provides a beam prediction model switching device, which is applied to a terminal, the terminal includes a first beam prediction model and multiple second beam prediction models, and the effective beam prediction model of the terminal is the first beam prediction model.
  • the device includes:
  • the measurement module 601 is used to measure the pilot signal of each beam included in the first monitoring beam set to obtain the signal quality corresponding to each beam included in the first monitoring beam set, and select the signal quality with the largest value from the signal qualities corresponding to each beam included in the first monitoring beam set as the measured signal quality;
  • the prediction module 602 is used to obtain the signal quality of each beam included in the first measurement beam set from the signal quality corresponding to each beam included in the first monitoring beam set, and input the signal quality of each beam included in the first measurement beam set into the first beam prediction model to obtain the first predicted signal quality of the predicted beam output by the first beam prediction model;
  • a calculation module 603, configured to calculate a first prediction error value of a first beam prediction model based on the first predicted signal quality and the measured signal quality;
  • the switching module 604 is used to select a target beam prediction model from multiple second beam prediction models if the first prediction error value is greater than or equal to the specified threshold value, and switch the effective beam prediction model to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • the switching module 604 is specifically configured to:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • the prediction module 602 is further used to deactivate the first beam prediction model and perform beam prediction using the full beam set if the first prediction error value and the second prediction error values of multiple second beam prediction models are both greater than or equal to a specified threshold value.
  • calculation module 603 is specifically used for:
  • the difference between the measured signal quality and the first predicted signal quality is used as the first prediction error value; or,
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by a first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • First prediction error value
  • is the filtering factor.
  • the device further comprises:
  • a receiving module is used to receive configuration information of a full beam set, configuration information of at least one measurement beam set, and configuration information of at least one monitoring beam set sent by a base station; and to receive a first RRC message sent by the base station, the first RRC message including an identifier of a first measurement beam set, an identifier of a first monitoring beam set, a specified threshold value, and a filter factor, the first RRC message being used to instruct the terminal to perform beam measurement using the first monitoring beam set and to perform beam prediction using the first measurement beam set.
  • the device further includes a sending module
  • a sending module used to send UCI to a base station, where the UCI includes beam indexes and signal qualities of multiple beams, and the signal qualities of the multiple beams are all minimum signal qualities;
  • the receiving module is also used to receive a first indication message sent by the base station, the first indication message includes an identifier of the full beam set, and the first indication message is used to instruct the terminal to use the configuration information of the full beam set to perform beam measurement and beam prediction.
  • the sending module is further used to send a second RRC message to the base station, where the second RRC message includes an identifier of a third measurement beam set corresponding to the target beam prediction model and an identifier of a third monitoring beam set;
  • the receiving module is also used to receive a second indication message sent by the base station, where the second indication message is used to instruct the terminal to perform beam measurement using the third monitoring beam set and to perform beam prediction using the third measurement beam set.
  • a beam prediction model switching device which is applied to a base station, the base station includes a first beam prediction model and multiple second beam prediction models, and the effective beam prediction model of the base station is the first beam prediction model.
  • the device includes:
  • the acquisition module 701 is used to obtain the signal quality corresponding to each beam included in the first monitoring beam set reported by the terminal, and select the signal quality with the largest value from the signal qualities corresponding to each beam included in the first monitoring beam set as the measured signal quality;
  • the prediction module 702 is used to obtain the signal quality of each beam included in the first measurement beam set from the signal quality corresponding to each beam included in the first monitoring beam set, and input the signal quality of each beam included in the first measurement beam set into the first beam prediction model to obtain the first predicted signal quality of the predicted beam output by the first beam prediction model;
  • a calculation module 703, configured to calculate a first prediction error value of a first beam prediction model based on the first predicted signal quality and the measured signal quality;
  • the switching module 704 is configured to select a target beam from a plurality of second beam prediction models if the first prediction error value is greater than or equal to a specified threshold value.
  • the target beam prediction model is switched from the effective beam prediction model to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • the switching module 704 is specifically configured to:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • the prediction module 702 is further configured to deactivate the first beam prediction model and perform beam prediction using the full beam set if the first prediction error value and the prediction error values of multiple second beam prediction models are both greater than or equal to a specified threshold value.
  • calculation module 703 is specifically used for:
  • the difference between the measured signal quality and the first predicted signal quality is used as the first prediction error value; or,
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by a first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • First prediction error value
  • is the filtering factor.
  • an embodiment of the present application provides a terminal, the terminal includes a first beam prediction model and multiple second beam prediction models, and the effective beam prediction model of the terminal is the first beam prediction model.
  • the terminal includes:
  • a machine-readable storage medium 802 stores machine-executable instructions that can be executed by the processor 801; the machine-executable instructions prompt the processor 801 to perform the following steps:
  • a target beam prediction model is selected from multiple second beam prediction models, and the effective beam prediction model is switched to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • machine executable instructions further cause the processor 801 to perform the following steps:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • machine executable instructions further cause the processor 801 to perform the following steps:
  • the first beam prediction model is disabled, and beam prediction is performed using the full beam set.
  • machine executable instructions further cause the processor 801 to perform the following steps:
  • the difference between the measured signal quality and the first predicted signal quality is used as the first prediction error value; or,
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a signal quality of a predicted beam output by a first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • First prediction error value
  • is the filtering factor.
  • machine executable instructions further cause the processor 801 to perform the following steps:
  • machine executable instructions further cause the processor 801 to perform the following steps:
  • the UCI includes beam indexes and signal qualities of multiple beams, and the signal qualities of the multiple beams are all minimum signal qualities;
  • a first indication message sent by a base station is received, where the first indication message includes an identifier of a full beam set, and the first indication message is used to instruct the terminal to use configuration information of the full beam set to perform beam measurement and beam prediction.
  • machine executable instructions further cause the processor 801 to perform the following steps:
  • the base station Sending a second RRC message to the base station, where the second RRC message includes an identifier of a third measurement beam set corresponding to the target beam prediction model and an identifier of a third monitoring beam set;
  • a second indication message sent by a base station is received, where the second indication message is used to instruct the terminal to perform beam measurement using a third monitoring beam set and to perform beam prediction using a third measurement beam set.
  • a communication bus 803 may also be included.
  • the processor 801, the machine-readable storage medium 802, and the transceiver 804 communicate with each other through the communication bus 803.
  • the communication bus 803 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus.
  • PCI Peripheral Component Interconnect
  • EISA Extended Industry Standard Architecture
  • the communication bus may be divided into an address bus, a data bus, a control bus, and the like.
  • the transceiver 804 may be a wireless communication module. Under the control of the processor 801 , the transceiver 804 exchanges data with other devices.
  • the machine-readable storage medium 802 may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk storage.
  • RAM random access memory
  • NVM non-volatile memory
  • the medium 802 may also be at least one storage device located far away from the aforementioned processor.
  • Processor 801 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • CPU central processing unit
  • NP network processor
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA field programmable gate array
  • the embodiment of the present application further provides a base station, the base station including a first beam prediction model and multiple second beam prediction models, the effective beam prediction model of the base station is the first beam prediction model, as shown in FIG9 , the base station includes:
  • a machine-readable storage medium 902 stores machine-executable instructions that can be executed by the processor 901; the machine-executable instructions prompt the processor 901 to perform the following steps:
  • a target beam prediction model is selected from multiple second beam prediction models, and the effective beam prediction model is switched to the target beam prediction model, and the prediction error value of the target beam prediction model is less than the specified threshold value.
  • machine executable instructions further cause the processor 901 to perform the following steps:
  • the second beam prediction model inputting the signal quality of each beam included in the measurement beam set corresponding to the second beam prediction model into the second beam prediction model to obtain a second predicted signal quality of the predicted beam output by the second beam prediction model;
  • the second beam prediction model is used as the target beam prediction model, and the traversal operation is terminated;
  • machine executable instructions further cause the processor 901 to perform the following steps:
  • the first beam prediction model is disabled and the full beam set is used for beam prediction.
  • machine executable instructions further cause the processor 901 to perform the following steps:
  • the difference between the measured signal quality and the first predicted signal quality is used as the first prediction error value; or,
  • the historical prediction error value is a difference between a measured signal quality obtained in a previous measurement period and a predicted signal quality output by a first beam prediction model in the previous measurement period;
  • the first prediction error value is calculated by the following formula:
  • First prediction error value
  • is the filtering factor.
  • a communication bus 903 may also be included.
  • the processor 901, the machine-readable storage medium 902, and the transceiver 904 are connected via
  • the communication bus 903 completes the communication between each other, and the communication bus 903 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc.
  • PCI Peripheral Component Interconnect
  • EISA Extended Industry Standard Architecture
  • the communication bus can be divided into an address bus, a data bus, a control bus, etc.
  • the transceiver 904 may be a wireless communication module. Under the control of the processor 901 , the transceiver 904 exchanges data with other devices.
  • the machine-readable storage medium 902 may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk storage.
  • the machine-readable storage medium 902 may also be at least one storage device located away from the aforementioned processor.
  • Processor 901 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • CPU central processing unit
  • NP network processor
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA field programmable gate array
  • a computer-readable storage medium in which a computer program is stored.
  • the computer program is executed by a processor, the steps of any of the above-mentioned beam prediction model switching methods are implemented.
  • a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any beam prediction model switching method in the above embodiments.
  • the computer program product includes one or more computer instructions.
  • the process or function described in the embodiment of the present application is generated in whole or in part.
  • the computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
  • the computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center.
  • the computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server, data center, etc. that contains one or more available media integrated.
  • the available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state drive (SSD)), etc.
  • a magnetic medium e.g., a floppy disk, a hard disk, a magnetic tape
  • an optical medium e.g., a DVD
  • a semiconductor medium e.g., a solid state drive (SSD)

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请实施例提供一种波束预测模型切换方法、装置、终端、基站及存储介质,涉及通信技术领域,该方法包括:获得第一监测波束集包括的各波束对应的信号质量,将数值最大的信号质量,作为实测信号质量;从第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并输入第一波束预测模型,获得第一波束预测模型输出的第一预测信号质量;基于第一预测信号质量和实测信号质量计算第一波束预测模型的第一预测误差值;若第一预测误差值大于等于指定门限值,则从多个第二波束预测模型中选择目标波束预测模型,将生效波束预测模型切换为目标波束预测模型。能够及时切换更适合当前场景的波束预测模型。

Description

波束预测模型切换方法、装置、终端、基站及存储介质 技术领域
本申请涉及通信技术领域,尤其涉及一种波束预测模型切换方法、装置、终端、基站及存储介质。
背景技术
在移动通信场景下,用户移动与信号被随机遮挡等因素都会导致信道时变,所以需要反复进行波束测量和跟踪,从而实时准确地确定基站与终端通信的信道信息。传统的波束预测方式为基站在所有发送波束方向上分别发送导频信号,从而预测得到最优发送波束的波束方向,然而这种方式会产生巨大的导频开销。
为了降低导频开销,产生了基于机器学习的波束预测算法,可以利用机器学习的方式训练波束预测模型,从而利用波束预测模型,通过对较少波束的测量结果预测最优发送波束的波束方向。
由于无线信道环境的时变性和场景的多样性,较难得到一种在不同时空均有较优泛化性能的模型。因此,目前可以训练多个波束预测模型,并选择合适的模型进行波束预测。但由于终端移动性和环境变化的影响,适合终端所处场景的波束训练模型会发生变化,如何选择合适的波束预测模型是目前亟需解决的问题。
发明内容
本申请实施例的目的在于提供一种波束预测模型切换方法、装置、终端、基站及存储介质,以及时切换更适合当前场景的波束预测模型,具体技术方案如下:
第一方面,本申请实施例提供一种波束预测模型切换方法,应用于终端,所述终端包括第一波束预测模型和多个第二波束预测模型,所述终端的生效波束预测模型为所述第一波束预测模型,所述方法包括:
对第一监测波束集包括的各波束的导频信号进行测量,得到所述第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
在一种可能的实现方式中,所述从所述多个第二波束预测模型中选择目标波束预测模型,包括:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
在一种可能的实现方式中,在所述基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值之后,所述方法还包括:
若所述第一预测误差值和所述多个第二波束预测模型的第二预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,并利用全量波束集进行波束预测。
在一种可能的实现方式中,所述基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值,包括:
将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算所述第一预测误差值:
所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
在一种可能的实现方式中,在所述对第一监测波束集包括的各波束的导频信号进行测量之前,所述方法还包括:
接收基站发送的全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息;
接收所述基站发送的第一RRC消息,所述第一RRC消息包括所述第一测量波束集的标识、所述第一监测波束集的标识、所述指定门限值以及所述滤波因子,所述第一RRC消息用于指示所述终端利用所述第一监测波束集进行波束测量,利用所述第一测量波束集进行波束预测。
在一种可能的实现方式中,在所述停用所述第一波束预测模型之后,所述方法还包括:
向所述基站发送UCI,所述UCI包括多个波束的波束索引和信号质量,所述多个波束的信号质量均为信号质量最小值;
接收所述基站发送的第一指示消息,所述第一指示消息包括所述全量波束集的标识,所述第一指示消息用于指示所述终端利用所述全量波束集的配置信息进行波束测量和波束预测。
在一种可能的实现方式中,在所述从所述多个第二波束预测模型中选择目标波束预测模型之后,所述方法还包括:
向所述基站发送第二RRC消息,所述第二RRC消息包括所述目标波束预测模型对应的第三测量波束集的标识以及第三监测波束集的标识;
接收所述基站发送的第二指示消息,所述第二指示消息用于指示所述终端利用所述第三监测波束集进行波束测量,并利用所述第三测量波束集进行波束预测。
第二方面,本申请实施例提供一种波束预测模型切换方法,应用于基站,所述基站包括第一波束预测模型和多个第二波束预测模型,所述基站的生效波束预测模型为所述第一波束预测模型,所述方法包括:
获取终端上报的第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号 质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
在一种可能的实现方式中,所述从所述多个第二波束预测模型中选择目标波束预测模型,包括:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
在一种可能的实现方式中,在所述基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值之后,所述方法还包括:
若所述第一预测误差值和所述多个第二波束预测模型的预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,利用全量波束集进行波束预测。
在一种可能的实现方式中,所述基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值,包括:
将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算所述第一预测误差值:
所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
第三方面,本申请实施例提供一种波束预测模型切换装置,应用于终端,所述终端包括第一波束预测模型和多个第二波束预测模型,所述终端的生效波束预测模型为所述第一波束预测模型,所述装置包括:
测量模块,用于对第一监测波束集包括的各波束的导频信号进行测量,得到所述第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
预测模块,用于从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
计算模块,用于基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
切换模块,用于若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
在一种可能的实现方式中,所述切换模块,具体用于:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
在一种可能的实现方式中,所述预测模块,还用于若所述第一预测误差值和所述多个第二波束预测模型的第二预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,并利用全量波束集进行波束预测。
在一种可能的实现方式中,所述计算模块,具体用于:
将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算所述第一预测误差值:
所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
在一种可能的实现方式中,所述装置还包括:
接收模块,用于接收基站发送的全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息;以及接收所述基站发送的第一RRC消息,所述第一RRC消息包括所述第一测量波束集的标识、所述第一监测波束集的标识、所述指定门限值以及所述滤波因子,所述第一RRC消息用于指示所述终端利用所述第一监测波束集进行波束测量,利用所述第一测量波束集进行波束预测。
在一种可能的实现方式中,所述装置还包括发送模块;
所述发送模块,用于向所述基站发送UCI,所述UCI包括多个波束的波束索引和信号质量,所述多个波束的信号质量均为信号质量最小值;
所述接收模块,还用于接收所述基站发送的第一指示消息,所述第一指示消息包括所述全量波束集的标识,所述第一指示消息用于指示所述终端利用所述全量波束集的配置信息进行波束测量和波束预测。
在一种可能的实现方式中,所述发送模块,还用于向所述基站发送第二RRC消息,所述第二RRC消息包括所述目标波束预测模型对应的第三测量波束集的标识以及第三监测波束集的标识;
所述接收模块,还用于接收所述基站发送的第二指示消息,所述第二指示消息用于指示所述终端利用所述第三监测波束集进行波束测量,并利用所述第三测量波束集进行波束预测。
第四方面,本申请实施例提供一种波束预测模型切换装置,应用于基站,所述基站包括第一波束预测模型和多个第二波束预测模型,所述基站的生效波束预测模型为所述第一波束预测模型,所述装置包 括:
获取模块,用于获取终端上报的第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
预测模块,用于从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
计算模块,用于基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
切换模块,用于若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
在一种可能的实现方式中,所述切换模块,具体用于:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
在一种可能的实现方式中,所述预测模块,还用于若所述第一预测误差值和所述多个第二波束预测模型的预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,利用全量波束集进行波束预测。
在一种可能的实现方式中,所述计算模块,具体用于:
将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算所述第一预测误差值:
所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
第五方面,本申请实施例提供一种终端,所述终端包括第一波束预测模型和多个第二波束预测模型,所述终端的生效波束预测模型为所述第一波束预测模型,所述终端包括:
处理器;收发器;
机器可读存储介质,所述机器可读存储介质存储有能够被所述处理器执行的机器可执行指令;所述机器可执行指令促使所述处理器执行以下步骤:
对第一监测波束集包括的各波束的导频信号进行测量,得到所述第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
若所述第一预测误差值和所述多个第二波束预测模型的第二预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,并利用全量波束集进行波束预测。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算所述第一预测误差值:
所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
接收基站发送的全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息;
接收所述基站发送的第一RRC消息,所述第一RRC消息包括所述第一测量波束集的标识、所述第一监测波束集的标识、所述指定门限值以及所述滤波因子,所述第一RRC消息用于指示所述终端利用所述第一监测波束集进行波束测量,利用所述第一测量波束集进行波束预测。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
向所述基站发送UCI,所述UCI包括多个波束的波束索引和信号质量,所述多个波束的信号质量均为信号质量最小值;
接收所述基站发送的第一指示消息,所述第一指示消息包括所述全量波束集的标识,所述第一指示消息用于指示所述终端利用所述全量波束集的配置信息进行波束测量和波束预测。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
向所述基站发送第二RRC消息,所述第二RRC消息包括所述目标波束预测模型对应的第三测量波束集的标识以及第三监测波束集的标识;
接收所述基站发送的第二指示消息,所述第二指示消息用于指示所述终端利用所述第三监测波束集进行波束测量,并利用所述第三测量波束集进行波束预测。
第六方面,本申请实施例提供一种基站,所述基站包括第一波束预测模型和多个第二波束预测模型,所述基站的生效波束预测模型为所述第一波束预测模型,所述基站包括:
处理器;收发器;
机器可读存储介质,所述机器可读存储介质存储有能够被所述处理器执行的机器可执行指令;所述机器可执行指令促使所述处理器执行以下步骤:
获取终端上报的第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
若所述第一预测误差值和所述多个第二波束预测模型的预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,利用全量波束集进行波束预测。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算所述第一预测误差值:
所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
第七方面,本申请实施例提供一种机器可读存储介质,存储有机器可执行指令,在被处理器调用和 执行时,所述机器可执行指令促使所述处理器:实现第一方面或第二方面所述的方法。
第八方面,本申请实施例提供一种计算机程序产品,所述计算机程序产品促使所述处理器:实现第一方面或第二方面所述的方法。
采用上述技术方案,终端测量得到第一监测波束集中各波束对应的信号质量,并将第一监测波束集包括的各波束对应的信号质量中数值最大的信号质量,作为实测信号质量。然后从第一监测波束集对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将第一测量波束集包括的各波束的信号量输入第一波束预测模型,以获得第一波束预测模型输出的预测波束的第一预测信号质量。由于实测信号质量是终端实际测量得到的最优信号质量,第一预测信号质量是第一波束预测模型预测得到的最优信号质量。因此,通过实测信号质量和第一预测信号质量可以得到第一波束预测模型的第一预测误差值,若该第一预测误差值大于等于指定门限值,则说明第一波束预测模型的误差较大,即第一波束预测模型不适用于基站与终端的当前通信场景。进而终端可及时从多个第二波束预测模型中选择目标波束预测模型,并将生效波束预测模型切换为目标波束预测模型。由于预测误差值小于指定门限值,因此第二预测模型更适用于当前通信场景,可以提升基站与终端之间的通信质量。
附图说明
此处所说明的附图用来提供对本申请的进一步理解,构成本申请的一部分,本申请的示意性实施例及其说明用于解释本申请,并不构成对本申请的不当限定。
图1为本申请实施例提供的一种波束预测模型切换方法流程图;
图2为本申请实施例提供的另一种波束预测模型切换方法的流程图;
图3为本申请实施例提供的另一种波束预测模型切换方法的流程图;
图4为本申请实施例提供的另一种波束预测模型切换方法的流程图;
图5为本申请实施例提供的另一种波束预测模型切换方法的流程图;
图6为本申请实施例提供的一种波束预测模型装置的结构示意图;
图7为本申请实施例提供的另一种波束预测模型装置的结构示意图;
图8为本申请实施例提供的一种终端的结构示意图;
图9为本申请实施例提供的一种终端的结构示意图。
具体实施方式
为使本申请的目的、技术方案、及优点更加清楚明白,以下参照附图并举实施例,对本申请进一步详细说明。显然,所描述的实施例仅仅是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员所获得的所有其他实施例,都属于本申请保护的范围。
本申请实施例中涉及的波束预测模型为预先训练好的人工智能(Artificial Intelligence,AI)模型,波束预测模型可以部署在终端,也可以部署在基站。波束预测模型用于利用少量波束的信号质量预测全量波束中的最优波束。波束预测模型的输入可以为少量波束的波束索引及对应的信号质量,输出可以为预测得到的最优波束的波束索引及信号质量,或者输出为预测得到的信号质量排前N的波束的波束索引及信号质量。
在终端中部署多个波束预测模型的情况下,终端进行波束测量后,利用波束测量结果对当前生效的波束预测模型进行性能监测,确定是否需要切换当前生效的波束预测模型。
在基站中部署多个波束预测模型的情况下,终端将波束测量结果上报给基站,基站根据波束测量结 果对当前生效的波束预测模型进行性能监测,以定是否需要切换当前生效的波束预测模型。
以下对本申请实施例提高的波束预测模型切换方法进行详细介绍。
本申请实施例提供了一种波束预测模型切换方法,该方法应用于终端,终端包括第一波束预测模型和多个第二波束预测模型,终端的生效波束预测模型为第一波束预测模型,如图1所示,该方法包括:
S101、对第一监测波束集包括的各波束的导频信号进行测量,得到第一监测波束集包括的各波束对应的信号质量,并从第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量。
其中,实测信号质量为第一监测波束集包括的各波束对应的信号质量中数值最大的信号质量,也就是实际测量得到的最优信号质量。
可以理解的,基站按照预先配置的测量周期,在第一监测波束集包括的各波束的波束方向上发送导频信号。相应地,终端对各波束方向上的导频信号进行测量,得到各波束对应的信号质量,并从测量得到的信号质量中确定出数值最大的信号质量,将数值最大的信号质量作为实测信号质量。
其中,信号质量可以通过层1参考信号接收功率(Layer1 Reference Signal Receiving Power,L1-RSRP)表示。L1-RSRP可用于表示物理层导频信号强度,L1-RSRP越大,所表示的信号质量越好。
S102、从第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将第一测量波束集包括的各波束的信号质量输入第一波束预测模型,获得第一波束预测模型输出的预测波束的第一预测信号质量。
其中,第一波束预测模型的输入参数为第一测量波束集包括的各波束的信号质量。第一测量波束集为第一监测波束集的子集,即第一监测波束集比第一测量波束集包括的波束更多,以达到对第一波束预测模型进行性能监测的目的。
由于第一测量波束集是第一监测波束集的子集,在终端测量得到第一监测波束集包括的各波束对应的信号质量之后,终端即可从中获取第一测量波束集中各波束对应的信号质量。
第一波束预测模型可以利用第一测量波束集中各波束对应的信号质量,预测出当前基站与终端之间的最优波束,即上述预测波束。并且第一波束预测模型还能够预测出该预测波束的信号质量,即第一预测信号质量。
S103、基于第一预测信号质量和实测信号质量计算第一波束预测模型的第一预测误差值。
可以理解的,第一预测信号质量是第一波束预测模型预测得到的最优的预测波束的信号质量,实测信号质量是实际测量得到的第一监测波束集中各波束中最优波束的信号质量。因此,第一预测误差值能够表示第一波束预测模型预测得到的最优波束的信号质量和实际测量得到的最优波束的实测信号质量之间的误差。
S104、若第一预测误差值大于等于指定门限值,则从多个第二波束预测模型中选择目标波束预测模型,将生效波束预测模型切换为目标波束预测模型,目标波束预测模型的预测误差值小于指定门限值。
可以理解的,若第一预测误差值大于等于指定门限值,表明第一波束预测模型的预测误差较大,第一波束预测模型无法较为准确地预测最优波束。进而终端可以从多个第二波束预测模型中选择预测误差值小于指定门限值的目标波束预测模型,以避免基站与终端之间的通信质量恶化。
采用上述方法,终端测量得到第一监测波束集中各波束对应的信号质量,并将第一监测波束集包括的各波束对应的信号质量中数值最大的信号质量,作为实测信号质量。然后从第一监测波束集对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将第一测量波束集包括的各波束的信号量输 入第一波束预测模型,以获得第一波束预测模型输出的预测波束的第一预测信号质量。由于实测信号质量是终端实际测量得到的最优信号质量,第一预测信号质量是第一波束预测模型预测得到的最优信号质量。因此,通过实测信号质量和第一预测信号质量可以得到第一波束预测模型的第一预测误差值,若该第一预测误差值大于等于指定门限值,则说明第一波束预测模型的误差较大,即第一波束预测模型不适用于基站与终端的当前通信场景。进而终端可及时从多个第二波束预测模型中选择目标波束预测模型,并将生效波束预测模型切换为目标波束预测模型。由于预测误差值小于指定门限值,因此第二预测模型更适用于当前通信场景,可以提升基站与终端之间的通信质量。
在执行图1所示的方法之前,基站还需对终端进行配置,即在S101之前,还需执行以下步骤:
终端接收基站发送的全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息。接收基站发送的第一RRC消息,第一RRC消息包括第一测量波束集的标识、第一监测波束集的标识、指定门限值以及滤波因子,第一RRC消息用于指示终端利用第一监测波束集进行波束测量,利用第一测量波束集进行波束预测。
一种实施方式中,至少一个测量波束集为基站支持的所有测量波束集,至少一个监测波束集为基站支持的所有监测波束集。
另一种实施方式中,至少一个测量波束集为终端中的各波束预测模型对应的测量波束集,至少一个监测波束集为终端中的各波束预测模型对应的监测波束集。
其中,波束预测模型的输入为测量波束集包括的各波束的信号质量。终端中不同波束预测模型的输入可以相同或不同,在输入不同的情况下,对应的测量波束集也不同。
不同测量波束集对应的监测波束集可以相同或不同。监测波束集可以预先设置,比如测量波束集和监测波束集可以为一对一的关系,可以在测量波束集的基础上增加一些波束,得到监测波束集。或者,测量波束集和监测波束集可以为多对一的关系,即多个测量波束集均为同一监测波束集的子集。
基站支持的每个波束对应一个波束索引。各波束集的配置信息均包括:波束集标识、波束索引、测量周期和波束数目。
终端可以存储全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息。进而当接收到第一RRC消息时,利用第一测量波束集的标识查找第一测量波束集的配置信息,利用第一监测波束集的标识查找第一监测波束集的配置信息。然后按照第一监测波束集的配置信息对第一监测波束集包括的各波束进行波束测量,并利用第一测量波束集包括的各波束的信号质量进行波束预测。可以理解为,基站可以通过第一RRC消息激活第一测量波束集和第一监测波束集,以使得终端开始使用第一测量波束集和第一监测波束集。
需要说明的是,终端中的各波束预测模型对应的输入参数具有一定的灵活性,若第一测量波束集与第一波束预测模型的当前输入参数不完全匹配,则可调整第一波束预测模型的输入参数,使得第一波束预测模型的输入参数与第一测量波束集适配。
采用该方法,基站可以预先向终端下发全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息,后续通过第一RRC消息即可为终端配置终端所需使用的测量波束集和监测波束集,进而无需在终端每次切换波束预测模型时重新下发测量波束集的配置信息和监测波束集的配置信息,可以提高模型切换效率。
在本申请的另一实施例中,针对上述S103、基于第一预测信号质量和实测信号质量计算第一波束预测模型的第一预测误差值,存在以下两种实现方式:
方式1、将实测信号质量和第一预测信号质量的差值作为第一预测误差值。
其中,终端在第一监测波束集的每个测量周期均可确定实测信号质量和第一预测信号质量,进而在每个测量周期计算第一预测误差值。
方式2、计算实测信号质量和第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算第一预测误差值:
第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;其中,α为滤波因子。
在信号质量为L1-RSRP的情况下,当前预测误差值可以表示为L1_RSRP_Gap_Current,历史预测误差值可以表示为L1_RSRP_Gap_Prev,则第一预测误差值Gap=|α*L1_RSRP_Gap_Current+(1-α)*L1_RSRP_Gap_Prev|。
在方式2中,由于在第一监测波束集的首个测量周期内,不存在历史预测误差值,所以可将首个测量周期计算得到的当前预测误差值作为首个测量周期的第一预测误差值。进而在第二个测量周期,可将首个测量周期的当前预测误差值作为历史预测误差值,进而利用上述公式计算第二个测量周期的第一预测误差值。同理,在第三个测量周期,可以将第二个测量周期的当前预测误差值作为历史预测误差值。
针对首个测量周期,上述第一预测误差值可以体现出实际测量得到实测信号质量和第一预测信号质量的差异,该差异可以反应第一波束预测模型在终端和基站之间的当前通信场景中的性能。且在方式2中,第一预测误差值是通过对当前预测误差值和历史预测误差值进行滤波得到的,能够准确反映第一波束预测模型在两个连续测量周期内的性能,使得后续能够更准确地基于第一预测误差值和指定门限值进行模型切换。由于切换条件越宽松,模型切换越频繁,相应地模型切换导致的空口开销越大。切换条件越严格,模型切换越不频繁,相应地可能会导致模型切换不及时。本申请实施例通过设置滤波因子和指定门限值,能够使得切换条件更加合理,能够在切换的实时性和空口开销之间进行权衡,最大化提升整体性能。
在本申请的一些实施例中,在第一预测误差值大于等于指定门限值的情况下,终端可以遍历多个第二波束预测模型,并从中选择目标波束预测模型。选择过程包括以下步骤:
依次对每个第二波束预测模型进行以下遍历操作:
步骤1、针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得第二波束预测模型输出的预测波束的第二预测信号质量。
其中,第二波束预测模型对应的测量波束集与第二波束测量模型的输入参数对应。终端可以从基站发送的至少一个测量波束集的配置信息中获取该测量波束集的配置信息,从而从该测量波束集的配置信息中获取该测量波束集包括的波束索引。
然后,终端根据测量波束集包括的波束索引,从第一监测波束集包括的各波束对应的信号质量中,获取测量波束集包括的各波束对应的信号质量。
若第一监测波束集包括该测量波束集的全部波束,则可直接获取该测量波束集包括的各波束对应的信号质量;若第一监测波束集不包括该测量波束集的全部波束,则终端可以从第一监测波束集中获取测量波束集不包括的部分波束对应的信号质量,从而获取到与该测量波束集包括的波束数量相同的信号质量,然后对第二波束预测模型的输入参数进行适应调整,并将获得的各波束对应的信号质量输入第二波束预测模型。
或者,第一RRC消息中可以包括多个监测波束集的标识,进而终端会对多个监测波束集包括的波束进行测量,从而得到每个监测波束集包括的各波束的信号质量。
在本步骤中,终端可从第二波束预测模型对应的监测波束集包括的各波束的信号质量中,获取第二波束预测模型对应的测量波束集包括的各波束的信号质量。进而可将该测量波束集包括的各波束的信号质量输入第二波束预测模型。
步骤2、将实测信号质量与第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值。
该实测信号质量可以为第一监测波束集包括的各波束的信号质量中的数值最大的信号质量。或者,在第一RRC消息中包括多个监测波束集的标识的情况下,该实测信号质量为第二波束预测模型对应的监测波束集包括的各波束的信号质量中的数值最大的信号质量。
步骤3、若第二预测误差值小于指定门限值,则将该第二波束预测模型作为目标波束预测模型,并结束遍历操作。
进而,终端将生效波束预测模型切换为目标波束预测模型。
步骤4、若第二预测误差值大于等于指定门限值,则继续遍历操作。
直至查找到第二预测误差值小于指定门限值的第二预测模型时,结束遍历操作。
若完成对所有第二波束预测模型的遍历操作后,确定所有第二波束预测模型的第二预测误差值均大于等于指定门限值,则停止遍历操作。
此时,终端可确定第一预测误差值和多个第二波束预测模型的预测误差值均大于等于指定门限值,即不存在适用于当前场景的波束预测模型,则停用第一波束预测模型,利用全量波束集进行波束预测。
在不存在适用于当前场景的波束预测模型的情况下,终端可以回退至非AI模式,即利用全量波束集进行波束预测,能够实现在波束预测模型性能恶化的情况下,或者在波束预测模型部署初期,波束预测模型的泛化性能不佳的情况下,快速回退到非AI模式,以避免影响基站与终端之间的通信质量。
其中,终端利用全量波束集进行波束预测,是指基站在所有波束方向上分别发送导频信号,终端对各波束方向的导频信号进行测量,从而预测得到最优信号质量对应的波束,后续终端与基站之前可利用该最优信号质量对应的波束进行通信。
具体的,在第一预测误差值和多个第二波束预测模型的预测误差值均大于等于指定门限值的情况下,停用第一波束预测模型之后,终端可以通过向基站发送上行控制信息(Uplink Control Information,UCI)的方式,通知基站各波束预测模型均已不适用,以向基站发起模型回退请求。在停用第一波束预测模型之后,终端可执行以下步骤:
步骤A、向基站发送UCI,
其中,本申请实施例中可以复用UCI测量上报格式,UCI包括多个波束的波束索引和信号质量,多个波束的信号质量均为信号质量最小值。其中,信号质量最小值为-140dBm(毫瓦分贝)。本申请实施例中通过将目前的UCI测量上报格式中的各信号质量设置为最小值,即可通知基站各波束预测模型均已不适用,无需设计专用的模型回退请求,实现较为简单。
UCI中包括的波束索引的数量是基站预先为终端配置的。
如表1所示,表1示例性示出了一种UCI格式。
表1

步骤B、接收基站发送的第一指示消息,第一指示消息包括全量波束集的标识。第一指示消息用于指示终端利用全量波束集的配置信息进行波束测量和波束预测。即第一指示消息激活全量波束集。
终端接收到第一指示消息后,即可停用第一波束预测模型,并回退至非AI模式。
第一指示消息可以为传输配置指示信息(Transmission Configuration Indicator,TCI)、媒体接入控制的控制单元(Media Access Control-Control Element,MAC-CE)消息或RRC消息。
采用上述方法,终端通过将UCI中的信号质量设置为最小值,从而通知基站当前不存在适用的波束预测模型,进而基站通过第一指示消息即可指示终端利用全量波束集进行波束预测,可以快速回退至非AI模式。并且,终端与基站之间可利用已有的消息类型实现模型回退,对目前的标准协议影响较小。
可以理解的是,在终端从多个第二波束预测模型中选择目标波束预测模型之后,终端还需向基站请求后续适用于第二波束预测模型的测量波束集和监测波束集。
终端向基站发送第二RRC消息,第二RRC消息包括目标波束预测模型对应的第三波束集的标识以及第三监测波束集的标识。进而终端接收基站发送的第二指示消息,第二指示消息用于指示终端利用第三监测波束集进行波束测量,并利用第三测量波束集进行波束预测。
其中,第二指示消息包括第三测量波束集的标识和第三监测波束集的标识。可以理解为,基站通过第二指示消息可以激活第三测量波束集和第三监测波束集,并去激活第一测量波束集和第一监测波束集。
第二指示消息可以为TCI消息、MAC-CE消息或RRC消息。
由于基站已提前向终端下发各测量波束集的配置信息和各监测波束集的配置信息,所以基站通过第二指示消息即可启动终端对第三测量波束集和第三监测波束集的测量,可以实现波束预测模型的快速切换。
以下结合具体例子介绍本申请实施例提供的波束预测模型切换方法,假设终端被配置为使用监测波束集Set_monitor进行波束测量,使用测量波束集Set_measure进行波束预测。如图2所示,该方法包括:
S201、对监测波束集Set_monitor对应的L1_RSRP依次遍历,获取性能最优波束的L1_RSRP,记为L1_RSRP_Best。
其中,性能最优波束即为实测信号质量对应的波束。终端从Set_monitor包括的各波束中,遍历查找L1_RSRP最大的波束,并将该L1_RSRP记为L1_RSRP_Best。图2中S201左侧的黑色方格为性能最优波束。
S202、从Set_monitor包括的各波束的L1_RSRP中,抽取测量波束集Set_measure包括的各波束的 L1_RSRP,作为当前波束预测模型的输入。
其中,图2中S202左侧的黑色方格为Set_measure包括的波束。
S203、基于当前波束预测模型,预测Set_measure中的最优波束及该波束的L1_RSRP,记为L1_RSRP_Predict。
其中,图2中S203左侧带阴影的方格为波束预测模型预测得到的最优波束。黑色方格为S201中的性能最优波束。
S204、计算L1_RSRP_Best与L1_RSRP_Predict之间的差值并进行滤波,得到预测误差值。
其中,进行滤波的方式可参考上述实施例中利用滤波因子计算第一预测误差值的公式。
S205、判断预测误差值是否小于指定门限值。
若是,执行S206,若否,执行S209。
S206、当前遍历的波束预测模型是否为终端中的生效波束预测模型。
若是,执行S207,若否,执行S208。
S207、波束预测模型不变,无需操作。
S208、发起模型切换请求。
其中,该模型切换请求可以为上述实施例中的第二RRC消息。
S209、判断是否已遍历所有的波束预测模型。
若是,执行S210,若否,执行S211。
S210、波束预测模型选择失败,回退到非AI模式。
其中,模型选择失败是指终端中部署的波束预测模型对应的预测误差值均大于等于指定门限值。
S211、选择下一波束预测模型。并针对下一波束预测模型返回S201,如此可对各波束预测模型进行遍历。
采用上述方法,通过设置滤波因子和指定门限值,可以在空口测量开销和波束预测性能间动态调整,及时切换合适的波束预测模型,适应不同场景的需求。并且,在各波束预测模型的性能均恶化的情况下,能够快速回退到非AI模式,以避免终端掉线。
在波束预测模型部署在终端的情况下,如图3所示,图3示出了终端与基站之间的交互流程,该方法包括:
S301、基站为终端下发全量波束集的配置信息、测量波束集的配置信息和监测波束集的配置信息。
S302、基站激活测量波束集和监测波束集,并下发滤波因子和指定门限值。
其中,该步骤可以通过上述实施例中介绍的第一RRC消息实现,此处不再赘述。
S303、终端对监测波束集包括的各波束的导频信号进行测量,获取每个波束对应的L1_RSRP,将L1_RSRP最大的波束定义为实测最优波束。
其中,实测最优波束为上述实测信号质量对应的波束。
S304、终端利用当前的波束预测模型进行波束预测,分析预测的最优波束和实测最优波束的性能差异。
该性能差异为上述实施例中介绍的预测误差值,计算预测误差值的方式可参考上述实施例中的相关描述,此处不再赘述。
S305、终端决策进行模型切换或者回退到非AI模式。
终端基于指定门限值确定是否进行模型切换以及是否回退到AI模式的方法可参考上述实施例中的 相关描述,此处不在赘述。
S306、终端向基站发送模型切换请求或回退至非AI模式的请求。
S307、基站对终端进行监测波束集和测量波束集的激活/去激活操作。
其中,该步骤可通过上述实施例中的第一指示消息或第二指示消息实现。
如此,在终端侧配置波束测量模型的情况下,可以实现对波束测量模型的合理切换,以保证基站与终端之间的通信质量。
对应于上述方法实施例,本申请实施例还提供一种波束预测模型切换方法,该方法应用于基站,基站包括第一波束预测模型和多个第二波束预测模型,基站的生效波束预测模型为第一波束预测模型,如图4所示,该方法包括:
S401、获取终端上报的第一监测波束集包括的各波束对应的信号质量,并从第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量。
其中,基站预先为终端配置了第一监测波束集的测量周期及上报方式,进而基站按照预先配置的测量周期,在第一监测波束集包括的各波束的波束方向上发送导频信号。相应地,终端对各波束方向上的导频信号进行测量,得到各波束对应的信号质量,向基站上报各波束对应的信号质量。
S402、从第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将第一测量波束集包括的各波束的信号质量输入第一波束预测模型,获得第一波束预测模型输出的预测波束的第一预测信号质量。
其中,第一测量波束集为第一监测波束集的子集。
S403、基于第一预测信号质量和实测信号质量计算第一波束预测模型的第一预测误差值。
S404、若第一预测误差值大于等于指定门限值,则从多个第二波束预测模型中选择目标波束预测模型,将生效波束预测模型切换为目标波束预测模型,目标波束预测模型的预测误差值小于指定门限值。
采用上述方法,基站获取终端上报的第一监测波束集中各波束对应的信号质量,并将第一监测波束集包括的各波束对应的信号质量中数值最大的信号质量,作为实测信号质量。然后从第一监测波束集对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将第一测量波束集包括的各波束的信号量输入第一波束预测模型,以获得第一波束预测模型输出的预测波束的第一预测信号质量。由于实测信号质量是终端实际测量得到的最优信号质量,第一预测信号质量是第一波束预测模型预测得到的最优信号质量。因此,通过实测信号质量和第一预测信号质量可以得到第一波束预测模型的第一预测误差值,若该第一预测误差值大于等于指定门限值,则说明第一波束预测模型的误差较大,即第一波束预测模型不适用于基站与终端的当前通信场景。进而基站可及时从多个第二波束预测模型中选择目标波束预测模型,并将生效波束预测模型切换为目标波束预测模型。由于预测误差值小于指定门限值,因此第二预测模型更适用于当前通信场景,可以提升基站与终端之间的通信质量。
需要说明的是,上述S402-S404的实现方式与终端侧S102-S104的实现方式相同,可参考上述实施例中的相关描述,此处不再赘述。
在基站确定第一预测误差值和多个第二波束预测模型的预测误差值均大于等于指定门限值的情况下,也可停用第一波束预测模型,并回退至非AI模式,利用全量波束集进行波束预测。
在回退至非AI模式之前,基站可为终端配置全量波束集的测量周期和上报方式,在回退至非AI模式后,基站在全量波束集包括的各波束的波束方向上发送导频信号,并接收终端上报的测量结果,进而依据测量结果选择信号质量最优的波束作为后续与终端通信的波束。
以下结合图5介绍在波束预测模型部署在基站的情况下,波束预测模型切换方法的完整流程,如图5所示,该方法包括:
S501、基站在第一监测波束集包括的各波束的波束方向上向终端发送导频信号。
S502、终端上报第一监测波束集中各波束的L1_RSRP。
S503、基站将L1_RSRP最大的波束定义为实测最优波束。
S504、基站利用当前的波束预测模型进行波束预测,分析预测的最优波束和实测最优波束的性能差异。
该性能差异为上述实施例中介绍的预测误差值,计算预测误差值的方式可参考上述实施例中的相关描述,此处不再赘述。
S505、基站决策进行模型切换或者回退到非AI模式。
基站基于指定门限值确定是否进行模型切换以及是否回退到AI模式的方法可参考上述终端侧实施例中的相关描述,此处不在赘述。
S506、基站向终端下发新的测量波束集的配置信息。
若基站决策进行模型切换,则向终端下发切换后的波束预测模型对应的监测波束集的配置信息和测量波束集的配置信。
若基站决策回退至非AI模式,则向终端下发全量波束集的配置信息。
采用上述方法,在基站侧配置波束测量模型的情况下,通过设置滤波因子和指定门限值,可以在空口测量开销和波束预测性能间动态调整,及时切换合适的波束预测模型,适应不同场景的需求。并且,在各波束预测模型的性能均恶化的情况下,能够快速回退到非AI模式,以避免终端掉线。
基于相同的构思,本申请实施例提供一种波束预测模型切换装置,应用于终端,终端包括第一波束预测模型和多个第二波束预测模型,终端的生效波束预测模型为第一波束预测模型,如图6所示,该装置包括:
测量模块601,用于对第一监测波束集包括的各波束的导频信号进行测量,得到第一监测波束集包括的各波束对应的信号质量,并从第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
预测模块602,用于从第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将第一测量波束集包括的各波束的信号质量输入第一波束预测模型,获得第一波束预测模型输出的预测波束的第一预测信号质量;
计算模块603,用于基于第一预测信号质量和实测信号质量计算第一波束预测模型的第一预测误差值;
切换模块604,用于若第一预测误差值大于等于指定门限值,则从多个第二波束预测模型中选择目标波束预测模型,将生效波束预测模型切换为目标波束预测模型,目标波束预测模型的预测误差值小于指定门限值。
可选的,切换模块604,具体用于:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将实测信号质量与第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若第二预测误差值小于指定门限值,则将该第二波束预测模型作为目标波束预测模型,并结束遍历操作;
若第二预测误差值大于等于指定门限值,则继续遍历操作。
可选的,预测模块602,还用于若第一预测误差值和多个第二波束预测模型的第二预测误差值均大于等于指定门限值,则停用第一波束预测模型,并利用全量波束集进行波束预测。
可选的,计算模块603,具体用于:
将实测信号质量和第一预测信号质量的差值作为第一预测误差值;或者,
计算实测信号质量和第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算第一预测误差值:
第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
可选的,该装置还包括:
接收模块,用于接收基站发送的全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息;以及接收基站发送的第一RRC消息,第一RRC消息包括第一测量波束集的标识、第一监测波束集的标识、指定门限值以及滤波因子,第一RRC消息用于指示终端利用第一监测波束集进行波束测量,利用所述第一测量波束集进行波束预测。
可选的,该装置还包括发送模块;
发送模块,用于向基站发送UCI,UCI包括多个波束的波束索引和信号质量,多个波束的信号质量均为信号质量最小值;
接收模块,还用于接收基站发送的第一指示消息,第一指示消息包括全量波束集的标识,第一指示消息用于指示终端利用全量波束集的配置信息进行波束测量和波束预测。
可选的,发送模块,还用于向基站发送第二RRC消息,第二RRC消息包括目标波束预测模型对应的第三测量波束集的标识以及第三监测波束集的标识;
接收模块,还用于接收基站发送的第二指示消息,第二指示消息用于指示终端利用第三监测波束集进行波束测量,并利用第三测量波束集进行波束预测。
在本申请的另一实施例中,还提供一种波束预测模型切换装置,应用于基站,基站包括第一波束预测模型和多个第二波束预测模型,基站的生效波束预测模型为第一波束预测模型,如图7所示,该装置包括:
获取模块701,用于获取终端上报的第一监测波束集包括的各波束对应的信号质量,并从第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
预测模块702,用于从第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将第一测量波束集包括的各波束的信号质量输入第一波束预测模型,获得第一波束预测模型输出的预测波束的第一预测信号质量;
计算模块703,用于基于第一预测信号质量和实测信号质量计算第一波束预测模型的第一预测误差值;
切换模块704,用于若第一预测误差值大于等于指定门限值,则从多个第二波束预测模型中选择目 标波束预测模型,将生效波束预测模型切换为目标波束预测模型,目标波束预测模型的预测误差值小于指定门限值。
可选的,切换模块704,具体用于:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将实测信号质量与第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若第二预测误差值小于指定门限值,则将该第二波束预测模型作为目标波束预测模型,并结束遍历操作;
若第二预测误差值大于等于指定门限值,则继续遍历操作。
可选的,预测模块702,还用于若第一预测误差值和多个第二波束预测模型的预测误差值均大于等于指定门限值,则停用第一波束预测模型,利用全量波束集进行波束预测。
可选的,计算模块703,具体用于:
将实测信号质量和第一预测信号质量的差值作为第一预测误差值;或者,
计算实测信号质量和第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算第一预测误差值:
第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
基于相同的构思,本申请实施例提供一种终端,终端包括第一波束预测模型和多个第二波束预测模型,终端的生效波束预测模型为第一波束预测模型,如图8所示,该终端包括:
处理器801;收发器804;
机器可读存储介质802,机器可读存储介质802存储有能够被处理器801执行的机器可执行指令;机器可执行指令促使处理器801执行以下步骤:
对第一监测波束集包括的各波束的导频信号进行测量,得到第一监测波束集包括的各波束对应的信号质量,并从第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
从第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将第一测量波束集包括的各波束的信号质量输入第一波束预测模型,获得第一波束预测模型输出的预测波束的第一预测信号质量;
基于第一预测信号质量和实测信号质量计算第一波束预测模型的第一预测误差值;
若第一预测误差值大于等于指定门限值,则从多个第二波束预测模型中选择目标波束预测模型,将生效波束预测模型切换为目标波束预测模型,目标波束预测模型的预测误差值小于指定门限值。
可选的,机器可执行指令还促使处理器801执行以下步骤:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将实测信号质量与第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若第二预测误差值小于指定门限值,则将该第二波束预测模型作为目标波束预测模型,并结束遍历操作;
若第二预测误差值大于等于指定门限值,则继续遍历操作。
可选的,机器可执行指令还促使处理器801执行以下步骤:
若第一预测误差值和多个第二波束预测模型的第二预测误差值均大于等于指定门限值,则停用第一波束预测模型,并利用全量波束集进行波束预测。
可选的,机器可执行指令还促使处理器801执行以下步骤:
将实测信号质量和第一预测信号质量的差值作为第一预测误差值;或者,
计算实测信号质量和第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期第一波束预测模型输出的预测波束的信号质量之间的差值;
通过以下公式计算第一预测误差值:
第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
可选的,机器可执行指令还促使处理器801执行以下步骤:
接收基站发送的全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息;
接收基站发送的第一RRC消息,第一RRC消息包括第一测量波束集的标识、第一监测波束集的标识、指定门限值以及滤波因子,第一RRC消息用于指示终端利用第一监测波束集进行波束测量,利用第一测量波束集进行波束预测。
可选的,机器可执行指令还促使处理器801执行以下步骤:
向基站发送UCI,UCI包括多个波束的波束索引和信号质量,多个波束的信号质量均为信号质量最小值;
接收基站发送的第一指示消息,第一指示消息包括全量波束集的标识,第一指示消息用于指示终端利用全量波束集的配置信息进行波束测量和波束预测。
可选的,机器可执行指令还促使处理器801执行以下步骤:
向基站发送第二RRC消息,第二RRC消息包括目标波束预测模型对应的第三测量波束集的标识以及第三监测波束集的标识;
接收基站发送的第二指示消息,第二指示消息用于指示终端利用第三监测波束集进行波束测量,并利用第三测量波束集进行波束预测。
在图8中,还可以包括通信总线803。处理器801、机器可读存储介质802及收发器804之间通过通信总线803完成相互间的通信,通信总线803可以是外设部件互连标准(Peripheral Component Interconnect,PCI)总线或扩展工业标准结构(Extended Industry Standard Architecture,EISA)总线等。该通信总线可以分为地址总线、数据总线、控制总线等。
收发器804可以为无线通信模块,收发器804在处理器801的控制下,与其他设备进行数据交互。
机器可读存储介质802可以包括随机存取存储器(Random Access Memory,RAM),也可以包括非易失性存储器(Non-Volatile Memory,NVM),例如至少一个磁盘存储器。另外,机器可读存储介 质802还可以是至少一个位于远离前述处理器的存储装置。
处理器801可以是通用处理器,包括中央处理器(Central Processing Unit,CPU)、网络处理器(Network Processor,NP)等;还可以是数字信号处理器(Digital Signal Processing,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现场可编程门阵列(Field-Programmable Gate Array,FPGA)或其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
本申请实施例还提供一种基站,基站包括第一波束预测模型和多个第二波束预测模型,基站的生效波束预测模型为第一波束预测模型,如图9所示,该基站包括:
处理器901;收发器904;
机器可读存储介质902,机器可读存储介质902存储有能够被处理器901执行的机器可执行指令;机器可执行指令促使处理器901执行以下步骤:
获取终端上报的第一监测波束集包括的各波束对应的信号质量,并从第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
从第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将第一测量波束集包括的各波束的信号质量输入第一波束预测模型,获得第一波束预测模型输出的预测波束的第一预测信号质量;
基于第一预测信号质量和实测信号质量计算第一波束预测模型的第一预测误差值;
若第一预测误差值大于等于指定门限值,则从多个第二波束预测模型中选择目标波束预测模型,将生效波束预测模型切换为目标波束预测模型,目标波束预测模型的预测误差值小于指定门限值。
可选的,机器可执行指令还促使处理器901执行以下步骤:
依次对每个第二波束预测模型进行以下遍历操作:
针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
将实测信号质量与第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
若第二预测误差值小于指定门限值,则将该第二波束预测模型作为目标波束预测模型,并结束遍历操作;
若第二预测误差值大于等于指定门限值,则继续遍历操作。
可选的,机器可执行指令还促使处理器901执行以下步骤:
若第一预测误差值和多个第二波束预测模型的预测误差值均大于等于指定门限值,则停用第一波束预测模型,利用全量波束集进行波束预测。
可选的,机器可执行指令还促使处理器901执行以下步骤:
将实测信号质量和第一预测信号质量的差值作为第一预测误差值;或者,
计算实测信号质量和第一预测信号质量的差值,得到当前预测误差值;
获取历史预测误差值,历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期第一波束预测模型输出的预测信号质量之间的差值;
通过以下公式计算第一预测误差值:
第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
其中,α为滤波因子。
在图9中,还可以包括通信总线903。处理器901、机器可读存储介质902及收发器904之间通过 通信总线903完成相互间的通信,通信总线903可以是外设部件互连标准(Peripheral Component Interconnect,PCI)总线或扩展工业标准结构(Extended Industry Standard Architecture,EISA)总线等。该通信总线可以分为地址总线、数据总线、控制总线等。
收发器904可以为无线通信模块,收发器904在处理器901的控制下,与其他设备进行数据交互。
机器可读存储介质902可以包括随机存取存储器(Random Access Memory,RAM),也可以包括非易失性存储器(Non-Volatile Memory,NVM),例如至少一个磁盘存储器。另外,机器可读存储介质902还可以是至少一个位于远离前述处理器的存储装置。
处理器901可以是通用处理器,包括中央处理器(Central Processing Unit,CPU)、网络处理器(Network Processor,NP)等;还可以是数字信号处理器(Digital Signal Processing,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现场可编程门阵列(Field-Programmable Gate Array,FPGA)或其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
在本申请提供的又一实施例中,还提供了一种计算机可读存储介质,该计算机可读存储介质内存储有计算机程序,所述计算机程序被处理器执行时实现上述任一波束预测模型切换方法的步骤。
在本申请提供的又一实施例中,还提供了一种包含指令的计算机程序产品,当其在计算机上运行时,使得计算机执行上述实施例中任一波束预测模型切换方法。
在上述实施例中,可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时,可以全部或部分地以计算机程序产品的形式实现。所述计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行所述计算机程序指令时,全部或部分地产生按照本申请实施例所述的流程或功能。所述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机指令可以存储在计算机可读存储介质中,或者从一个计算机可读存储介质向另一个计算机可读存储介质传输,例如,所述计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。所述计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。所述可用介质可以是磁性介质,(例如,软盘、硬盘、磁带)、光介质(例如,DVD)、或者半导体介质(例如固态硬盘Solid State Disk(SSD))等。
需要说明的是,在本文中,诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来,而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且,术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
本说明书中的各个实施例均采用相关的方式描述,各个实施例之间相同相似的部分互相参见即可,每个实施例重点说明的都是与其他实施例的不同之处。尤其,对于装置实施例而言,由于其基本相似于方法实施例,所以描述的比较简单,相关之处参见方法实施例的部分说明即可。
以上所述仅为本申请的较佳实施例,并不用以限制本申请,凡在本申请的精神和原则之内,所做的任何修改、等同替换、改进等,均应包含在本申请保护的范围之内。

Claims (35)

  1. 一种波束预测模型切换方法,其特征在于,应用于终端,所述终端包括第一波束预测模型和多个第二波束预测模型,所述终端的生效波束预测模型为所述第一波束预测模型,所述方法包括:
    对第一监测波束集包括的各波束的导频信号进行测量,得到所述第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
    从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
    基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
    若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
  2. 根据权利要求1所述的方法,其特征在于,所述从所述多个第二波束预测模型中选择目标波束预测模型,包括:
    依次对每个第二波束预测模型进行以下遍历操作:
    针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
    将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
    若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
    若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
  3. 根据权利要求1或2所述的方法,其特征在于,在所述基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值之后,所述方法还包括:
    若所述第一预测误差值和所述多个第二波束预测模型的第二预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,并利用全量波束集进行波束预测。
  4. 根据权利要求3所述的方法,其特征在于,所述基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值,包括:
    将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
    计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
    获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
    通过以下公式计算所述第一预测误差值:
    所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
    其中,α为滤波因子。
  5. 根据权利要求4所述的方法,其特征在于,在所述对第一监测波束集包括的各波束的导频信号进行测量之前,所述方法还包括:
    接收基站发送的全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集 的配置信息;
    接收所述基站发送的第一RRC消息,所述第一RRC消息包括所述第一测量波束集的标识、所述第一监测波束集的标识、所述指定门限值以及所述滤波因子,所述第一RRC消息用于指示所述终端利用所述第一监测波束集进行波束测量,利用所述第一测量波束集进行波束预测。
  6. 根据权利要求5所述的方法,其特征在于,在所述停用所述第一波束预测模型之后,所述方法还包括:
    向所述基站发送UCI,所述UCI包括多个波束的波束索引和信号质量,所述多个波束的信号质量均为信号质量最小值;
    接收所述基站发送的第一指示消息,所述第一指示消息包括所述全量波束集的标识,所述第一指示消息用于指示所述终端利用所述全量波束集的配置信息进行波束测量和波束预测。
  7. 根据权利要求5所述的方法,其特征在于,在所述从所述多个第二波束预测模型中选择目标波束预测模型之后,所述方法还包括:
    向所述基站发送第二RRC消息,所述第二RRC消息包括所述目标波束预测模型对应的第三测量波束集的标识以及第三监测波束集的标识;
    接收所述基站发送的第二指示消息,所述第二指示消息用于指示所述终端利用所述第三监测波束集进行波束测量,并利用所述第三测量波束集进行波束预测。
  8. 一种波束预测模型切换方法,其特征在于,应用于基站,所述基站包括第一波束预测模型和多个第二波束预测模型,所述基站的生效波束预测模型为所述第一波束预测模型,所述方法包括:
    获取终端上报的第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
    从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
    基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
    若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
  9. 根据权利要求8所述的方法,其特征在于,所述从所述多个第二波束预测模型中选择目标波束预测模型,包括:
    依次对每个第二波束预测模型进行以下遍历操作:
    针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
    将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
    若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
    若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
  10. 根据权利要求8或9所述的方法,其特征在于,在所述基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值之后,所述方法还包括:
    若所述第一预测误差值和所述多个第二波束预测模型的预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,利用全量波束集进行波束预测。
  11. 根据权利要求8所述的方法,其特征在于,所述基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值,包括:
    将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
    计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
    获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
    通过以下公式计算所述第一预测误差值:
    所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
    其中,α为滤波因子。
  12. 一种波束预测模型切换装置,其特征在于,应用于终端,所述终端包括第一波束预测模型和多个第二波束预测模型,所述终端的生效波束预测模型为所述第一波束预测模型,所述装置包括:
    测量模块,用于对第一监测波束集包括的各波束的导频信号进行测量,得到所述第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
    预测模块,用于从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
    计算模块,用于基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
    切换模块,用于若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
  13. 根据权利要求12所述的装置,其特征在于,所述切换模块,具体用于:
    依次对每个第二波束预测模型进行以下遍历操作:
    针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
    将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
    若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
    若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
  14. 根据权利要求12或13所述的装置,其特征在于,
    所述预测模块,还用于若所述第一预测误差值和所述多个第二波束预测模型的第二预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,并利用全量波束集进行波束预测。
  15. 根据权利要求14所述的装置,其特征在于,所述计算模块,具体用于:
    将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
    计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
    获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
    通过以下公式计算所述第一预测误差值:
    所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
    其中,α为滤波因子。
  16. 根据权利要求15所述的装置,其特征在于,所述装置还包括:
    接收模块,用于接收基站发送的全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息;以及接收所述基站发送的第一RRC消息,所述第一RRC消息包括所述第一测量波束集的标识、所述第一监测波束集的标识、所述指定门限值以及所述滤波因子,所述第一RRC消息用于指示所述终端利用所述第一监测波束集进行波束测量,利用所述第一测量波束集进行波束预测。
  17. 根据权利要求16所述的装置,其特征在于,所述装置还包括发送模块;
    所述发送模块,用于向所述基站发送UCI,所述UCI包括多个波束的波束索引和信号质量,所述多个波束的信号质量均为信号质量最小值;
    所述接收模块,还用于接收所述基站发送的第一指示消息,所述第一指示消息包括所述全量波束集的标识,所述第一指示消息用于指示所述终端利用所述全量波束集的配置信息进行波束测量和波束预测。
  18. 根据权利要求16所述的装置,其特征在于,
    所述发送模块,还用于向所述基站发送第二RRC消息,所述第二RRC消息包括所述目标波束预测模型对应的第三测量波束集的标识以及第三监测波束集的标识;
    所述接收模块,还用于接收所述基站发送的第二指示消息,所述第二指示消息用于指示所述终端利用所述第三监测波束集进行波束测量,并利用所述第三测量波束集进行波束预测。
  19. 一种波束预测模型切换装置,其特征在于,应用于基站,所述基站包括第一波束预测模型和多个第二波束预测模型,所述基站的生效波束预测模型为所述第一波束预测模型,所述装置包括:
    获取模块,用于获取终端上报的第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
    预测模块,用于从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
    计算模块,用于基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
    切换模块,用于若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
  20. 根据权利要求19所述的装置,其特征在于,所述切换模块,具体用于:
    依次对每个第二波束预测模型进行以下遍历操作:
    针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
    将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
    若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型, 并结束所述遍历操作;
    若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
  21. 根据权利要求19或20所述的装置,其特征在于,
    所述预测模块,还用于若所述第一预测误差值和所述多个第二波束预测模型的预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,利用全量波束集进行波束预测。
  22. 根据权利要求19所述的装置,其特征在于,所述计算模块,具体用于:
    将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
    计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
    获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
    通过以下公式计算所述第一预测误差值:
    所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
    其中,α为滤波因子。
  23. 一种终端,其特征在于,所述终端包括第一波束预测模型和多个第二波束预测模型,所述终端的生效波束预测模型为所述第一波束预测模型,所述终端包括:
    处理器;
    收发器;
    机器可读存储介质,所述机器可读存储介质存储有能够被所述处理器执行的机器可执行指令;所述机器可执行指令促使所述处理器执行以下步骤:
    对第一监测波束集包括的各波束的导频信号进行测量,得到所述第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
    从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
    基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
    若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
  24. 根据权利要求23所述的终端,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:
    依次对每个第二波束预测模型进行以下遍历操作:
    针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
    将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
    若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
    若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
  25. 根据权利要求23或24所述的终端,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:
    若所述第一预测误差值和所述多个第二波束预测模型的第二预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,并利用全量波束集进行波束预测。
  26. 根据权利要求25所述的终端,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:
    将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
    计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
    获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
    通过以下公式计算所述第一预测误差值:
    所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
    其中,α为滤波因子。
  27. 根据权利要求26所述的终端,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:
    接收基站发送的全量波束集的配置信息、至少一个测量波束集的配置信息以及至少一个监测波束集的配置信息;
    接收所述基站发送的第一RRC消息,所述第一RRC消息包括所述第一测量波束集的标识、所述第一监测波束集的标识、所述指定门限值以及所述滤波因子,所述第一RRC消息用于指示所述终端利用所述第一监测波束集进行波束测量,利用所述第一测量波束集进行波束预测。
  28. 根据权利要求27所述的终端,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:
    向所述基站发送UCI,所述UCI包括多个波束的波束索引和信号质量,所述多个波束的信号质量均为信号质量最小值;
    接收所述基站发送的第一指示消息,所述第一指示消息包括所述全量波束集的标识,所述第一指示消息用于指示所述终端利用所述全量波束集的配置信息进行波束测量和波束预测。
  29. 根据权利要求27所述的终端,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:
    向所述基站发送第二RRC消息,所述第二RRC消息包括所述目标波束预测模型对应的第三测量波束集的标识以及第三监测波束集的标识;
    接收所述基站发送的第二指示消息,所述第二指示消息用于指示所述终端利用所述第三监测波束集进行波束测量,并利用所述第三测量波束集进行波束预测。
  30. 一种基站,其特征在于,所述基站包括第一波束预测模型和多个第二波束预测模型,所述基站的生效波束预测模型为所述第一波束预测模型,所述基站包括:
    处理器;
    收发器;
    机器可读存储介质,所述机器可读存储介质存储有能够被所述处理器执行的机器可执行指令;所述机器可执行指令促使所述处理器执行以下步骤:
    获取终端上报的第一监测波束集包括的各波束对应的信号质量,并从所述第一监测波束集包括的各波束对应的信号质量中选择数值最大的信号质量,作为实测信号质量;
    从所述第一监测波束集包括的各波束对应的信号质量中获取第一测量波束集包括的各波束的信号质量,并将所述第一测量波束集包括的各波束的信号质量输入所述第一波束预测模型,获得所述第一波束预测模型输出的预测波束的第一预测信号质量;
    基于所述第一预测信号质量和所述实测信号质量计算所述第一波束预测模型的第一预测误差值;
    若所述第一预测误差值大于等于指定门限值,则从所述多个第二波束预测模型中选择目标波束预测模型,将所述生效波束预测模型切换为所述目标波束预测模型,所述目标波束预测模型的预测误差值小于所述指定门限值。
  31. 根据权利要求30所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:
    依次对每个第二波束预测模型进行以下遍历操作:
    针对该第二波束预测模型,将该第二波束预测模型对应的测量波束集包括的各波束的信号质量输入该第二波束预测模型,获得该第二波束预测模型输出的预测波束的第二预测信号质量;
    将所述实测信号质量与所述第二预测信号质量的差值作为该第二波束预测模型的第二预测误差值;
    若所述第二预测误差值小于所述指定门限值,则将该第二波束预测模型作为所述目标波束预测模型,并结束所述遍历操作;
    若所述第二预测误差值大于等于所述指定门限值,则继续所述遍历操作。
  32. 根据权利要求30或31所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:
    若所述第一预测误差值和所述多个第二波束预测模型的预测误差值均大于等于所述指定门限值,则停用所述第一波束预测模型,利用全量波束集进行波束预测。
  33. 根据权利要求30所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:
    将所述实测信号质量和所述第一预测信号质量的差值作为所述第一预测误差值;或者,
    计算所述实测信号质量和所述第一预测信号质量的差值,得到当前预测误差值;
    获取历史预测误差值,所述历史预测误差值为上一测量周期测量得到的实测信号质量和上一测量周期所述第一波束预测模型输出的预测信号质量之间的差值;
    通过以下公式计算所述第一预测误差值:
    所述第一预测误差值=|α*当前预测误差值+(1-α)*历史预测误差值|;
    其中,α为滤波因子。
  34. 一种机器可读存储介质,其特征在于,存储有机器可执行指令,在被处理器调用和执行时,所述机器可执行指令促使所述处理器:实现权利要求1-11任一所述的方法。
  35. 一种计算机程序产品,其特征在于,所述计算机程序产品促使所述处理器:实现权利要求1-11任一所述的方法。
PCT/CN2023/121924 2023-09-27 2023-09-27 波束预测模型切换方法、装置、终端、基站及存储介质 Pending WO2025065321A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2023/121924 WO2025065321A1 (zh) 2023-09-27 2023-09-27 波束预测模型切换方法、装置、终端、基站及存储介质
CN202380010925.1A CN120113284A (zh) 2023-09-27 2023-09-27 波束预测模型切换方法、装置、终端、基站及存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2023/121924 WO2025065321A1 (zh) 2023-09-27 2023-09-27 波束预测模型切换方法、装置、终端、基站及存储介质

Publications (1)

Publication Number Publication Date
WO2025065321A1 true WO2025065321A1 (zh) 2025-04-03

Family

ID=95204039

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/121924 Pending WO2025065321A1 (zh) 2023-09-27 2023-09-27 波束预测模型切换方法、装置、终端、基站及存储介质

Country Status (2)

Country Link
CN (1) CN120113284A (zh)
WO (1) WO2025065321A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111565062A (zh) * 2020-04-15 2020-08-21 中国联合网络通信集团有限公司 一种波束切换方法和装置
CN111865446A (zh) * 2020-07-29 2020-10-30 中南大学 利用网络环境上下文信息实现的智能波束配准方法与装置
CN113438002A (zh) * 2021-06-07 2021-09-24 北京邮电大学 基于lstm的模拟波束切换方法、装置、设备及介质
CN115398820A (zh) * 2020-04-16 2022-11-25 高通股份有限公司 波束成形的通信中的机器学习模型选择
US20230261728A1 (en) * 2022-02-15 2023-08-17 Qualcomm Incorporated Enhanced beam failure detection

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111565062A (zh) * 2020-04-15 2020-08-21 中国联合网络通信集团有限公司 一种波束切换方法和装置
CN115398820A (zh) * 2020-04-16 2022-11-25 高通股份有限公司 波束成形的通信中的机器学习模型选择
CN111865446A (zh) * 2020-07-29 2020-10-30 中南大学 利用网络环境上下文信息实现的智能波束配准方法与装置
CN113438002A (zh) * 2021-06-07 2021-09-24 北京邮电大学 基于lstm的模拟波束切换方法、装置、设备及介质
US20230261728A1 (en) * 2022-02-15 2023-08-17 Qualcomm Incorporated Enhanced beam failure detection

Also Published As

Publication number Publication date
CN120113284A (zh) 2025-06-06

Similar Documents

Publication Publication Date Title
CN107333293B (zh) 测量信息上报方法及装置
CN109089327B (zh) 无线网络连接方法、装置、终端及存储介质
CN113438702B (zh) 小区切换方法、装置、电子设备及可读存储介质
WO2019056906A1 (zh) 小区测量方法、装置、终端及存储介质
WO2019140646A1 (zh) 切换处理方法、网络设备、ue及计算机存储介质
KR20200004338A (ko) 업링크 신호의 전송 파라미터를 결정하는 방법, 단말기 및 네크워크 장치
CN107295558A (zh) 小区测量方法及装置
CN109413701B (zh) 一种VoLTE业务切换方法及系统
WO2024094157A1 (zh) Ai模型更新方法及通信装置
CN110572855B (zh) 一种Volte语音业务的异系统切换方法及装置
US20180288655A1 (en) Path processing method and apparatus, and terminal
US9974063B2 (en) Radio network control method and radio network controller
WO2024094103A1 (zh) Ai模型更新方法及通信装置
TW202121874A (zh) 網路路徑選擇方法及應用其之網路節點裝置
WO2025065321A1 (zh) 波束预测模型切换方法、装置、终端、基站及存储介质
WO2024066620A1 (zh) 波束指示方法、装置及存储介质
CN109561478B (zh) 用户终端盲重定向方法、装置、计算机存储介质及设备
WO2025176099A1 (zh) 波束管理方法、装置及存储介质
WO2025213974A1 (zh) 基于事件触发的层一测量上报方法、设备、装置及存储介质
CN115297514B (zh) 一种连接切换管理系统及方法
CN110636563B (zh) 通信方法和装置
CN114710814B (zh) 终端设备的控制方法、装置、系统、存储介质及电子设备
WO2024067795A1 (zh) 一种配置方法、终端、网络设备、电子设备及介质
CN118158752A (zh) 信号质量测量方法、装置、电子设备、存储介质及产品
WO2024032764A1 (zh) 信息处理方法、装置、节点设备及介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23953478

Country of ref document: EP

Kind code of ref document: A1

WWP Wipo information: published in national office

Ref document number: 202380010925.1

Country of ref document: CN