Beam prediction model training method and apparatus, and beam prediction method and apparatus
By sending pilot signals in full in mobile communication scenarios and sampling and processing measurement reports, training deep learning models, solving the problems of poor generalization performance and high computing power demand in the prior art, and achieving efficient beam prediction.
Patent Information
- Application Number
- PCT/CN2023/092867
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-09
- Publication Date
- 2024-11-14
AI Technical Summary
In mobile communication scenarios, it is difficult for the prior art to train a beam prediction model with relatively optimized performance under smaller computing power requirements, and traditional methods require a lot of pilot overhead and complex judgment logic.
By sending pilot signals in full between the base station and the terminal, receiving measurement report messages, filling in one-dimensional arrays as label data, sampling according to the preset sampling mode of different scenarios, and using multiple sets of training data to train the deep learning model, we obtain Beam prediction model.
It reduces the computing power requirements of base stations and terminals, reduces pilot overhead and data acquisition difficulty, and improves the generalization performance of the beam prediction model.
Smart Images

Figure CN2023092867_14112024_PF_FP_ABST
Abstract
Description
Beam prediction model training method, beam prediction method and device Technical Field
[0001] The present application relates to the field of wireless communication technology, and in particular to a beam prediction model training method, a beam prediction method, and a device. Background Art
[0002] In mobile communications, factors such as user movement and random signal obstruction can cause time-varying channels. Therefore, repeated beam measurement and tracking are required to accurately determine the channel information between base stations and terminals in real time. Traditional beam prediction involves the base station sending pilot signals in all transmit beam directions to predict the optimal transmit beam direction. However, this approach incurs significant pilot overhead.
[0003] In order to reduce pilot overhead, a beam prediction algorithm based on machine learning has been developed. The beam prediction model can be trained using machine learning, and then the beam prediction model can be used to predict the beam direction of the optimal transmission beam based on the measurement results of fewer beams.
[0004] Currently, multiple beam prediction models can be trained, each applicable to a specific scenario. Base stations can select a beam prediction model based on the actual scenario when performing beam prediction. Due to the complexity of wireless environments and the large number of scenarios involved, a large number of beam prediction models need to be trained. The generalization performance of a single beam training model is poor, and the base station's decision logic for selecting a beam training model is complex.
[0005] Alternatively, you can use training data from multiple scenarios to train a beam prediction model applicable to multiple scenarios. This approach can yield good generalization performance, but it requires a large amount of training data. Beam measurements must be performed for each scenario to collect training data, which in turn places high demands on the computing power of the equipment used to train and apply the beam prediction model.
[0006] Therefore, how to train a beam prediction model with better generalization performance while requiring less computing power is an urgent problem that needs to be solved.
[0007] Summary of the Invention
[0008] The purpose of the embodiments of the present application is to provide a beam prediction model training method, a beam prediction method, and an apparatus to obtain a beam prediction model with excellent generalization performance under low computing power requirements. The specific technical solutions are as follows:
[0009] In a first aspect, an embodiment of the present application provides a beam prediction model training method, which is applied to a base station and includes:
[0010] Using the full transmit beam to send a pilot signal to the terminal;
[0011] receiving a measurement report message sent by the terminal, where the measurement report message includes measurement results of a first preset number of beam pairs, where the first preset number is the product of the number of transmit beams of the base station and the number of receive beams of the terminal;
[0012] Filling the measurement results of the first preset number of beam pairs into elements included in a one-dimensional array, and using the one-dimensional array as label data, where the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, where the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal;
[0013] The elements are sampled according to the preset sampling modes corresponding to different scenarios, and invalid values are filled in the unsampled elements to obtain training data corresponding to each preset sampling mode;
[0014] A deep learning model is trained using multiple sets of training data and the label data to obtain a beam prediction model.
[0015] In a possible implementation manner, the receiving a measurement report message sent by the terminal includes:
[0016] receiving a first measurement report message sent by the terminal through physical layer signaling, where the first measurement report message includes a valid value bitmap field and measurement results of the first preset number of beam pairs, where the valid value bitmap field is used to indicate a position of the measurement result in the first measurement report message; or
[0017] Receive a second measurement report message sent by the terminal through MAC layer signaling, where the second measurement report message includes the number of measurement results and the measurement results of the first preset number of beam pairs.
[0018] In one possible implementation, the measurement result of each beam pair includes the beam ID of the receive beam, the pilot index of the transmit beam, and the pilot signal quality of the beam pair;
[0019] Filling the measurement results of the first preset number of beam pairs into the elements of the one-dimensional array includes:
[0020] For each beam pair measurement result, convert the pilot index of the transmit beam included in the beam pair measurement result into a beam ID;
[0021] Calculate a storage sequence number for the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receive beam in the beam pair, Tj is the beam ID of the transmit beam in the beam pair, and R_max is the maximum number of receive beams;
[0022] The pilot signal quality included in the measurement result of the beam pair is filled into the element indicated by the storage sequence number.
[0023] In a second aspect, an embodiment of the present application provides a beam prediction method, which is applied to a base station and includes:
[0024] Use the transmit beam corresponding to the current scenario to send a pilot signal to the terminal;
[0025] receiving a measurement report message sent by the terminal after receiving the pilot signal on a receive beam, where the transmit beam and the receive beam form a beam pair, and the measurement report message includes a measurement result of the beam pair;
[0026] Filling the measurement results of the beam pairs into the elements included in the one-dimensional array, and filling unoccupied elements with invalid values; the number of the elements is the product of the maximum number of transmit beams and the maximum number of receive beams, the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal;
[0027] The elements are input into a beam prediction model to obtain an output array of the beam prediction model, where the elements of the output array include prediction results of the beam pairs.
[0028] In a possible implementation manner, the receiving a measurement report message sent by the terminal after receiving the pilot signal on a receive beam includes:
[0029] receiving a third measurement report message sent by the terminal through physical layer signaling, where the third measurement report message includes a valid value bitmap field and a measurement result of the beam pair, where the valid value bitmap field is used to indicate a position of the measurement result in the third measurement report message; or
[0030] Receive a fourth measurement report message sent by the terminal through MAC layer signaling, where the fourth measurement report message includes the number of measurement results and the measurement results of the beam pair.
[0031] In a possible implementation, the measurement result of each beam pair includes a beam ID of a receiving beam, a pilot index of a transmitting beam, and a pilot signal quality of the beam pair;
[0032] Filling the measurement results of the beam pair into the elements of the one-dimensional array includes:
[0033] For each beam pair, convert the pilot index of the transmit beam included in the measurement result of the beam pair into a beam ID;
[0034] Calculate a storage sequence number for the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receive beam in the beam pair, Tj is the beam ID of the transmit beam in the beam pair, and R_max is the maximum number of receive beams;
[0035] The measurement result of the beam pair is filled into the element indicated by the storage sequence number.
[0036] In a possible implementation, after obtaining the output array of the beam prediction model, the method further includes:
[0037] Obtaining a first number of elements including optimal transmit beams from the elements of the output array;
[0038] selecting a second number of receive beams from the prediction results included in the first number of elements;
[0039] A beam prediction message is sent to the terminal through physical layer signaling or MAC layer signaling, where the beam prediction message includes the second number, beam IDs of the second number of receiving beams, and priorities of the second number of receiving beams.
[0040] In a possible implementation, before using the transmit beam corresponding to the current scenario to send a pilot signal to the terminal, the method further includes:
[0041] Sending a first RRC message to the terminal, where the CSI measurement configuration field of the first RRC message includes a newly added field, and a value of the newly added field is used to indicate the number of receive beams used by the terminal when performing beam measurement;
[0042] Receive a second RRC message sent by the terminal, where the second RRC message is used to indicate that the terminal supports beam measurement according to the number of receiving beams included in the newly added field.
[0043] In a third aspect, an embodiment of the present application provides a beam prediction model training device, which is applied to a base station and includes:
[0044] A sending module, configured to send a pilot signal to a terminal using a full transmit beam;
[0045] a receiving module, configured to receive a measurement report message sent by the terminal, where the measurement report message includes measurement results of a first preset number of beam pairs, where the first preset number is the product of the number of transmit beams of the base station and the number of receive beams of the terminal;
[0046] a filling module, configured to fill the measurement results of the first preset number of beam pairs into elements included in a one-dimensional array, and use the one-dimensional array as label data, where the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, where the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal;
[0047] A sampling module is used to sample the elements according to the preset sampling modes corresponding to different scenarios, fill the unsampled elements with invalid values, and obtain training data corresponding to each preset sampling mode;
[0048] The training module is used to train a deep learning model using multiple sets of training data and the label data to obtain a beam prediction model.
[0049] In a possible implementation, the receiving module is specifically configured to:
[0050] receiving a first measurement report message sent by the terminal through physical layer signaling, where the first measurement report message includes a valid value bitmap field and measurement results of the first preset number of beam pairs, where the valid value bitmap field is used to indicate a position of the measurement result in the first measurement report message; or
[0051] Receive a second measurement report message sent by the terminal through MAC layer signaling, where the second measurement report message includes the number of measurement results and the measurement results of the first preset number of beam pairs.
[0052] In one possible implementation, the measurement result of each beam pair includes the beam ID of the receive beam, the pilot index of the transmit beam, and the pilot signal quality of the beam pair;
[0053] The filling module is specifically used for:
[0054] For each beam pair measurement result, convert the pilot index of the transmit beam included in the beam pair measurement result into a beam ID;
[0055] Calculate a storage sequence number for the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receive beam in the beam pair, Tj is the beam ID of the transmit beam in the beam pair, and R_max is the maximum number of receive beams;
[0056] The pilot signal quality included in the measurement result of the beam pair is filled into the element indicated by the storage sequence number.
[0057] In a fourth aspect, an embodiment of the present application provides a beam prediction device, which is applied to a base station and includes:
[0058] A sending module, configured to send a pilot signal to the terminal using a transmit beam corresponding to the current scenario;
[0059] a receiving module, configured to receive a measurement report message sent by the terminal after receiving the pilot signal on a receive beam, where the transmit beam and the receive beam form a beam pair, and the measurement report message includes a measurement result of the beam pair;
[0060] a filling module, configured to fill the measurement results of the beam pair into the elements included in the one-dimensional array, and fill unoccupied elements with invalid values; the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, the maximum number of transmit beams being the maximum number of transmit beams of the base station, and the maximum number of receive beams being the maximum number of receive beams of the terminal;
[0061] An acquisition module is used to input the elements into a beam prediction model to obtain an output array of the beam prediction model, where the elements of the output array include prediction results of the beam pairs.
[0062] In a possible implementation, the receiving module is specifically configured to:
[0063] receiving a third measurement report message sent by the terminal through physical layer signaling, where the third measurement report message includes a valid value bitmap field and a measurement result of the beam pair, where the valid value bitmap field is used to indicate a position of the measurement result in the third measurement report message; or
[0064] Receive a fourth measurement report message sent by the terminal through MAC layer signaling, where the fourth measurement report message includes the number of measurement results and the measurement results of the beam pair.
[0065] In a possible implementation, the measurement result of each beam pair includes a beam ID of a receiving beam, a pilot index of a transmitting beam, and a pilot signal quality of the beam pair;
[0066] The filling module is specifically used for:
[0067] For each beam pair, convert the pilot index of the transmit beam included in the measurement result of the beam pair into a beam ID;
[0068] Calculate a storage sequence number for the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receive beam in the beam pair, Tj is the beam ID of the transmit beam in the beam pair, and R_max is the maximum number of receive beams;
[0069] The measurement result of the beam pair is filled into the element indicated by the storage sequence number.
[0070] In a possible implementation, the apparatus further includes a selection module:
[0071] The acquisition module is further configured to acquire a first number of elements including the optimal transmit beam from the elements of the output array;
[0072] The selection module is configured to select a second number of receive beams from the prediction results included in the first number of elements;
[0073] The sending module is also used to send a beam prediction message to the terminal through physical layer signaling or MAC layer signaling, and the beam prediction message includes the second number, the beam ID of the second number of receiving beams, and the priority of the second number of receiving beams.
[0074] In one possible implementation,
[0075] The sending module is further configured to, before sending a pilot signal to the terminal using the transmit beam corresponding to the current scenario, send a first RRC message to the terminal, where the CSI measurement configuration field of the first RRC message includes a newly added field, and a value of the newly added field is used to indicate the number of receive beams used by the terminal when performing beam measurement;
[0076] The receiving module is further used to receive a second RRC message sent by the terminal, where the second RRC message is used to indicate that the terminal supports beam measurement according to the number of receiving beams included in the newly added field.
[0077] In a fifth aspect, an embodiment of the present application provides a base station, including:
[0078] processor;
[0079] transceiver;
[0080] A machine-readable storage medium storing machine-executable instructions capable of being executed by the processor; the machine-executable instructions prompting the processor to perform the following steps:
[0081] Sending a pilot signal to the terminal by using a full transmit beam through the transceiver;
[0082] Receiving, by the transceiver, a measurement report message sent by the terminal, where the measurement report message includes measurement results of a first preset number of beam pairs, where the first preset number is a product of the number of transmit beams of the base station and the number of receive beams of the terminal;
[0083] Filling the measurement results of the first preset number of beam pairs into elements included in a one-dimensional array, and using the one-dimensional array as label data, where the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, where the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal;
[0084] The elements are sampled according to the preset sampling modes corresponding to different scenarios, and invalid values are filled in the unsampled elements to obtain training data corresponding to each preset sampling mode;
[0085] A deep learning model is trained using multiple sets of training data and the label data to obtain a beam prediction model.
[0086] In one possible implementation, the machine executable instructions further cause the processor to execute the following steps:
[0087] receiving a first measurement report message sent by the terminal through physical layer signaling, where the first measurement report message includes a valid value bitmap field and measurement results of the first preset number of beam pairs, where the valid value bitmap field is used to indicate a position of the measurement result in the first measurement report message; or
[0088] Receive a second measurement report message sent by the terminal through MAC layer signaling, where the second measurement report message includes the number of measurement results and the measurement results of the first preset number of beam pairs.
[0089] In one possible implementation, the measurement result of each beam pair includes the beam ID of the receive beam, the pilot index of the transmit beam, and the pilot signal quality of the beam pair;
[0090] The machine executable instructions further cause the processor to perform the following steps:
[0091] For each beam pair measurement result, convert the pilot index of the transmit beam included in the beam pair measurement result into a beam ID;
[0092] Calculate a storage sequence number for the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receive beam in the beam pair, Tj is the beam ID of the transmit beam in the beam pair, and R_max is the maximum number of receive beams;
[0093] The pilot signal quality included in the measurement result of the beam pair is filled into the element indicated by the storage sequence number.
[0094] In a sixth aspect, an embodiment of the present application provides a base station, including:
[0095] processor;
[0096] transceiver;
[0097] A machine-readable storage medium storing machine-executable instructions capable of being executed by the processor; the machine-executable instructions prompting the processor to perform the following steps:
[0098] Sending a pilot signal to the terminal via the transceiver using a transmit beam corresponding to the current scenario;
[0099] receiving, by the transceiver, a measurement report message sent by the terminal after receiving the pilot signal on a receive beam, the transmit beam and the receive beam forming a beam pair, the measurement report message including a measurement result of the beam pair;
[0100] Filling the measurement results of the beam pairs into the elements included in the one-dimensional array, and filling unoccupied elements with invalid values; the number of the elements is the product of the maximum number of transmit beams and the maximum number of receive beams, the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal;
[0101] The elements are input into a beam prediction model to obtain an output array of the beam prediction model, where the elements of the output array include prediction results of the beam pairs.
[0102] In one possible implementation, the machine executable instructions further cause the processor to execute the following steps:
[0103] receiving, by the transceiver, a third measurement report message sent by the terminal through physical layer signaling, where the third measurement report message includes a valid value bitmap field and a measurement result of the beam pair, where the valid value bitmap field is used to indicate a position of the measurement result in the third measurement report message; or
[0104] A fourth measurement report message sent by the terminal through MAC layer signaling is received through the transceiver, where the fourth measurement report message includes the number of measurement results and the measurement results of the beam pair.
[0105] In a possible implementation, the measurement result of each beam pair includes a beam ID of a receiving beam, a pilot index of a transmitting beam, and a pilot signal quality of the beam pair;
[0106] The machine executable instructions further cause the processor to perform the following steps:
[0107] For each beam pair, convert the pilot index of the transmit beam included in the measurement result of the beam pair into a beam ID;
[0108] Calculate a storage sequence number for the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receive beam in the beam pair, Tj is the beam ID of the transmit beam in the beam pair, and R_max is the maximum number of receive beams;
[0109] The measurement result of the beam pair is filled into the element indicated by the storage sequence number.
[0110] In one possible implementation, the machine executable instructions further cause the processor to execute the following steps:
[0111] Obtaining a first number of elements including optimal transmit beams from the elements of the output array;
[0112] selecting a second number of receive beams from the prediction results included in the first number of elements;
[0113] The transceiver is used to send a beam prediction message to the terminal through physical layer signaling or MAC layer signaling, where the beam prediction message includes the second number, beam IDs of the second number of receiving beams, and priorities of the second number of receiving beams.
[0114] In one possible implementation, the machine executable instructions further cause the processor to execute the following steps:
[0115] Sending a first RRC message to the terminal through the transceiver, where the CSI measurement configuration field of the first RRC message includes a newly added field, and a value of the newly added field is used to indicate the number of receive beams used by the terminal when performing beam measurement;
[0116] A second RRC message sent by the terminal is received through the transceiver, where the second RRC message is used to indicate that the terminal supports beam measurement according to the number of receiving beams included in the newly added field.
[0117] In a seventh aspect, an embodiment of the present application provides a machine-readable storage medium storing machine-executable instructions. When called and executed by a processor, the machine-executable instructions prompt the processor to implement the method described in the first aspect or the second aspect.
[0118] In an eighth 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.
[0119] Using the above method, the base station can use the full transmit beam to send a pilot signal to the terminal and receive a measurement report message sent by the terminal. The measurement report message includes the measurement results of the first preset number of beam pairs, that is, the terminal has measured the full number of beam pairs between the base station and the terminal. The base station can fill the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array. Since the number of elements included in the one-dimensional array is the product of the maximum number of transmit beams and the maximum number of receive beams, the one-dimensional array is suitable for carrying measurement results in various scenarios. This is equivalent to unifying the format of model input data in different scenarios in the embodiment of the present application. The model does not need to distinguish between the data of each scenario, which can reduce the complexity of the model. The base station can also sample the elements according to the preset sampling mode corresponding to different scenarios. A set of training data can be obtained based on each preset sampling mode. In this way, multiple sets of training data can be obtained based on the measurement results reported by the terminal at one time, and there is no need to perform beam measurement for each scenario separately. This reduces the pilot overhead between the base station and the terminal and the difficulty of obtaining training data for each scenario, and reduces the computing power requirements of the base station and the terminal. Moreover, because the preset sampling patterns corresponding to different scenarios are used when sampling the one-dimensional array, it is equivalent to simulating multiple scenarios, which can make the beam prediction model trained using these training data have better generalization performance. BRIEF DESCRIPTION OF THE DRAWINGS
[0120] The drawings described herein are used to provide further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute improper limitations on the present application.
[0121] FIG1 is an exemplary schematic diagram of an AI / ML-based beam prediction structure provided in an embodiment of the present application;
[0122] FIG2 is a flow chart of a beam prediction model training method provided in an embodiment of the present application;
[0123] FIG3 is an exemplary schematic diagram of a one-dimensional array provided in an embodiment of the present application;
[0124] FIG4 is an exemplary schematic diagram of a beam pair measurement result format provided in an embodiment of the present application;
[0125] FIG5 is a flow chart of another beam prediction model training method provided in an embodiment of the present application;
[0126] FIG6 is an exemplary schematic diagram of a preset sampling mode provided in an embodiment of the present application;
[0127] FIG7 is a flow chart of a beam prediction method provided in an embodiment of the present application;
[0128] FIG8 is an exemplary schematic diagram of a radio resource control protocol message format provided in an embodiment of the present application;
[0129] FIG9 is an interactive flow chart of a beam prediction method provided in an embodiment of the present application;
[0130] FIG10 is a schematic structural diagram of a beam prediction model training device provided in an embodiment of the present application;
[0131] FIG11 is a schematic structural diagram of a beam prediction device provided in an embodiment of the present application;
[0132] FIG12 is a schematic structural diagram of a base station provided in an embodiment of the present application;
[0133] FIG13 is a schematic diagram of the structure of another base station provided in an embodiment of the present application. DETAILED DESCRIPTION
[0134] To make the objectives, technical solutions, and advantages of the present invention more clearly understood, the present invention is further described below with reference to the accompanying drawings and examples. It should be understood that the described examples are only some examples of the present invention, not all examples. All other examples derived by persons of ordinary skill in the art based on the examples of the present invention fall within the scope of protection of the present invention.
[0135] In mobile communication scenarios, machine learning can be used to train a beam prediction model, which can then be used to predict the beam direction of the optimal beam based on the measurement results of fewer beams.
[0136] As shown in Figure 1, Figure 1 is an exemplary schematic diagram of a beam prediction structure based on artificial intelligence (AI) / machine learning (ML) provided in an embodiment of the present application, including a data acquisition module 101, a model training module 102, a model reasoning module 103 and an execution module 104.
[0137] The data collection module 101 may be deployed in a base station, or part of its functions may be deployed in the base station and part of its functions may be deployed in other network devices.
[0138] The model training module 102 may be deployed in a base station or in other network devices.
[0139] The model inference module 103 and the execution module 104 are both deployed in the base station.
[0140] Data acquisition module 101 is used to collect training data, label data corresponding to the training data, and inference data, and to identify, clean, and normalize the training and inference data to ensure that the format and quality of the processed training and inference data meet the requirements of model training. Both the training and inference data can be the Layer 1 Reference Signal Received Power (L1-RSRP) of each beam. L1-RSRP can be used to represent the strength of the physical layer pilot signal.
[0141] The training data and the label data corresponding to the training data are used to train the beam prediction model. The inference data is the data that needs to be input into the beam prediction model after the trained beam prediction model is deployed.
[0142] During the model training phase, the data acquisition module 101 may input the training data and the label data corresponding to the training data into the model training module 102. The model training module 102 is used to perform model training based on the training data sent by the data acquisition module 101. Specifically, a machine learning algorithm such as a deep neural network (DNN) or a long short-term memory network (LSTM) may be used for training to obtain a beam prediction model. For example, the model training module 102 may input the training data into the model, obtain the output of the model, calculate the loss function value based on the output of the model and the label data corresponding to the training data, and then adjust the model parameters based on the loss function value until the model converges, at which point the trained model is used as the beam prediction model.
[0143] Furthermore, the model training module 102 can perform model deployment in the model reasoning module 103 , that is, deploy the trained beam prediction model in the model reasoning module 103 .
[0144] During the beam prediction phase, the data acquisition module 101 may input inference data into the model inference module 103. The model inference module 103 may process the inference data using the beam prediction model to obtain an inference result output by the beam prediction model and send the inference result to the execution module 104. The inference result may be the beam predicted by the beam prediction model.
[0145] The execution module 104 is a module in the base station used for communicating with the terminal. The execution module 104 is used to send signals to the terminal using a transmit beam and receive signals sent by the terminal using a receive beam.
[0146] For example, the execution module 104 may send a signal to the terminal according to the transmit beam predicted by the beam prediction model.
[0147] Execution module 104 can also use some transmit beams to transmit pilot signals to the terminal and receive the L1-RSRP corresponding to each transmit beam fed back by the terminal. Execution module 104 can then provide data feedback to data acquisition module 101, specifically, feeding back the L1-RSRP corresponding to each transmit beam fed back by the terminal to data acquisition module 101. Data acquisition module 101 can use the obtained L1-RSRP corresponding to each transmit beam as inference data.
[0148] The model reasoning module 103 can also provide model performance feedback to the model training module 102. Specifically, it can provide feedback on the accuracy of the beam predicted by the beam prediction model. If the accuracy is less than the preset accuracy threshold, the model training module 102 can train the beam prediction model again to optimize the parameters of the beam prediction model, and update the beam prediction model in the model reasoning module 103 based on the optimized parameters.
[0149] Among them, the scenario can be defined by the following three factors:
[0150] Factor 1: Base station antenna type and beam type;
[0151] Factor 2: Terminal antenna type and beam type;
[0152] Factor 3: Environmental information of the wireless channel.
[0153] Specifically, the environmental information of the wireless channel may include the area (eg, urban area and suburban area), cell radius, operating frequency of the wireless channel, base station height, terminal moving speed, and the like.
[0154] It can be seen that a variety of scenarios can be defined through the above three factors.
[0155] Due to the diversity of scenarios, how to train a beam prediction model with better generalization performance while requiring less computing power is an urgent problem that needs to be solved.
[0156] To solve the above problems, the present application provides a beam prediction model training method, which is applied to a base station. As shown in FIG2 , the method includes:
[0157] S201: Send a pilot signal to the terminal using a full transmit beam.
[0158] S202: Receive a measurement report message sent by the terminal.
[0159] The measurement report message includes measurement results of a first preset number of beam pairs, where the first preset number is the product of the number of transmitting beams of the base station and the number of receiving beams of the terminal.
[0160] A terminal's receive beam and a base station's transmit beam form a beam pair. The beam pair measurement result refers to the terminal's measurement of the pilot signal sent by the base station in the direction of the transmit beam, using the receive beam. The measurement result can be the signal quality of the pilot signal, which can be represented by the Reference Signal Received Power (RSRP).
[0161] For example, if the base station has T transmit beams and the terminal has R receive beams, the first preset number is T*R. Assuming the terminal uses receive beam Ra to receive a pilot signal sent by the base station using beam Tb, Ra and Tb form a beam pair. The measurement result of this beam pair is the measurement result of the pilot signal after the terminal uses receive beam Ra to receive the pilot signal sent by the base station using transmit beam Tb.
[0162] S203: Fill the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array, and use the one-dimensional array as label data.
[0163] The number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams. The maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal.
[0164] The number of transmit beams of a base station may vary in different scenarios, and the maximum transmit beam is the maximum number of transmit beams of the base station in different scenarios. Similarly, the number of receive beams of a terminal in different scenarios may also vary, and the maximum receive beam is the maximum number of transmit beams of the terminal in different scenarios.
[0165] As shown in FIG3 , FIG3 exemplarily shows a one-dimensional array, where each rectangular box in the one-dimensional array represents an element, and each element can be filled with the measurement result of a beam pair.
[0166] It should be noted that when the first preset number is smaller than the number of elements included in the one-dimensional array, after the base station fills the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array, the one-dimensional array also includes unoccupied elements. The base station can set the unoccupied elements in the one-dimensional array to invalid values (V_invalid).
[0167] S204 : Sample elements according to preset sampling patterns corresponding to different scenarios, fill unsampled elements with invalid values, and obtain training data corresponding to each preset sampling pattern.
[0168] The preset sampling mode is pre-set according to the scene, and each scene corresponds to a preset sampling mode. The elements to be sampled in the preset sampling mode depend on the beam pair used in beam prediction in the corresponding scene.
[0169] For example, in one scenario, during beam prediction, if the base station uses 16 transmit beams to send pilot signals for beam measurement, and the terminal uses one receive beam to receive the pilot signals for beam measurement, then in this scenario, 16 beam pairs are used between the base station and the terminal for beam prediction. Accordingly, the preset sampling pattern for this scenario is to sample the elements corresponding to these 16 beam pairs in a one-dimensional array.
[0170] It can be understood that after sampling the elements corresponding to the above 16 beam pairs, the remaining elements in the one-dimensional array are filled with invalid values, so that a training data can be obtained.
[0171] Among them, the value of the invalid value can affect the convergence performance of the beam prediction model. The value of the invalid value can be configured according to actual needs, and the invalid value used in the model training stage and the model inference stage is the same.
[0172] By sampling a one-dimensional array using preset sampling patterns for different scenarios, multiple sets of training data can be obtained, each corresponding to a scenario. Furthermore, the labels corresponding to the multiple sets of training data obtained by sampling the same one-dimensional array are the same.
[0173] In an embodiment of the present application, the beam ID of the transmitting beam and the beam identification (ID) of the receiving beam included in the beam pair can be used to locate the element position in the one-dimensional array. The correspondence between the beam ID included in the beam pair and the element position will be introduced in detail in subsequent embodiments.
[0174] S205: Use multiple sets of training data and label data to train a deep learning model to obtain a beam prediction model.
[0175] Before training the beam prediction model, the method of S201-S204 can be used to generate multiple one-dimensional arrays, each one-dimensional array can be used as label data, and training data can be generated based on each one-dimensional array. Then, multiple groups of training data corresponding to each one-dimensional array can be obtained, and the beam prediction model can be trained using the training data and label data corresponding to the multiple one-dimensional arrays.
[0176] Using the above method, the base station can use the full transmit beam to send a pilot signal to the terminal and receive a measurement report message sent by the terminal. The measurement report message includes the measurement results of the first preset number of beam pairs, that is, the terminal has measured the full number of beam pairs between the base station and the terminal. The base station can fill the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array. Since the number of elements included in the one-dimensional array is the product of the maximum number of transmit beams and the maximum number of receive beams, the one-dimensional array is suitable for carrying measurement results in various scenarios. This is equivalent to unifying the format of model input data in different scenarios in the embodiment of the present application. The model does not need to distinguish between the data of each scenario, which can reduce the complexity of the model. The base station can also sample the elements according to the preset sampling mode corresponding to different scenarios. A set of training data can be obtained based on each preset sampling mode. In this way, multiple sets of training data can be obtained based on the measurement results reported by the terminal at one time, and there is no need to perform beam measurement for each scenario separately. This reduces the pilot overhead between the base station and the terminal and the difficulty of obtaining training data for each scenario, and reduces the computing power requirements of the base station and the terminal. Moreover, because the preset sampling patterns corresponding to different scenarios are used when sampling the one-dimensional array, it is equivalent to simulating multiple scenarios, which can make the beam prediction model trained using these training data have better generalization performance.
[0177] In some embodiments of the present application, as shown in FIG4 , the measurement result of each beam pair includes the beam ID of the receive beam, the pilot index of the transmit beam, and the pilot signal quality of the beam pair. The pilot index of the transmit beam is the Channel-State Information Reference Signal Resource Index (CRI), and the pilot signal quality of the beam pair can be the pilot signal strength.
[0178] The above-mentioned S202, receiving the measurement report message sent by the terminal, specifically includes the following two implementation methods:
[0179] In one implementation, the base station receives a first measurement report message sent by the terminal through physical layer signaling. The first measurement report message includes a valid value bitmap field and measurement results of a first preset number of beam pairs. The valid value bitmap field is used to indicate the position of the measurement results in the first measurement report message.
[0180] The number of measurement results that can be carried by the first measurement report message is greater than or equal to a first preset number, for example, the number may be the same as the number of elements in the one-dimensional array, or may also be the first preset number. When the terminal reports the measurement result through the first measurement report message, the position used to carry the measurement result in the first measurement report message may not be fully occupied, and the valid value bitmap field may be used to indicate the position occupied by the measurement result.
[0181] For example, if the first bit in the valid value bitmap field is set to 1, it means that the first position used to carry the measurement result in the first measurement report message has the measurement result of the beam pair; if the second bit in the valid value bitmap field is set to 0, it means that the second position used to carry the measurement result in the first measurement report message does not have the measurement result of the beam pair.
[0182] Taking the first preset number equal to N as an example, the fields included in the first measurement report message and the meaning of each field are shown in Table 1.
[0183] Table 1
[0184] RxBeamId is the beam ID of the receive beam.
[0185] In another implementation manner, the base station receives a second measurement report message sent by the terminal through media access control (MAC) layer signaling, where the second measurement report message includes the number of measurement results and measurement results of a first preset number of beam pairs.
[0186] The number of measurement results that can be carried by the second measurement report message is a first preset number.
[0187] Taking the first preset number equal to N as an example, the fields included in the second measurement report message and the meaning of each field are shown in Table 2.
[0188] Table 2
[0189] Using the above two implementation methods, the terminal can use physical layer signaling or MAC layer signaling to report the measurement results of the beam pair, and the embodiment of the present application specifies the reporting format of the terminal, so that the base station in different scenarios can accurately obtain the measurement results reported by each terminal and fill the obtained measurement results into a one-dimensional array.
[0190] After the base station receives the measurement results of each beam pair, the above S203 of filling the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array can be specifically implemented as follows:
[0191] Step 1: For each beam pair measurement result, convert the pilot index of the transmit beam included in the beam pair measurement result into a beam ID.
[0192] The base station stores the correspondence between the pilot index and the beam ID of the transmitted beam. The beam ID of the transmitted beam is numbered starting from 0. For example, if the number of transmitted beams of the base station is T, the value range of the beam ID of the transmitted beam is an integer from 0 to T-1.
[0193] Step 2: Calculate the storage serial number of the measurement result of the beam pair. The storage serial number is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams.
[0194] The terminal can number its own receive beam by its beam ID. While the terminal and the base station are connected, the beam ID of the receive beam remains unchanged. The beam ID of a receive beam starts at 0. For example, if the number of receive beams is R, the beam ID of a receive beam ranges from 0 to R-1.
[0195] The storage sequence number is the position number of each element in the one-dimensional array. Taking the one-dimensional array shown in Figure 3 as an example, the base station numbers the position of each element in the one-dimensional array from top to bottom. The position number of the first element at the top is 1, and the position number of the second element is 2, and so on, to obtain the position number of each element in the one-dimensional array.
[0196] Step 3: Fill the pilot signal quality included in the measurement result of the beam pair into the element indicated by the storage sequence number.
[0197] Assuming that the storage sequence number of the measurement result of a certain beam pair calculated by the base station is 2, the base station fills the pilot signal quality included in the measurement result of the beam pair into the element with position number 2 in the one-dimensional array shown in FIG3 .
[0198] Using this method, after receiving the measurement results of the beam pair, the base station can calculate a storage serial number based on the beam ID of the receiving beam and the beam ID of the transmitting beam of the beam pair, and fill the pilot signal quality included in the measurement results of the beam pair into the element indicated by the storage serial number. Since the number of elements included in the one-dimensional array is the product of the maximum number of transmitting beams and the maximum number of receiving beams, in any scenario, the measurement results of the received beam pair can be filled into an element of the one-dimensional array, that is, the one-dimensional array can carry the measurement results in different scenarios. The present application sets the training data in the form of a one-dimensional array, and then the beam prediction model does not need to distinguish the data of each scenario, which reduces the complexity of the beam prediction model and can save computing resources when the base station uses the beam prediction model.
[0199] The following describes the beam prediction model training method provided by the embodiment of the present application with reference to a specific example. As shown in FIG5 , the method includes:
[0200] S501. The terminal reports the RSRP signal strengths of T*R beam pairs to the base station.
[0201] Among them, T is the number of transmission beams of the base station, and R is the number of reception beams of the terminal. For example, the base station sends pilot signals to the terminal in the beam directions of 16 transmission beams, and the terminal uses 2 reception beams to receive the pilot signals. Then, the measurement results fed back by the terminal include a total of 16 * 2 = 32 beam pair measurement results. The measurement result of each beam pair includes the RSRP corresponding to the beam pair.
[0202] S502. The base station fills the RSRPs of T * R beam pairs into the elements included in a one-dimensional array, and fills the unoccupied elements in the Tmax * Rmax elements included in the one-dimensional array with invalid values, serving as the label data of the model.
[0203] Among them, the one-dimensional array includes Tmax * Rmax elements, that is, this one-dimensional array can be used to store the RSRPs of at most Tmax * Rmax beam pairs. When T < T_max or R < R_max, the base station can fill the unoccupied elements in the one-dimensional array with invalid values, that is, fill the elements corresponding to the beam pairs not reported by the terminal among the Tmax * Rmax beam pairs with invalid values.
[0204] Since the one-dimensional array includes the RSRPs of T * R beam pairs, which is equivalent to including the real measurement results of all beam pairs, using this one-dimensional array as label data can improve the model accuracy of the trained beam prediction model.
[0205] S503. The base station samples the measurement results of the terminal's T * R beam pairs according to a preset sampling pattern, and fills the non-sampled part in the Tmax * Rmax beam pairs with an invalid value V_invalid, serving as a set of training data of the model.
[0206] Among them, the base station can sample the measurement results stored in the one-dimensional array using the preset sampling patterns corresponding to different scenarios to obtain a set of training data corresponding to each scenario.
[0207] As an example, the number of transmission beams of the base station is T, the number of reception beams of the terminal is R, the maximum number of transmission beams is Tmax, and the maximum number of reception beams is Rmax. Then, the one-dimensional array includes Tmax * Rmax elements, and the terminal has reported a total of T * R beam pair measurement results.
[0208] As shown in Figure 6, the base station can divide the measurement results of T * R beam pairs into R groups, and fill the measurement results of each group into the one-dimensional array respectively. The shaded rectangular cells in Figure 6 are the elements filled with measurement results, and the blank rectangular cells represent the elements filled with invalid values (V_invalid).
[0209] Referring to the one-dimensional array shown in the left column of FIG6 , the first set of measurement results includes T valid values, such as the RSRP of T beam pairs consisting of receive beam 0 and T transmit beams. When T is less than Tmax, the first set of measurement results also includes Tmax minus T invalid values (V_invalid). Furthermore, the second set of measurement results includes T valid values, such as the RSRP of T beam pairs consisting of receive beam 1 and T transmit beams. Similarly, the third set of measurement results includes the RSRP of T beam pairs consisting of receive beam 2 and T transmit beams. Similarly, the Rth set of measurement results includes the RSRP of T beam pairs consisting of receive beam R-1 and T transmit beams.
[0210] Figure 6 exemplifies two preset sampling patterns, sampling pattern 1 and sampling pattern 2. The base station samples the one-dimensional array on the left side of Figure 6 using sampling pattern 1, resulting in a one-dimensional array in the middle of Figure 6 . This array can be used as a set of training data. The base station samples the one-dimensional array on the left side of Figure 6 using sampling pattern 2, resulting in a one-dimensional array on the right side of Figure 6 . This array can also be used as a set of training data. Furthermore, the one-dimensional array on the left side of Figure 6 can serve as the label data corresponding to the two types of training data described above.
[0211] It's understandable that each preset sampling pattern corresponds to a specific scenario. The more preset sampling patterns there are, the better the model generalization performance, but also the correspondingly greater the model measurement overhead. During training, the number and types of preset sampling patterns can be adjusted to balance model generalization performance and model measurement overhead, making the beam prediction model adaptable to beam prediction in a variety of scenarios.
[0212] S504: Repeat S501 to S503 to construct a training dataset and a label dataset.
[0213] It is understood that each execution of S501 to S503 can generate multiple sets of training data and corresponding label data. By executing these multiple times, measurement results reported by different terminals can be collected, thereby generating training data and label data multiple times. All the obtained training data can be constructed into a training dataset, and the label data corresponding to the training data can be constructed into a label dataset.
[0214] S505: Perform model training based on the constructed training data set and label data set, and store model parameters.
[0215] It is understood that the base station can use a supervised training method to train the deep learning model and use the trained model as the beam prediction model. After training, the base station can store the relevant parameters of the beam prediction model for subsequent beam prediction using the beam prediction model and subsequent optimization of the beam prediction model.
[0216] Using the above method, the terminal measures all beams between the base station and the terminal, and the terminal reports the RSRP of T*R beam pairs. The base station fills the measurement results of the beam pairs reported by the terminal into a one-dimensional array. Since the number of elements included in the one-dimensional array is the product of the maximum number of transmit beams Tmax and the maximum number of receive beams Rmax, the one-dimensional array is applicable to measurement results in various scenarios. This is equivalent to unifying the format of model input data in different scenarios in the embodiment of the present application. The model does not need to distinguish data from each scenario, which can reduce the complexity of the model. The base station then samples the elements in the one-dimensional array according to the preset sampling mode corresponding to different scenarios. A set of training data can be obtained based on each preset sampling mode. In this way, multiple training data can be obtained based on the measurement results reported by the terminal once, and there is no need to perform beam measurement for each scenario separately. This reduces the pilot overhead between the base station and the terminal, the difficulty of obtaining training data for each scenario, and reduces the computing power requirements of the base station and the terminal. In addition, because the preset sampling modes corresponding to different scenarios are used when sampling the one-dimensional array, it is equivalent to simulating multiple scenarios, which can make the beam prediction model trained using this training data have better generalization capabilities.
[0217] Furthermore, by adjusting the number of preset sampling patterns and flexibly configuring them, embodiments of the present application enable dynamic adjustments between beam prediction performance and measurement overhead, making the method more adaptable to specific scenarios. Simulation experiments demonstrated that the beam prediction model trained using this method achieved over 80% prediction accuracy, and compared to existing technologies, the air interface overhead and measurement duration of the beam prediction process were reduced by 75%.
[0218] Based on the above embodiment, the base station can use the trained beam prediction model to perform beam prediction. Corresponding to the above embodiment, an embodiment of the present application provides a beam prediction method, which is applied to the base station. As shown in FIG7 , the method includes:
[0219] S701: Send a pilot signal to the terminal using a transmit beam corresponding to the current scenario.
[0220] Among them, the transmitting beam corresponding to the current scene is specifically the transmitting beam corresponding to the preset sampling mode of the current scene. The preset sampling mode is one of the preset sampling modes used when training the beam prediction model, and the currently used preset sampling mode is the sampling mode corresponding to the current scene in which the current base station and terminal are located.
[0221] The base station stores the beam ID of the transmit beam corresponding to each preset sampling pattern. During the model training phase, the preset sampling pattern can be used to sample the one-dimensional array. The beam ID of the transmit beam corresponding to the sampled element is the beam ID of the transmit beam corresponding to the preset sampling pattern.
[0222] The number of transmitted beams corresponding to the preset sampling mode affects the accuracy of the beam prediction results. Fewer transmitted beams reduce the air interface measurement overhead for the base station and terminal, and shorten the terminal measurement latency. However, this may also result in a decrease in the accuracy of the beam prediction results output by the beam prediction model. When setting the preset sampling mode, embodiments of the present application can balance the number of transmitted beams with the model's prediction accuracy.
[0223] S702: After receiving the pilot signal on the receiving beam, the receiving terminal sends a measurement report message.
[0224] The transmit beam in S701 and the receive beam in S702 form a beam pair. Accordingly, the measurement report message includes the measurement result of the beam pair.
[0225] The terminal can use a preset receive beam to receive the pilot signal sent by the base station. The terminal's preset receive beam and the base station's transmit beam used to send the pilot signal form a beam pair. The base station can use multiple transmit beams to send the pilot signal, and the terminal can use at least one receive beam to receive the pilot signal. The measurement report message reported by the terminal includes measurement results for multiple beam pairs.
[0226] S703: Fill the measurement results of the beam pair into the elements included in the one-dimensional array, and fill the unoccupied elements with invalid values.
[0227] The number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams. The maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal.
[0228] Unoccupied elements in the one-dimensional array can be uniformly filled with invalid values (V_invalid), and the value of the invalid value is the same as the value of the invalid value in the beam prediction model training method provided in the above embodiment.
[0229] S704: Input the elements into the beam prediction model to obtain an output array of the beam prediction model, where the elements of the output array include the prediction results of the beam pairs.
[0230] The beam prediction model is a model trained using the beam prediction model training method in the above embodiment.
[0231] The output array has the same data format as the one-dimensional array, and each element in the output array is populated with the prediction result of a beam pair. Specifically, the output array has the same number of elements as the one-dimensional array. The prediction result for each beam pair can be the signal quality of the beam pair, with the beam pair with the highest signal quality being the optimal beam pair recommended by the beam prediction model. Alternatively, the prediction result for each beam pair can be the probability of the beam pair being the optimal beam pair, with the beam pair with the highest probability being the optimal beam pair.
[0232] During the beam prediction process, the base station may predict an optimal transmit beam, or may predict an optimal beam pair.
[0233] Taking the prediction result as the signal quality of the beam pair as an example, when the base station needs to predict the optimal transmission beam, the base station can determine the beam pair corresponding to the highest signal quality in the output array and use the transmission beam in the beam pair as the optimal transmission beam.
[0234] Specifically, the base station can determine the element with the highest signal quality stored in the output array, and determine the beam pair corresponding to the element based on the position of the element in the output array. Specifically, it can determine the beam ID of the transmitting beam and the beam ID of the receiving beam included in the beam pair, and then use the transmitting beam in the beam pair as the optimal transmitting beam.
[0235] When the base station needs to predict the optimal beam pair, the base station may use the beam pair corresponding to the highest signal quality in the output array as the optimal beam pair.
[0236] Using the above method, the base station uses the transmission beam corresponding to the current scenario to send a pilot signal to the terminal, and fills the measurement results of the beam pair reported by the terminal into the elements included in the one-dimensional array, and then inputs the one-dimensional array into the beam prediction model to obtain an output array. Since the number of elements included in the one-dimensional array is the product of the maximum number of transmission beams and the maximum number of reception beams, the measurement results reported by the terminal can all be filled into the elements included in the one-dimensional array, and there will be no situation where the one-dimensional array includes insufficient elements. The number of elements input into the beam prediction model in different scenarios is the same, so the beam prediction method provided in the embodiment of the present application is applicable to various scenarios, and the beam prediction model does not need to distinguish between different scenarios, which reduces the complexity of the model. In addition, the beam prediction model has good generalization performance and can accurately predict the beams of base stations and terminals in various scenarios.
[0237] In some embodiments of the present application, before the above S701, using the transmit beam corresponding to the current scenario to send a pilot signal to the terminal, the method further includes:
[0238] A first RRC message is sent to the terminal, and a second RRC message is received from the terminal.
[0239] The Channel State Information (CSI) measurement configuration field of the first Radio Resource Control (RRC) message includes a new field. The value of the new field is used to indicate the number of receive beams used by the terminal when performing beam measurement. The second RRC message is used to indicate that the terminal supports beam measurement according to the number of receive beams included in the new field. Both the first RRC message and the second RRC message belong to Layer 3 (L3) signaling.
[0240] In the 3rd Generation Partnership Project (3GPP) system, as shown in Figure 8, the base station can add the terminal receive beam number ueRxBeamNum in the Channel State Information-Measurement configuration (CSI-MeasConfig) of the RRC protocol message. ueRxBeamNum is used to indicate the number of receive beams used by the terminal when performing beam measurement. As an example, in Figure 8, ueRxBeamNum can be set to an integer (INTEGER) (1..8), which represents an integer between 1 and 8, and the ueRxBeamNum is optional.
[0241] The CSI-MeasConfig sequence (SQUENCE) also includes other bit fields. Other bit fields in CSI-MeasConfig are used to indicate the time-frequency resource and repetition period of the pilot signal. This application does not impose specific restrictions on how to use other bit fields to indicate the time-frequency resource and repetition period of the pilot signal, and reference may be made to the relevant 3GPP protocol provisions.
[0242] Using the above method, the base station adds a new bit field in the CSI measurement configuration field to convey the number of receive beams to the terminal, so that the terminal can perform pilot signal measurement in accordance with the instructions of the base station.
[0243] In some embodiments of the present application, the base station may receive the measurement report message sent by the terminal in the following two ways:
[0244] Method 1: The receiving terminal sends a third measurement report message through physical layer signaling. The third measurement report message includes a valid value bitmap field and the measurement result of the beam pair. The valid value bitmap field is used to indicate the position of the measurement result in the third measurement report message.
[0245] The format of the third measurement report message is the same as that of the first measurement report message in the above embodiment. For specific content, refer to the relevant description in the above embodiment. The difference is that the first measurement report message includes the measurement results of all beam pairs, while the third measurement report message includes the measurement results of a smaller number of beam pairs. By using the position of the bit with a value of 1 in the valid value bitmap field, the base station can quickly locate and obtain the measurement results of the beam pair from the third measurement report message.
[0246] For example, there are 64 beam pairs between the base station and the terminal, and the number of transmitting beams corresponding to the current scenario is 16. The terminal uses 1 receiving beam to receive the pilot signal sent by the base station using 16 transmitting beams. In this case, the third measurement report message includes the measurement results of 16 beam pairs, and the value of the bit corresponding to the position of these 16 beam pairs in the valid bitmap field is 1.
[0247] Method 2: The fourth measurement report message is sent by the receiving terminal through MAC layer signaling. The fourth measurement report message includes the number of measurement results and the measurement results of the beam pair.
[0248] The format of the fourth measurement report message is the same as that of the second measurement report message in the above embodiment. For specific content, please refer to the relevant description in the above embodiment.
[0249] As an example, there are 64 beam pairs between the base station and the terminal, and the number of transmitting beams corresponding to the current scenario is 16. The terminal uses 2 receiving beams to receive the pilot signal sent by the base station using 16 transmitting beams. In this case, the third measurement report message includes the measurement results of 32 beam pairs, and the number of measurement results included in the fourth measurement report message is 32.
[0250] Using the above two methods, the terminal can report the measurement results of the beam pair to the base station in a unified format, making this method applicable to various scenarios and facilitating the base station to subsequently use the measurement results of the beam pair to generate a one-dimensional array.
[0251] In some embodiments of the present application, the measurement result of each beam pair includes the beam ID of the receive beam, the pilot index of the transmit beam, and the pilot signal quality of the beam pair. The above S703, filling the beam pair measurement results into the elements included in the one-dimensional array, can be specifically implemented as follows:
[0252] Step A: For each beam pair, convert the pilot index of the transmit beam included in the measurement result of the beam pair into a beam ID.
[0253] The base station stores a correspondence between the pilot index and the beam ID of the transmitted beam. Based on the correspondence, the base station can convert the pilot index of the transmitted beam included in the measurement result of the beam pair into the beam ID.
[0254] Step B: Calculate the storage serial number of the measurement result of the beam pair, the storage serial number is: Ri*R_max+Tj, Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams.
[0255] Step C: Fill the measurement result of the beam pair into the element indicated by the storage sequence number.
[0256] The method of filling the measurement results of each beam pair into the one-dimensional array in steps AC can be referred to the relevant introduction in the above embodiment, and will not be repeated here.
[0257] Since the data structure and filling method of the one-dimensional array in the beam prediction process are the same as the data structure and filling method of the one-dimensional array in the model training process, that is, the embodiment of the present application unifies the structure of the input data of the beam prediction model, so that the beam prediction method can be applied to a variety of scenarios, and can accurately use the measurement results of a small number of beam pairs for beam prediction.
[0258] In addition to predicting the optimal transmit beam, the base station can also predict multiple receive beams for the terminal and notify the terminal of the predicted receive beams, so that the terminal can select a receive beam from the predicted receive beams for receiving subsequent signals sent by the base station. Based on this, after the above S704, inputting the elements into the beam prediction model and obtaining the output array of the beam prediction model, the method further includes:
[0259] A first number of elements including the optimal transmit beams are obtained from the elements of the output array, and a second number of receive beams are selected from the prediction results included in the first number of elements. A beam prediction message is then sent to the terminal via physical layer signaling or MAC layer signaling. The beam prediction message includes the second number of beams, beam IDs of the second number of receive beams, and priorities of the second number of receive beams.
[0260] The method for determining the optimal transmit beam can be referred to the relevant description in the above embodiment. After determining the optimal transmit beam, the base station can filter out the elements of the output array to which the optimal transmit beam belongs, and then select a second number of receive beams based on the predicted results of the filtered elements in descending order.
[0261] Taking signal quality as an example, after the base station obtains and filters out the elements that belong to the optimal transmit beam, it then selects a second number of elements in descending order of signal quality from the filtered elements, and then obtains the beam IDs of the receive beams corresponding to the second number of elements. It is understood that higher signal quality is associated with a higher priority receive beam.
[0262] The second number is the number of receiving beams recommended by the base station to the terminal, and the second number is less than or equal to the maximum number of receiving beams supported by the terminal. The number of receiving beams selected by the base station can be preset. Alternatively, the base station can sort the signal qualities of the elements included in the first number of elements in descending order, compare the difference between the first signal quality and the second signal quality, and if the difference is less than a specified threshold, continue to compare the difference between the second signal quality and the third signal quality, and so on, until the difference between the mth signal quality and the m+1th signal quality is greater than the specified threshold, then stop comparing, or stop comparing when m is the maximum number of receiving beams supported by the terminal. The base station then selects the receiving beams corresponding to the elements with the signal qualities from 1 to m and recommends them to the terminal.
[0263] In this embodiment, a physical layer signaling message format is redefined. The beam prediction message sent via both physical layer signaling and MAC layer signaling includes the predicted number of receive beams and the beam IDs of the second number of receive beams, arranged in priority order. The fields included in the beam prediction message are shown in Table 3.
[0264] Table 3
[0265] Among them, the first priority in the above table is the highest priority. After the terminal receives the beam prediction message, it can select one or more receiving beams according to the priority of each receiving beam included in the beam prediction message as the receiving beam to be used in subsequent communications with the base station. For example, the terminal can select a receiving beam with the highest priority according to its own configuration. When the receiving beam is currently not suitable for receiving the signal of the base station, it can select a receiving beam with the second highest priority. Alternatively, the terminal can also select a specified number of receiving beams arranged in descending order of priority according to its own configuration. The specified number is less than or equal to the maximum number of receiving beams supported by the terminal.
[0266] It can be understood that after the terminal receives the beam prediction message, it can use the priority of the second number of receiving beams indicated by the beam prediction message to select the receiving beam for receiving the signals subsequently sent by the base station, which can improve the communication quality between the base station and the terminal.
[0267] As shown in FIG9 , FIG9 is an interactive flow chart of a beam prediction method provided in an embodiment of the present application, the method including:
[0268] S901. A base station sends a first radio resource control (RRC) message to a terminal.
[0269] The first RRC message includes the time-frequency resources, repetition and period of the pilot signal and the number of receiving beams when the terminal performs beam measurement.
[0270] S902. The terminal replies with a second RRC message to the base station.
[0271] If the terminal supports the measurement mode indicated by the base station through the first RRC message, the terminal feeds back a second RRC message to the base station.
[0272] S903: The base station performs beamforming of the pilot signal in a preset sampling mode and sends the pilot signal.
[0273] It is understandable that the base station transmits the pilot signal using a transmit beam in a preset sampling pattern, wherein the preset sampling pattern is one of the multiple sampling patterns in the beam prediction model training method provided in the above embodiment.
[0274] S904: The terminal uses a preset receiving beam to measure the pilot signal strength.
[0275] After receiving the first RRC message sent by the base station, the terminal can obtain the number of receive beams in the first RRC message. When the terminal needs to measure the pilot signal strength, the number of preset receive beams used is the same as the number of receive beams obtained from the first RRC message.
[0276] S905. The terminal reports the measurement results of each beam pair to the base station.
[0277] The measurement result of each beam pair includes the pilot index, the beam ID of the received beam, and the signal quality of the pilot signal.
[0278] S906: Construct input data of a beam prediction model based on the measurement results of each beam pair.
[0279] It is understandable that the base station can fill the measurement results into a one-dimensional array and use the one-dimensional array as input data for the beam prediction model. The method of filling the measurement results into the one-dimensional array is described in the above embodiment and will not be repeated here.
[0280] S907: Use the beam prediction model to predict the optimal beam or beam pair.
[0281] It can be understood that the base station can obtain the output array of the beam prediction model, and the output array includes the prediction result of each beam pair.
[0282] S908: When a beam pair needs to be predicted, the base station recommends the predicted receiving beam to the terminal.
[0283] Among them, the base station can use MAC layer signaling or physical layer signaling to send a beam number prediction message to the terminal. The message format of the beam prediction message refers to the relevant description in the above embodiment and will not be repeated here.
[0284] Using the above method, the base station adds a new field to the CSI measurement configuration field to convey the number of receive beams to the terminal. The base station then uses the preset sampling mode to perform beamforming of the pilot signal. The preset sampling mode corresponds to the current scenario of the base station and the terminal, and the preset sampling mode is one of the preset sampling modes in the beam prediction model training phase, so the beam prediction waveform can be applied to beam prediction in this scenario. Since the base station and the terminal do not need to collect measurement data for all beams in the preset sampling mode, the base station and the terminal can complete the prediction of the optimal beam or beam pair by collecting less data, which improves the efficiency of beam prediction and reduces the computing power resources occupied by the beam prediction process.
[0285] Based on the same concept, an embodiment of the present application provides a beam prediction model training device, which is applied to a base station. As shown in FIG10 , the device includes:
[0286] A sending module 1001 is configured to send a pilot signal to a terminal using a full transmit beam;
[0287] A receiving module 1002 is configured to receive a measurement report message sent by a terminal, where the measurement report message includes measurement results of a first preset number of beam pairs, where the first preset number is the product of the number of transmit beams of the base station and the number of receive beams of the terminal;
[0288] A filling module 1003 is configured to fill the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array, and use the one-dimensional array as label data, where the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, where the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal;
[0289] Sampling module 1004, configured to sample elements according to preset sampling patterns corresponding to different scenarios, fill unsampled elements with invalid values, and obtain training data corresponding to each preset sampling pattern;
[0290] The training module 1005 is used to train the deep learning model using multiple sets of training data and label data to obtain a beam prediction model.
[0291] Optionally, the receiving module 1002 is specifically configured to:
[0292] receiving a first measurement report message sent by a terminal through physical layer signaling, where the first measurement report message includes a valid value bitmap field and measurement results of a first preset number of beam pairs, where the valid value bitmap field is used to indicate a position of the measurement results in the first measurement report message; or
[0293] The receiving terminal sends a second measurement report message through MAC layer signaling, where the second measurement report message includes the number of measurement results and the measurement results of a first preset number of beam pairs.
[0294] Optionally, the measurement result of each beam pair includes the beam ID of the receiving beam, the pilot index of the transmitting beam, and the pilot signal quality of the beam pair;
[0295] Fill in module 1003, specifically for:
[0296] For each beam pair measurement result, convert the pilot index of the transmit beam included in the beam pair measurement result into a beam ID;
[0297] Calculate the storage sequence number of the measurement result of the beam pair, which is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams;
[0298] The pilot signal quality included in the measurement result of the beam pair is filled into the element indicated by the storage sequence number.
[0299] The embodiment of the present application further provides a beam prediction device, which is applied to a base station. As shown in FIG11 , the device includes:
[0300] The sending module 1101 is configured to send a pilot signal to the terminal using a transmit beam corresponding to the current scenario;
[0301] The receiving module 1102 is configured to receive a measurement report message sent by the terminal after receiving a pilot signal on a receiving beam, where the transmitting beam and the receiving beam form a beam pair, and the measurement report message includes a measurement result of the beam pair;
[0302] A filling module 1103 is configured to fill the measurement results of the beam pairs into the elements included in the one-dimensional array and fill unoccupied elements with invalid values; the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, where the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal;
[0303] The acquisition module 1104 is configured to input elements into the beam prediction model and acquire an output array of the beam prediction model, where the elements of the output array include prediction results of the beam pairs.
[0304] Optionally, the receiving module 1102 is specifically configured to:
[0305] receiving a third measurement report message sent by the terminal through physical layer signaling, where the third measurement report message includes a valid value bitmap field and a measurement result of the beam pair, where the valid value bitmap field is used to indicate a position of the measurement result in the third measurement report message; or
[0306] The receiving terminal sends a fourth measurement report message through MAC layer signaling, where the fourth measurement report message includes the number of measurement results and the measurement results of the beam pairs.
[0307] Optionally, the measurement result of each beam pair includes the beam ID of the receiving beam, the pilot index of the transmitting beam, and the pilot signal quality of the beam pair;
[0308] Fill in module 1103, specifically for:
[0309] For each beam pair, convert the pilot index of the transmit beam included in the measurement result of the beam pair into a beam ID;
[0310] Calculate the storage sequence number of the measurement result of the beam pair, which is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams;
[0311] The measurement result of the beam pair is filled in the element indicated by the storage sequence number.
[0312] Optionally, the device further includes a selection module:
[0313] The acquisition module 1104 is further configured to acquire a first number of elements including the optimal transmit beam from the elements of the output array;
[0314] A selection module, configured to select a second number of receive beams from the prediction results included in the first number of elements;
[0315] The sending module 1101 is further used to send a beam prediction message to the terminal through physical layer signaling or MAC layer signaling, where the beam prediction message includes the second number, the beam IDs of the second number of receiving beams, and the priorities of the second number of receiving beams.
[0316] Optional,
[0317] The sending module 1101 is further configured to send a first RRC message to the terminal before sending a pilot signal to the terminal using the transmit beam corresponding to the current scenario, wherein the CSI measurement configuration field of the first RRC message includes a newly added field, and the value of the newly added field is used to indicate the number of receive beams used by the terminal for beam measurement;
[0318] The receiving module 1102 is further used to receive a second RRC message sent by the terminal, where the second RRC message is used to indicate that the terminal supports beam measurement according to the number of receiving beams included in the newly added field.
[0319] An embodiment of the present application provides a base station, as shown in FIG12 , including:
[0320] Processor 1201;
[0321] transceiver 1204;
[0322] A machine-readable storage medium 1202 stores machine-executable instructions that can be executed by the processor 1201; the machine-executable instructions prompt the processor 1201 to perform the following steps:
[0323] Transmitting a pilot signal to the terminal using a full transmit beam via transceiver 1204;
[0324] Receiving, through the transceiver 1204, a measurement report message sent by the terminal, where the measurement report message includes measurement results of a first preset number of beam pairs, where the first preset number is the product of the number of transmit beams of the base station and the number of receive beams of the terminal;
[0325] Filling the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array, and using the one-dimensional array as label data, where the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, where the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal;
[0326] Sample the elements according to the preset sampling patterns corresponding to different scenarios, fill the unsampled elements with invalid values, and obtain training data corresponding to each preset sampling pattern;
[0327] The deep learning model is trained using multiple sets of training data and label data to obtain a beam prediction model.
[0328] Optionally, the machine executable instructions further cause the processor 1201 to perform the following steps:
[0329] Receiving, through transceiver 1204, a first measurement report message sent by the terminal through physical layer signaling, where the first measurement report message includes a valid value bitmap field and measurement results of a first preset number of beam pairs, where the valid value bitmap field is used to indicate a position of the measurement result in the first measurement report message; or
[0330] A second measurement report message sent by the terminal through MAC layer signaling is received through the transceiver 1204, where the second measurement report message includes the number of measurement results and the measurement results of a first preset number of beam pairs.
[0331] Optionally, the measurement result of each beam pair includes the beam ID of the receiving beam, the pilot index of the transmitting beam, and the pilot signal quality of the beam pair;
[0332] The machine executable instructions further cause the processor 1201 to perform the following steps:
[0333] For each beam pair measurement result, convert the pilot index of the transmit beam included in the beam pair measurement result into a beam ID;
[0334] Calculate the storage sequence number of the measurement result of the beam pair, which is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams;
[0335] The pilot signal quality included in the measurement result of the beam pair is filled into the element indicated by the storage sequence number.
[0336] In Figure 12 , a communication bus 1203 may also be included. The processor 1201, the machine-readable storage medium 1202, and the transceiver 1204 communicate with each other via the communication bus 1203. The communication bus 1203 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus may be divided into an address bus, a data bus, a control bus, and the like.
[0337] The transceiver 1204 may be a wireless communication module. Under the control of the processor 1201 , the transceiver 1204 exchanges data with other devices.
[0338] The machine-readable storage medium 1202 may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Alternatively, the machine-readable storage medium may be at least one storage device located remote from the processor.
[0339] The processor 1201 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may 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, or discrete hardware components.
[0340] The embodiment of the present application further provides a base station, as shown in FIG13 , the base station including:
[0341] Processor 1301;
[0342] transceiver 1304;
[0343] Machine-readable storage medium 1302, the machine-readable storage medium 1302 stores machine-executable instructions that can be executed by processor 1301; the machine-executable instructions prompt processor 1301 to perform the following steps:
[0344] Using the transmit beam corresponding to the current scenario, a pilot signal is sent to the terminal via transceiver 1304;
[0345] The transceiver 1304 receives a measurement report message sent by the terminal after receiving the pilot signal on the receiving beam. The transmitting beam and the receiving beam form a beam pair. The measurement report message includes a measurement result of the beam pair.
[0346] Fill the beam pair measurement results into the elements of the one-dimensional array, and fill the unoccupied elements with invalid values; the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, where the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal;
[0347] Input the elements into the beam prediction model and obtain the output array of the beam prediction model. The elements of the output array include the prediction results of the beam pairs.
[0348] Optionally, the machine executable instructions further cause the processor 1301 to perform the following steps:
[0349] Receiving, through transceiver 1304, a third measurement report message sent by the terminal through physical layer signaling, where the third measurement report message includes a valid value bitmap field and a measurement result of the beam pair, where the valid value bitmap field is used to indicate a position of the measurement result in the third measurement report message; or
[0350] The fourth measurement report message sent by the terminal through MAC layer signaling is received through the transceiver 1304. The fourth measurement report message includes the number of measurement results and the measurement results of the beam pairs.
[0351] Optionally, the measurement result of each beam pair includes the beam ID of the receiving beam, the pilot index of the transmitting beam, and the pilot signal quality of the beam pair;
[0352] The machine executable instructions further cause the processor 1301 to perform the following steps:
[0353] For each beam pair, convert the pilot index of the transmit beam included in the measurement result of the beam pair into a beam ID;
[0354] Calculate the storage sequence number of the measurement result of the beam pair, which is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams;
[0355] The measurement result of the beam pair is filled in the element indicated by the storage sequence number.
[0356] Optionally, the machine executable instructions further cause the processor 1301 to perform the following steps:
[0357] Obtaining a first number of elements including the optimal transmit beam from the elements of the output array;
[0358] Selecting a second number of receive beams from the prediction results included in the first number of elements;
[0359] The transceiver 1304 is used to send a beam prediction message to the terminal through physical layer signaling or MAC layer signaling, where the beam prediction message includes the second number, beam IDs of the second number of receiving beams, and priorities of the second number of receiving beams.
[0360] Optionally, the machine executable instructions further cause the processor 1301 to perform the following steps:
[0361] Sending a first RRC message to the terminal through the transceiver 1304, where the CSI measurement configuration field of the first RRC message includes a newly added field, where the value of the newly added field is used to indicate the number of receive beams used by the terminal when performing beam measurement;
[0362] A second RRC message sent by the terminal is received through the transceiver 1304, where the second RRC message is used to indicate that the terminal supports beam measurement according to the number of receive beams included in the newly added field.
[0363] In Figure 13 , a communication bus 1303 may also be included. The processor 1301, the machine-readable storage medium 1302, and the transceiver 1304 communicate with each other via the communication bus 1303. The communication bus 1303 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus may be divided into an address bus, a data bus, a control bus, and the like.
[0364] The transceiver 1304 may be a wireless communication module. Under the control of the processor 1301 , the transceiver 1304 exchanges data with other devices.
[0365] The machine-readable storage medium 1302 may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Alternatively, the machine-readable storage medium may be at least one storage device located remotely from the processor.
[0366] Processor 1301 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 device, discrete gate or transistor logic device, discrete hardware component.
[0367] Based on the same inventive concept, in accordance with the beam prediction model training method and beam prediction method provided in the above-mentioned embodiments of the present application, the embodiments of the present application further provide a machine-readable storage medium storing machine-executable instructions executable by a processor. The machine-executable instructions cause the processor to implement the steps of the above-mentioned beam prediction model training method or beam prediction method.
[0368] In another embodiment provided in the present application, a computer program product comprising instructions is also provided, which, when executed on a computer, enables the computer to execute the steps of the beam prediction model training method or the beam prediction method in the above-mentioned embodiment.
[0369] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.
[0370] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A beam prediction model training method, characterized in that: The method is applied to a base station, and the method includes: Sending a pilot signal to the terminal using a full transmit beam; receiving a measurement report message sent by the terminal, where the measurement report message includes measurement results of a first preset number of beam pairs, where the first preset number is the product of the number of transmit beams of the base station and the number of receive beams of the terminal; Filling the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array, and using the one-dimensional array as label data, the number of the elements is the product of the maximum number of transmit beams and the maximum number of receive beams, the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal; The elements are sampled according to preset sampling modes corresponding to different scenarios, and invalid values are filled in the unsampled elements to obtain training data corresponding to each preset sampling mode; The deep learning model is trained using multiple sets of training data and the label data to obtain a beam prediction model.
2. The method according to claim 1, characterized in that The receiving a measurement report message sent by the terminal includes: receiving a first measurement report message sent by the terminal through physical layer signaling, wherein the first measurement report message includes a valid value bitmap field and measurement results of the first preset number of beam pairs, and the valid value bitmap field is used to indicate the position of the measurement results in the first measurement report message; or, Receive a second measurement report message sent by the terminal through MAC layer signaling, where the second measurement report message includes the number of measurement results and the measurement results of the first preset number of beam pairs.
3. The method according to claim 1 or 2, characterized in that: The measurement results of each beam pair include the beam ID of the receiving beam, the pilot index of the transmitting beam, and the pilot signal quality of the beam pair; Filling the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array includes: For each beam pair measurement result, convert the pilot index of the transmit beam included in the beam pair measurement result into a beam ID; Calculate the storage sequence number of the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams; Fill the pilot signal quality included in the measurement result of the beam pair into the element indicated by the storage sequence number.
4. A beam prediction method, characterized in that: The method is applied to a base station, and the method includes: Send a pilot signal to the terminal using the transmit beam corresponding to the current scenario; receiving a measurement report message sent by the terminal after receiving the pilot signal on a receiving beam, wherein the transmitting beam and the receiving beam form a beam pair, and the measurement report message includes a measurement result of the beam pair; Fill the measurement results of the beam pairs into the elements included in the one-dimensional array, and fill the unoccupied elements with invalid values; the number of the elements is the product of the maximum number of transmit beams and the maximum number of receive beams, the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal; The elements are input into a beam prediction model to obtain an output array of the beam prediction model, wherein the elements of the output array include prediction results of the beam pairs.
5. The method according to claim 4, characterized in that The receiving a measurement report message sent by the terminal after receiving the pilot signal on the receiving beam includes: receiving a third measurement report message sent by the terminal through physical layer signaling, where the third measurement report message includes a valid value bitmap field and a measurement result of the beam pair, where the valid value bitmap field is used to indicate a position of the measurement result in the third measurement report message; or, Receive a fourth measurement report message sent by the terminal through MAC layer signaling, where the fourth measurement report message includes the number of measurement results and the measurement results of the beam pair.
6. The method according to claim 4 or 5, characterized in that: The measurement result of each beam pair includes a beam ID of a receiving beam, a pilot index of a transmitting beam, and a pilot signal quality of the beam pair; Filling the measurement results of the beam pair into the elements of the one-dimensional array comprises: For each beam pair, convert the pilot index of the transmit beam included in the measurement result of the beam pair into a beam ID; Calculate the storage sequence number of the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams; The measurement result of the beam pair is filled into the element indicated by the storage sequence number.
7. The method according to claim 4, characterized in that After obtaining the output array of the beam prediction model, the method further includes: Obtaining a first number of elements including an optimal transmit beam from the elements of the output array; Selecting a second number of receive beams from the prediction results included in the first number of elements; A beam prediction message is sent to the terminal through physical layer signaling or MAC layer signaling, and the beam prediction message includes the second number, beam IDs of the second number of receiving beams, and priorities of the second number of receiving beams.
8. The method according to claim 4, characterized in that Before sending the pilot signal to the terminal using the transmit beam corresponding to the current scenario, the method further includes: Sending a first RRC message to the terminal, where the CSI measurement configuration field of the first RRC message includes a newly added field, and a value of the newly added field is used to indicate the number of receive beams used by the terminal when performing beam measurement; Receive a second RRC message sent by the terminal, where the second RRC message is used to indicate that the terminal supports beam measurement according to the number of receiving beams included in the newly added field.
9. A beam prediction model training device, characterized in that: The device is applied to a base station, and the device includes: A sending module, used to send a pilot signal to a terminal using a full transmit beam; A receiving module, configured to receive a measurement report message sent by the terminal, wherein the measurement report message includes measurement results of a first preset number of beam pairs, where the first preset number is the product of the number of transmit beams of the base station and the number of receive beams of the terminal; A filling module, used to fill the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array, and use the one-dimensional array as label data, the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal; A sampling module, used to sample the elements according to preset sampling modes corresponding to different scenarios, fill unsampled elements with invalid values, and obtain training data corresponding to each preset sampling mode; The training module is used to train the deep learning model using multiple sets of training data and the label data to obtain a beam prediction model.
10. The device according to claim 9, characterized in that The receiving module is specifically used for: receiving a first measurement report message sent by the terminal through physical layer signaling, wherein the first measurement report message includes a valid value bitmap field and measurement results of the first preset number of beam pairs, and the valid value bitmap field is used to indicate the position of the measurement results in the first measurement report message; or, Receive a second measurement report message sent by the terminal through MAC layer signaling, where the second measurement report message includes the number of measurement results and the measurement results of the first preset number of beam pairs.
11. The device according to claim 9 or 10, characterized in that The measurement results of each beam pair include the beam ID of the receiving beam, the pilot index of the transmitting beam, and the pilot signal quality of the beam pair; The filling module is specifically used for: For each beam pair measurement result, convert the pilot index of the transmit beam included in the beam pair measurement result into a beam ID; Calculate the storage sequence number of the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams; Fill the pilot signal quality included in the measurement result of the beam pair into the element indicated by the storage sequence number.
12. A beam prediction device, characterized in that: The device is applied to a base station, and the device includes: A sending module, used to send a pilot signal to the terminal using a sending beam corresponding to the current scenario; A receiving module, configured to receive a measurement report message sent by the terminal after receiving the pilot signal on a receiving beam, wherein the transmitting beam and the receiving beam form a beam pair, and the measurement report message includes a measurement result of the beam pair; A filling module is used to fill the measurement results of the beam pair into the elements included in the one-dimensional array, and fill the unoccupied elements with invalid values; the number of elements is the product of the maximum number of transmit beams and the maximum number of receive beams, the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal; The acquisition module is used to input the elements into the beam prediction model to obtain the output array of the beam prediction model, and the elements of the output array include the prediction results of the beam pairs.
13. The device according to claim 12, characterized in that The receiving module is specifically used for: receiving a third measurement report message sent by the terminal through physical layer signaling, where the third measurement report message includes a valid value bitmap field and a measurement result of the beam pair, where the valid value bitmap field is used to indicate a position of the measurement result in the third measurement report message; or, Receive a fourth measurement report message sent by the terminal through MAC layer signaling, where the fourth measurement report message includes the number of measurement results and the measurement results of the beam pair.
14. The device according to claim 12 or 13, characterized in that The measurement result of each beam pair includes a beam ID of a receiving beam, a pilot index of a transmitting beam, and a pilot signal quality of the beam pair; The filling module is specifically used for: For each beam pair, convert the pilot index of the transmit beam included in the measurement result of the beam pair into a beam ID; Calculate the storage sequence number of the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams; The measurement result of the beam pair is filled into the element indicated by the storage sequence number.
15. The device according to claim 12, characterized in that The device also includes a selection module: The acquisition module is further used to acquire a first number of elements including the optimal transmit beam from the elements of the output array; The selection module is used to select a second number of receiving beams from the prediction results included in the first number of elements; The sending module is also used to send a beam prediction message to the terminal through physical layer signaling or MAC layer signaling, and the beam prediction message includes the second number, the beam IDs of the second number of receiving beams, and the priorities of the second number of receiving beams.
16. The device according to claim 12, characterized in that The sending module is further configured to send a first RRC message to the terminal before sending a pilot signal to the terminal using the transmit beam corresponding to the current scenario, wherein the CSI measurement configuration field of the first RRC message includes a newly added field, and the value of the newly added field is used to indicate the number of receive beams used by the terminal for beam measurement; The receiving module is further configured to receive a second RRC message sent by the terminal, wherein the second RRC message is used to indicate that the terminal Supporting beam measurement according to the number of receive beams included in the newly added field.
17. A base station, characterized in that: include: processor; Transceiver; 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: Sending a pilot signal to a terminal by using a full transmit beam through the transceiver; Receiving, by the transceiver, a measurement report message sent by the terminal, where the measurement report message includes measurement results of a first preset number of beam pairs, where the first preset number is the product of the number of transmit beams of the base station and the number of receive beams of the terminal; Filling the measurement results of the first preset number of beam pairs into the elements included in the one-dimensional array, and using the one-dimensional array as label data, the number of the elements is the product of the maximum number of transmit beams and the maximum number of receive beams, the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal; The elements are sampled according to preset sampling modes corresponding to different scenarios, and invalid values are filled in the unsampled elements to obtain training data corresponding to each preset sampling mode; The deep learning model is trained using multiple sets of training data and the label data to obtain a beam prediction model.
18. The base station according to claim 17, characterized in that: The machine executable instructions further cause the processor to perform the following steps: receiving, by the transceiver, a first measurement report message sent by the terminal through physical layer signaling, wherein the first measurement report message includes a valid value bitmap field and measurement results of the first preset number of beam pairs, and the valid value bitmap field is used to indicate the position of the measurement results in the first measurement report message; or, A second measurement report message sent by the terminal through MAC layer signaling is received by the transceiver, where the second measurement report message includes the number of measurement results and the measurement results of the first preset number of beam pairs.
19. The base station according to claim 17 or 18, characterized in that: The measurement results of each beam pair include the beam ID of the receiving beam, the pilot index of the transmitting beam, and the pilot signal quality of the beam pair; The machine executable instructions further cause the processor to perform the following steps: For each beam pair measurement result, convert the pilot index of the transmit beam included in the beam pair measurement result into a beam ID; Calculate the storage sequence number of the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams; Fill the pilot signal quality included in the measurement result of the beam pair into the element indicated by the storage sequence number.
20. A base station, characterized in that: include: processor; Transceiver; 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: Sending a pilot signal to the terminal through the transceiver using a transmit beam corresponding to the current scenario; Receiving, by the transceiver, a measurement report message sent by the terminal after receiving the pilot signal on a receiving beam, the transmitting beam and the receiving beam forming a beam pair, and the measurement report message including a measurement result of the beam pair; Fill the measurement results of the beam pair into the elements included in the one-dimensional array, and fill the unoccupied elements with invalid values; The number is the product of the maximum number of transmit beams and the maximum number of receive beams, the maximum number of transmit beams is the maximum number of transmit beams of the base station, and the maximum number of receive beams is the maximum number of receive beams of the terminal; The elements are input into a beam prediction model to obtain an output array of the beam prediction model, wherein the elements of the output array include prediction results of the beam pairs.
21. The base station according to claim 20, characterized in that The machine executable instructions further cause the processor to perform the following steps: receiving, by the transceiver, a third measurement report message sent by the terminal through physical layer signaling, wherein the third measurement report message includes a valid value bitmap field and a measurement result of the beam pair, wherein the valid value bitmap field is used to indicate a position of the measurement result in the third measurement report message; or, A fourth measurement report message sent by the terminal through MAC layer signaling is received through the transceiver, and the fourth measurement report message includes the number of measurement results and the measurement results of the beam pair.
22. The base station according to claim 20 or 21, characterized in that: The measurement result of each beam pair includes a beam ID of a receiving beam, a pilot index of a transmitting beam, and a pilot signal quality of the beam pair; The machine executable instructions further cause the processor to perform the following steps: For each beam pair, convert the pilot index of the transmit beam included in the measurement result of the beam pair into a beam ID; Calculate the storage sequence number of the measurement result of the beam pair, where the storage sequence number is: Ri*R_max+Tj, where Ri is the beam ID of the receiving beam in the beam pair, Tj is the beam ID of the transmitting beam in the beam pair, and R_max is the maximum number of receiving beams; The measurement result of the beam pair is filled into the element indicated by the storage sequence number.
23. The base station according to claim 20, characterized in that The machine executable instructions further cause the processor to perform the following steps: Obtaining a first number of elements including an optimal transmit beam from the elements of the output array; Selecting a second number of receive beams from the prediction results included in the first number of elements; The transceiver is used to send a beam prediction message to the terminal through physical layer signaling or MAC layer signaling, and the beam prediction message includes the second number, the beam IDs of the second number of receiving beams, and the priorities of the second number of receiving beams.
24. The base station according to claim 20, characterized in that The machine executable instructions further cause the processor to perform the following steps: Sending a first RRC message to the terminal through the transceiver, wherein the CSI measurement configuration field of the first RRC message includes a newly added field, and a value of the newly added field is used to indicate the number of receiving beams used by the terminal when performing beam measurement; A second RRC message sent by the terminal is received through the transceiver, where the second RRC message is used to indicate that the terminal supports beam measurement according to the number of receiving beams included in the newly added field.
25. A machine-readable storage medium, characterized in that: Machine executable instructions are stored, and when called and executed by a processor, the machine executable instructions prompt the processor to implement any method described in claims 1-3 or 4-8.
26. A computer program product, characterized in that The computer program product causes the processor to implement the method of any one of claims 1-3 or 4-8.
Citation Information
Patent Citations
Method for determining optimal beam and an electronic device thereof
US20200212991A1
Beam Measurement Method, Network Device, and User Equipment
US20210160724A1
Predictive methods for SSB beam measurements
US20220408381A1
Enhanced beam failure indicator and applications
US20230104716A1
