WO2024159458A1 - 训练数据采集方法、装置、基站、终端及存储介质 - Google Patents
训练数据采集方法、装置、基站、终端及存储介质 Download PDFInfo
- Publication number
- WO2024159458A1 WO2024159458A1 PCT/CN2023/074151 CN2023074151W WO2024159458A1 WO 2024159458 A1 WO2024159458 A1 WO 2024159458A1 CN 2023074151 W CN2023074151 W CN 2023074151W WO 2024159458 A1 WO2024159458 A1 WO 2024159458A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- measurement
- terminal
- measurement result
- training
- working
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/24—Cell structures
- H04W16/28—Cell structures using beam steering
Definitions
- the present application relates to the field of communication technology, and in particular to a training data collection method, device, base station, terminal and storage medium.
- the beam prediction model can be trained by machine learning, and then the beam prediction model can be used to predict the beam direction of the optimal beam through the measurement results of fewer beams. How to obtain training data for training the beam prediction model is an urgent problem to be solved.
- the purpose of the embodiments of the present application is to provide a training data collection method, device, base station, terminal and storage medium to accurately obtain training data for training a beam prediction model.
- the specific technical solution is as follows:
- an embodiment of the present application provides a training data collection method, which is applied to a base station and includes:
- Training data is generated based on the first measurement result, and label data corresponding to the training data is generated based on the second measurement result.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the first measurement timestamp is a start timestamp for the terminal to measure the training beam set
- the second measurement timestamp is a start timestamp for the terminal to measure the working beam set
- the generating of training data based on the first measurement result, and generating label data corresponding to the training data based on the second measurement result includes:
- the measurement data included in the first measurement result is used as training data, and the measurement data included in the second measurement result is used as label data corresponding to the training data.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the first measurement timestamp is a start timestamp for the terminal to measure the training beam set
- the second measurement timestamp is a start timestamp for the terminal to measure the working beam set
- the generating of training data based on the first measurement result, and generating label data corresponding to the training data based on the second measurement result includes:
- a first measurement result and a second measurement result whose absolute value of the difference between the first measurement timestamp and the second measurement timestamp is less than a preset time threshold are combined as a measurement result;
- the measurement data included in the first measurement result in the measurement result combination is used as training data, and the measurement data included in the second measurement result in the measurement result combination is used as label data.
- the method before generating training data based on the first measurement result and generating label data corresponding to the training data based on the second measurement result, the method further includes:
- the logical ID of each beam included in the first measurement result is converted into a physical ID
- the physical ID of each beam included in the second measurement result is converted into a physical ID.
- the logical ID of the beam is converted into a physical ID
- the physical ID of each beam of the base station is determined by the base station based on the elevation angle and azimuth angle of each beam.
- the sending a pilot signal for beam measurement to a terminal in each beam direction of a training beam set, and sending a pilot signal for beam measurement to the terminal in each beam direction of a working beam set includes:
- a cell common pilot signal for beam measurement is sent in each beam direction of the training beam set, and a terminal-specific pilot signal for beam measurement is sent to the terminal in each beam direction of the working beam.
- the first measurement result of the training beam set and the second measurement result of the working beam set sent by the receiving terminal include:
- a first RRC message and a second RRC message sent by the terminal through high-layer signaling are received, where the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate a beam in a training beam set or a working beam set, and the measurement time field is used to indicate a start timestamp of beam measurement;
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set, and the value of the beam number field in the second RRC message is a preset number, which is less than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- the method before sending a pilot signal for beam measurement to a terminal in each beam direction of a training beam set, the method includes:
- the inter-system message includes a first field and a second field, where the first field is used to indicate starting beam measurement, and the second field is used to carry the preset time threshold;
- a first confirmation message replied by the terminal is received, where the first confirmation message is used to indicate that the terminal supports beam measurement of the training beam set.
- the method before sending a pilot signal for beam measurement to the terminal in each beam direction of the working beam set, the method further includes:
- a second confirmation message replied by the terminal is received, where the second confirmation message is used to indicate that the terminal supports beam measurement of the working beam set.
- the first measurement control message and the second measurement control message both include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data of the beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of the preset number of working beams, where the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in order of signal quality from high to low.
- the method further includes:
- the moving speed of the terminal is less than or equal to a preset speed threshold, updating the beams included in the working beam set to beams in a specified range, where the specified range is working beams in a specified elevation angle range and a specified azimuth angle range;
- the specified elevation angle range is the elevation angle range obtained by expanding the elevation angle range of the preset number of working beams according to the first preset step size
- the specified azimuth angle range is the azimuth angle range obtained by expanding the azimuth angle range of the preset number of working beams according to the second preset step size.
- the method further includes:
- the ratio of the smaller one of the first change period and the second change period to the preset coefficient is used as the measurement period after the training beam set and the working beam set are updated.
- an embodiment of the present application provides a training data collection method, which is applied to a terminal and includes:
- a first measurement result of the training beam set and a second measurement result of the working beam set are sent to the base station, wherein the first measurement result is a data source of the training data, and the second measurement result is a data source of the label data of the training data.
- the sending, to the base station, a first measurement result of the training beam set and a second measurement result of the working beam set includes:
- a first RRC message and a second RRC message are sent to the base station through high-layer signaling, where the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate a beam in a training beam set or a working beam set, and the measurement time field is used to indicate a start timestamp of beam measurement;
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set
- the value of the beam number field in the second RRC message is the number of beams included in the training beam set.
- the value of the beam number field in the message is a preset number, and the preset number is smaller than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- the method before performing beam measurement on a pilot signal sent by the base station in each beam direction of the training beam set, the method further includes:
- the method further includes:
- the terminal If the terminal supports beam measurement of the working beam set, it replies a second confirmation message to the base station.
- the first measurement control message and the second measurement control message both include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data for a beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of the preset number of working beams, where the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in order of signal quality from high to low.
- an embodiment of the present application provides a training data collection device, which is applied to a base station, and includes:
- a sending module used to send a pilot signal for beam measurement to the terminal in each beam direction of the training beam set, and send a pilot signal for beam measurement to the terminal in each beam direction of the working beam set;
- a receiving module configured to receive a first measurement result of the training beam set and a second measurement result of the working beam set sent by a terminal;
- a generating module is used to generate training data based on the first measurement result, and to generate label data corresponding to the training data based on the second measurement result.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the first measurement timestamp is a start timestamp for the terminal to measure the training beam set
- the second measurement timestamp is a start timestamp for the terminal to measure the working beam set
- the generating module is specifically used for:
- the measurement data included in the first measurement result is used as training data, and the measurement data included in the second measurement result is used as label data corresponding to the training data.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the first measurement timestamp is a start timestamp for the terminal to measure the training beam set
- the second measurement timestamp is a start timestamp for the terminal to measure the working beam set
- the generating module is specifically used for:
- a first measurement result and a second measurement result whose absolute value of the difference between the first measurement timestamp and the second measurement timestamp is less than a preset time threshold are combined as a measurement result;
- the measurement data included in the first measurement result in the measurement result combination is used as training data, and the measurement data included in the second measurement result in the measurement result combination is used as label data.
- the device further includes:
- a conversion module is used to convert the logical ID of each beam included in the first measurement result into a physical ID, and convert the logical ID of each beam included in the second measurement result into a physical ID.
- the physical ID of each beam of the base station is determined by the base station based on the pitch angle and azimuth angle of each beam.
- the sending module is specifically configured to:
- a cell common pilot signal for beam measurement is sent in each beam direction of the training beam set, and a terminal-specific pilot signal for beam measurement is sent to the terminal in each beam direction of the working beam.
- the receiving module is specifically configured to:
- a first RRC message and a second RRC message sent by the terminal through high-layer signaling are received, where the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate a beam in a training beam set or a working beam set, and the measurement time field is used to indicate a start timestamp of beam measurement;
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set, and the value of the beam number field in the second RRC message is a preset number, which is less than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- the receiving module is further used to receive an inter-system message sent by an OAM device, where the inter-system message includes a first field and a second field, where the first field is used to indicate starting beam measurement, and the second field is used to carry the preset time threshold;
- the sending module is configured to send a first measurement control message to the terminal in response to receiving an inter-system message sent by the OAM device, wherein the first measurement control message is used to instruct the terminal to perform beam measurement on the training beam set at a first specified time-frequency position;
- the receiving module is used to receive a first confirmation message replied by the terminal, where the first confirmation message is used to indicate that the terminal supports beam measurement of the training beam set.
- the sending module is further used to send a second measurement control message to the terminal in response to the inter-system message, where the second measurement control message is used to instruct the terminal to perform beam measurement on the working beam set at a second specified time-frequency position;
- the receiving module is further used to receive a second confirmation message replied by the terminal, where the second confirmation message is used to indicate that the terminal supports beam measurement of the working beam set.
- the first measurement control message and the second measurement control message both include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data for a beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of the preset number of working beams, where the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in order of signal quality from high to low.
- the device further includes:
- An updating module configured to update the beams included in the working beam set to beams within a specified range if the moving speed of the terminal is less than or equal to a preset speed threshold, wherein the specified range is a working beam within a specified elevation angle range and a specified azimuth angle range;
- the specified elevation angle range is the elevation angle range obtained by expanding the elevation angle range of the preset number of working beams according to the first preset step size
- the specified azimuth angle range is the azimuth angle range obtained by expanding the azimuth angle range of the preset number of working beams according to the second preset step size.
- the device further includes:
- a calculation module used for calculating the pitch angle change rate and the azimuth angle change rate of the terminal based on the position of the base station, the position of the terminal and the moving speed of the terminal;
- the calculation module is further used to calculate a first change period of the beam of the base station in the pitch angle dimension and a second change period in the azimuth angle dimension, wherein the first change period is a ratio of the beam width in the pitch angle dimension to the pitch angle change rate, and the second change period is a ratio of the beam width in the azimuth angle dimension to the azimuth angle change rate;
- the updating module is further configured to use a ratio of a smaller one of the first changing period and the second changing period to a preset coefficient as a measurement period after the training beam set and the working beam set are updated.
- an embodiment of the present application provides a training data collection device, which is applied to a terminal and includes:
- a measurement module configured to perform beam measurement on a pilot signal sent by the base station to the terminal in each beam direction of a training beam set, and perform beam measurement on a pilot signal sent by the base station to the terminal in each beam direction of a working beam set;
- a sending module is used to send a first measurement result of the training beam set and a second measurement result of the working beam set to the base station, wherein the first measurement result is a data source of the training data, and the second measurement result is a data source of the label data of the training data.
- the sending module is specifically configured to:
- a first RRC message and a second RRC message are sent to the base station through high-layer signaling, where the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate a beam in a training beam set or a working beam set, and the measurement time field is used to indicate a start timestamp of beam measurement;
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set, and the value of the beam number field in the second RRC message is a preset number, which is smaller than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- the device further includes:
- a receiving module configured to receive a first measurement control message sent by the base station, where the first measurement control message is used to instruct the terminal to perform beam measurement on the training beam set at a first specified time-frequency position;
- the sending module replies a first confirmation message to the base station;
- the receiving module is further used to receive a second measurement control message sent by the base station, where the second measurement control message is used to instruct the terminal to perform beam measurement on the working beam set at a second specified time-frequency position;
- the sending module replies a second confirmation message to the base station.
- the first measurement control message and the second measurement control message both include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data for a beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of the preset number of working beams, where the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in order of signal quality from high to low.
- an embodiment of the present application provides a base station, including: a processor; a 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:
- Training data is generated based on the first measurement result, and label data corresponding to the training data is generated based on the second measurement result.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the first measurement timestamp is a start timestamp for the terminal to measure the training beam set
- the second measurement timestamp is a start timestamp for the terminal to measure the working beam set
- the machine executable instructions further cause the processor to perform the following steps:
- the measurement data included in the first measurement result is used as training data, and the measurement data included in the second measurement result is used as label data corresponding to the training data.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the first measurement timestamp is a start timestamp for the terminal to measure the training beam set
- the second measurement timestamp is a start timestamp for the terminal to measure the working beam set
- the machine executable instructions further cause the processor to perform the following steps:
- a first measurement result and a second measurement result whose absolute value of the difference between the first measurement timestamp and the second measurement timestamp is less than a preset time threshold are combined as a measurement result;
- the measurement data included in the first measurement result in the measurement result combination is used as training data, and the measurement data included in the second measurement result in the measurement result combination is used as label data.
- machine executable instructions further cause the processor to perform the following steps:
- the logical ID of each beam included in the first measurement result is converted into a physical ID
- the logical ID of each beam included in the second measurement result is converted into a physical ID.
- the physical ID of each beam of the base station is determined by the base station based on the elevation angle and azimuth angle of each beam.
- machine executable instructions further cause the processor to perform the following steps:
- the transceiver sends a cell common pilot signal for beam measurement in each beam direction of the training beam set, and sends a terminal-specific pilot signal for beam measurement to the terminal in each beam direction of the working beam.
- machine executable instructions further cause the processor to perform the following steps:
- the transceiver receiving, by the transceiver, a first measurement report message and a second measurement report message sent by the terminal through physical layer signaling, wherein the first measurement report message includes the first measurement result, and the second measurement report message includes the second measurement result;
- a first RRC message and a second RRC message sent by the terminal through high-layer signaling are received through the transceiver, where the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate a beam in a training beam set or a working beam set, and the measurement time field is used to indicate a start timestamp of beam measurement;
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set, and the value of the beam number field in the second RRC message is a preset number, which is less than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- machine executable instructions further cause the processor to perform the following steps:
- an inter-system message sent by an OAM device comprising a first field and a second field, the first field being used to indicate starting beam measurement, and the second field being used to carry the preset time threshold;
- a first confirmation message replied by the terminal is received through the transceiver, where the first confirmation message is used to indicate that the terminal supports beam measurement of the training beam set.
- machine executable instructions further cause the processor to perform the following steps:
- the transceiver receives a second confirmation message replied by the terminal, wherein the second confirmation message is used to indicate that the terminal supports the
- the working beam set is used to perform beam measurement.
- the first measurement control message and the second measurement control message both include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data for a beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of the preset number of working beams, where the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in order of signal quality from high to low.
- machine executable instructions further cause the processor to perform the following steps:
- the moving speed of the terminal is less than or equal to a preset speed threshold, updating the beams included in the working beam set to beams in a specified range, where the specified range is working beams in a specified elevation angle range and a specified azimuth angle range;
- the specified elevation angle range is the elevation angle range obtained by expanding the elevation angle range of the preset number of working beams according to the first preset step size
- the specified azimuth angle range is the azimuth angle range obtained by expanding the azimuth angle range of the preset number of working beams according to the second preset step size.
- machine executable instructions further cause the processor to perform the following steps:
- the ratio of the smaller one of the first change period and the second change period to the preset coefficient is used as the measurement period after the training beam set and the working beam set are updated.
- an embodiment of the present application provides a terminal, including: a processor; a 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:
- a first measurement result of the training beam set and a second measurement result of the working beam set are sent to the base station through the transceiver, wherein the first measurement result is a data source of the training data, and the second measurement result is a data source of the label data of the training data.
- machine executable instructions further cause the processor to perform the following steps:
- a first RRC message and a second RRC message are sent to the base station through the transceiver through high-layer signaling, where the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate the beam in the training beam set or the working beam set, and the measurement time field is used to indicate the beam measurement time.
- the starting timestamp of the quantity is used to indicate the beam in the training beam set or the working beam set, and the measurement time field is used to indicate the beam measurement time.
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set, and the value of the beam number field in the second RRC message is a preset number, which is smaller than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- machine executable instructions further cause the processor to perform the following steps:
- the terminal If the terminal supports beam measurement of the training beam set, replying a first confirmation message to the base station through the transceiver;
- the machine executable instructions further cause the processor to perform the following steps:
- a second confirmation message is replied to the base station through the transceiver.
- the first measurement control message and the second measurement control message both include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data for a beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of the preset number of working beams, where the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in order of signal quality from high to low.
- an embodiment of the present application provides a machine-readable storage medium storing machine-executable instructions.
- the machine-executable instructions When called and executed by a processor, the machine-executable instructions prompt the processor to implement the method steps described in any one of the first aspect or the second aspect.
- an embodiment of the present application provides a computer program product, which prompts the processor to implement the method steps described in any one of the first aspect or the second aspect.
- the base station can send a pilot signal for beam measurement to the terminal in each beam direction of the training beam set and the working beam set, generate training data based on the first measurement result of the terminal on the training beam set, and generate label data corresponding to the training data based on the second measurement result of the terminal on the working beam set.
- the first measurement result and the second measurement result are both obtained by actual measurement of the terminal, so the training data and label data generated by this method are consistent with the actual communication situation between the base station and the terminal, that is, the method can accurately obtain the training data used to train the beam prediction model.
- FIG1 is an architecture diagram of a mobile communication system provided in an embodiment of the present application.
- FIG2 is an exemplary schematic diagram of an AI/ML-based beam prediction structure provided in an embodiment of the present application.
- FIG3 is a flow chart of a training data collection method provided in an embodiment of the present application.
- FIG4 is an exemplary schematic diagram of a mapping relationship between a time-frequency resource of a beam and a logical ID provided in an embodiment of the present application;
- FIG5 is an exemplary schematic diagram of the azimuth angle and elevation angle of a beam provided in an embodiment of the present application.
- FIG6 is an exemplary schematic diagram of a numbering result of a physical ID provided in an embodiment of the present application.
- FIG7 is an exemplary schematic diagram of a method for constructing a training data set during a training data collection process provided by an embodiment of the present application
- FIG8 is a flow chart of another training data collection method provided in an embodiment of the present application.
- FIG9 is a flow chart of a method for adjusting a working beam set during a training data collection process provided by an embodiment of the present application.
- FIG10a is a flow chart of a method for adjusting a measurement period during a training data collection process provided by an embodiment of the present application
- FIG10b is an exemplary schematic diagram of a relative position relationship between a terminal and a base station provided in an embodiment of the present application
- FIG11 is a flow chart of another training data collection method provided in an embodiment of the present application.
- FIG12 is a flow chart of another training data collection method provided in an embodiment of the present application.
- FIG13 is a flowchart of a method for a base station to determine an optimal beam provided by an embodiment of the present application
- FIG14 is a schematic diagram of the structure of a training data acquisition device provided in an embodiment of the present application.
- FIG15 is a schematic diagram of the structure of another training data acquisition device provided in an embodiment of the present application.
- FIG16 is a schematic diagram of the structure of a base station provided in an embodiment of the present application.
- FIG. 17 is a schematic diagram of the structure of a terminal provided in an embodiment of the present application.
- the mobile communication system includes a base station 101 and multiple terminals 102 within the coverage of the base station 101.
- the base station 101 can use multiple beams with different directions to cover the cell it serves.
- the purpose of beam prediction is to determine the optimal transmission beam that can be used by the base station 101 to communicate with each terminal 102.
- the traditional beam prediction method is that the base station performs full-beam scanning, that is, the base station sends a pilot signal to the terminal through a beam in each direction respectively, the terminal measures each pilot signal and reports the measurement result of each pilot signal to the base station, and then the base station selects the optimal transmission beam for the terminal based on the measurement result.
- a beam prediction method based on artificial intelligence/machine learning has been developed. That is, beam prediction is performed using a beam prediction model. By measuring a small number of beams (for example, 16), the better beams among all the working beams (for example, 64) can be predicted, and then the optimal transmitting beam can be selected from the 16 better beams. This can achieve a higher beam prediction effect and significantly reduce the pilot overhead.
- AI/ML artificial intelligence/machine learning
- Figure 2 is an exemplary schematic diagram of an AI/ML-based beam prediction structure provided in an embodiment of the present application, including a data acquisition module 201, a model training module 202, a model reasoning module 203 and an execution module 204.
- the data collection module 201 may be deployed in a base station, or part of its functions may be deployed in a base station and part of its functions may be deployed in other network devices.
- the model training module 202 may be deployed in a base station or in other network devices.
- the model reasoning module 203 and the execution module 204 are both deployed in the base station.
- the data acquisition module 201 is used to collect training data, label data corresponding to the training data and inference data, and to identify, clean and normalize the training data and inference data so that the format and quality of the processed training data and inference data meet the requirements of model training.
- Both the training data and the inference data can be the Layer 1 Reference Signal Receiving Power (L1-RSRP) of each beam, and L1-RSRP can be used to represent the physical layer pilot signal strength.
- L1-RSRP Layer 1 Reference Signal Receiving Power
- the training data and the label data corresponding to the training data are used to train the beam prediction model, and the inference data is the data that needs to be input into the beam prediction model after the trained beam prediction model is deployed.
- the data acquisition module 201 can input the training data and the label data corresponding to the training data into the model training module 202, and the model training module 202 is used to perform model training according to the training data sent by the data acquisition module 201.
- a machine learning algorithm such as a deep neural network (DNN) or a long short-term memory network (LSTM) can be used for training to obtain a beam prediction model.
- the model training module 202 can input the training data into the model, obtain the output result of the model, calculate the loss function value based on the output result 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, and the trained model is used as the beam prediction model.
- DNN deep neural network
- LSTM long short-term memory network
- model training module 202 can perform model deployment in the model reasoning module 203 , that is, deploy the trained beam prediction model in the model reasoning module 203 .
- the data acquisition module 201 can input the inference data into the model inference module 203, and the model inference module 203 can process the inference data using the beam prediction model, thereby obtaining the inference result output by the beam prediction model, and sending the inference result to the execution module 204.
- the inference result can be the beam predicted by the beam prediction model.
- the execution module 204 is a module in the base station used for communicating with the terminal.
- the execution module 204 is used to send a signal to the terminal using a transmit beam and receive a signal sent by the terminal using a receive beam.
- the execution module 204 may send a signal to the terminal according to the transmit beam predicted by the beam prediction model.
- the execution module 204 may also use part of the transmit beam to transmit a pilot signal to the terminal, and receive the L1-RSRP corresponding to each transmit beam fed back by the terminal. Then, the execution module 204 may feed back data to the data acquisition module 201, that is, feed back the L1-RSRP corresponding to each transmit beam fed back by the terminal to the data acquisition module 201.
- the data acquisition module may use the obtained L1-RSRP corresponding to each transmit beam as inference data.
- the model reasoning module 203 can also provide model performance feedback to the model training module 202, specifically, the accuracy of the beam predicted by the beam prediction model. If the accuracy is less than a preset accuracy threshold, the model training module 202 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 203 based on the optimized parameters.
- an embodiment of the present application provides a training data collection method, which is applied to a base station. As shown in FIG3 , the method includes:
- the working beam set can include all working beams, and the full working beams refer to beams that work normally in the base station.
- the working beam set includes 64 working beams.
- the training beam set includes beams used for beam prediction, and the number of beams included in the training beam set is the same as the number of training data that needs to be input into the beam prediction model at one time.
- the beam prediction model can use the L1-RSRP of the 16 beams in the training beam set to predict the better beam among all working beams.
- S302 Receive a first measurement result of a training beam set and a second measurement result of a working beam set sent by a terminal.
- the terminal may measure each pilot signal sent by the base station, and use the measurement result of the pilot signal as the measurement result of the beam corresponding to the pilot signal.
- the first measurement result includes the measurement result of the terminal on each beam in the training beam set
- the second measurement result includes the measurement result of the terminal on a preset number of working beams ranked before signal quality in the working beam set.
- the preset number is the same as the number of beams predicted by the beam prediction model. If the beam prediction model is used to predict five better beams, that is, the five beams with the top signal quality rankings, then the preset number is five.
- S303 Generate training data based on the first measurement result, and generate label data corresponding to the training data based on the second measurement result.
- the base station can send a pilot signal for beam measurement to the terminal in each beam direction of the training beam set and the working beam set, generate training data based on the first measurement result of the terminal on the training beam set, and generate label data corresponding to the training data based on the second measurement result of the terminal on the working beam set.
- the first measurement result and the second measurement result are both obtained by actual measurement of the terminal, so the training data and label data generated by this method are consistent with the actual communication situation between the base station and the terminal, that is, the method can accurately obtain the training data used to train the beam prediction model.
- sending a pilot signal for beam measurement to the terminal in each beam direction of the training beam set, and sending a pilot signal for beam measurement to the terminal in each beam direction of the working beam set specifically includes the following two implementation methods:
- Method 1 Send a terminal-specific pilot signal for beam measurement to the terminal in each beam direction of the training beam set, and send a terminal-specific pilot signal for beam measurement to the terminal in each beam direction of the working beam.
- the base station can send pilot signals for beam measurement to multiple terminals according to the pre-configured time-frequency position. That is, the base station can send terminal-specific pilot signals to each terminal in each beam direction of the training beam set, and send terminal-specific pilot signals to each terminal in each beam direction of the working beam. Furthermore, each terminal that receives the terminal-specific pilot signal can feed back the first measurement result and the second measurement result to the base station. In this way, the terminal can more accurately measure each beam of the training beam set and the working beam through the terminal-specific pilot signal.
- Method 2 Send a cell common pilot signal for beam measurement in each beam direction of the training beam set, and send a terminal dedicated pilot signal for beam measurement to the terminal in each beam direction of the working beam.
- the cell common pilot signal is a pilot signal that can be received by all terminals in the base station coverage area.
- the cell common pilot signal can be a synchronization broadcast block signal (Synchronization Signal Block, SSB).
- each terminal After the base station sends the cell common pilot signal in each beam direction of the training beam set, all terminals within the coverage of the base station can feedback the first measurement result for each cell common pilot signal. Similarly, after the base station sends a terminal-specific pilot signal to each terminal in each beam direction of the working beam, each terminal can feedback the second measurement result corresponding to its own terminal-specific pilot signal. In this way, sending a cell common pilot signal to the terminal can save pilot overhead and improve measurement efficiency compared to sending a terminal-specific pilot signal to each terminal separately.
- S302 receiving a first measurement result of a training beam set and a second measurement result of a working beam set sent by a terminal, specifically includes the following two implementation methods:
- Method 1 receiving a first measurement report message and a second measurement report message sent by a terminal through physical layer signaling.
- the physical layer signaling is specifically layer 1 (Layer 1, L1) signaling, the first measurement report message includes the first measurement result, and the second measurement report message includes the second measurement result.
- the formats of the first measurement report message and the second measurement report message comply with the provisions of the Radio Resource Control (RRC) protocol, which will not be repeated here.
- RRC Radio Resource Control
- the first measurement result may specifically include a first measurement timestamp and measurement data of the terminal on each beam in the training beam set.
- the first measurement timestamp is a start timestamp of the terminal measuring the training beam set.
- the second measurement result may specifically include a second measurement timestamp and the measurement result of the terminal on a preset number of working beams in the working beam set.
- the second measurement timestamp is the starting timestamp of the terminal's measurement of the working beam set.
- the terminal can periodically measure the training beam set and the working beam set. After the terminal performs beam measurement on the training beam set in each period, it will feedback a first measurement result to the base station once. The first measurement timestamp included in the first measurement result is the starting timestamp of the measurement of the training beam set in the period. Similarly, after the terminal performs beam measurement on the working beam set in each period, it will feedback a second measurement result to the base station once. The second measurement timestamp included in the second measurement result is the starting timestamp of the measurement of the working beam set in the period.
- the transmission delay of the first measurement report message and the second measurement report message transmitted by physical layer signaling is low, which can make the beam measurement more timely.
- Method 2 receiving a first RRC message and a second RRC message sent by a terminal through high-layer signaling, wherein the first RRC message includes a first measurement result, and the second RRC message includes a second measurement result.
- the measurement results can be transmitted by means of physical layer signaling with low latency.
- the training data in the embodiments of the present application are used to train the beam prediction model, and the latency requirements are not high. Therefore, both the first measurement result and the second measurement result can be transmitted by high-level signaling. This can reduce the capacity requirements for the physical layer control channel, and high-level signaling has high flexibility and scalability.
- the RRC message format of high-level signaling transmission can be set according to the requirements of the training data. The use of high-level signaling helps to improve the performance of the beam prediction algorithm.
- the embodiments of the present application define the format of an RRC message transmitted using high-layer signaling.
- the message formats of the first RRC message and the second RRC message are introduced below.
- Both the first RRC message and the second RRC message include a message type field, a measurement time field, a beam number field, a logical identity (Identity, ID) of each beam, and measurement data.
- the message type field is used to indicate the type of beam being measured, that is, to indicate the beams included in the training beam set or the working beam set. For example, if the value of the message type field is 1, it indicates that the measurement results of the beams included in the training beam set are carried; if the value of the message type field is 2, it indicates that the measurement results of the beams included in the working beam set are carried.
- the value of the message type field in the first RRC message is 1, and the value of the message type field in the second RRC message is 2.
- the measurement time field is used to indicate the starting timestamp of the beam measurement.
- the starting timestamp is the starting timestamp when the terminal starts measuring the training beam set or the working beam set.
- the starting timestamp can be represented by the system frame number (SFN) or the coordinated universal time (UTC).
- the beam number field in the first RRC message is used to indicate the reported number of beams included in the training beam set this time.
- the value of the beam number field is the number of beams included in the training beam set. Accordingly, the first RRC message carries the logical ID and measurement data of each beam in the training beam set.
- the beam number field in the second RRC message is used to indicate the number of working beams reported this time, and the value of the beam number field is a preset number. Accordingly, the second RRC message carries the logical IDs and measurement data of the preset number of beams in the working beam set.
- the preset number is less than the number of beams included in the working beam set, and the preset number is the same as the number of beams predicted by the beam prediction model. If the beam prediction model is used to predict 5 better beams, the preset number is 5.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field.
- the terminal location field is used to indicate the geographical location of the terminal.
- the terminal can obtain its own geographical location through Global Navigation Satellite System (GNSS) positioning or other positioning methods.
- GNSS Global Navigation Satellite System
- the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- the terminal speed field in the first RRC message is used to indicate the average moving speed of the terminal during the measurement of each beam in the training beam set
- the terminal speed field in the second RRC message is used to indicate the average moving speed of the terminal during the measurement of each beam in the training beam set.
- the field is used to indicate the average moving speed of the terminal during the measurement of each beam in the working beam set.
- the moving speed includes the moving speed of the terminal in both the horizontal and vertical directions.
- the first measurement result includes a first measurement timestamp and measurement data
- the second measurement result includes a second measurement timestamp and measurement data.
- the first measurement timestamp is the starting timestamp of the beam measurement of the training beam set, for example, the first measurement timestamp is specifically the value of the measurement time field carried in the first RRC message.
- the second measurement timestamp is the starting timestamp of the beam measurement of the working beam set, for example, the second measurement timestamp is specifically the value of the measurement time field carried in the second RRC message.
- the measurement data included in the first measurement result is the logical ID and signal quality of each beam included in the training beam set.
- the signal quality can be specifically represented by L1-RSRP.
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of a preset number of working beams, wherein the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in descending order of signal quality.
- the second measurement result includes the logical ID and signal quality of the working beam with signal quality TOP-K, or the second measurement result includes the logical ID of the working beam with signal quality TOP-K.
- K is a preset number. Assuming the preset number is 5, the second measurement result includes the logical IDs and signal qualities of the top 5 working beams in signal quality among the 64 working beams, or includes the logical IDs of the top 5 working beams in signal quality.
- the second measurement result includes the signal quality depends on whether the output of the beam prediction model includes the signal quality. If the output of the beam prediction model is the IDs and signal qualities of a preset number of beams, the second measurement result includes the logical IDs and signal qualities of the preset number of working beams; if the output of the beam prediction model is the IDs of a preset number of beams, the second measurement result includes the logical IDs of the preset number of working beams.
- the logical ID of the beam can also be called the index number of the beam. It is different, so different base stations may set different logical IDs for beams in the same beam direction.
- the horizontal axis on the left side of FIG4 represents time
- the vertical axis represents frequency
- the filled positions of the four columns from left to right represent the time-frequency resources of the four beams, respectively.
- the physical IDs of the four beams are index 1, index 2, index 3, and index 4.
- the time-frequency resources occupied by each beam on the right side of FIG4 are the time-frequency resources represented by the same filled positions on the left side of FIG4 .
- the logical IDs of the four beams of base station 1 are index 1, index 2, index 3, and index 4, respectively, but the logical IDs of the four beams in base station 2 are index 3, index 4, index 1, and index 2, respectively.
- the beam corresponding to index 1 in base station 1 has a different direction from the beam corresponding to index 1 in base station 2, and the beam directions corresponding to index 2 in base station 1 and base station 2 are different, and the same is true for index 3 and index 4.
- an embodiment of the present application provides a method for matching the logical ID of the beam with the real physical beam, that is, before the base station generates training data based on the first measurement result and generates label data corresponding to the training data based on the second measurement result, the base station can convert the logical ID of each beam included in the first measurement result into a physical ID, and convert the logical ID of each beam included in the second measurement result into a physical ID.
- each beam of the base station is determined based on the elevation angle and azimuth angle of the beam. Specifically, each beam can be numbered based on the elevation angle and azimuth angle in the Cartesian coordinate system, and the number is used as the physical ID of the beam.
- the elevation angle is the angle between the beam pointing and the horizontal plane
- the azimuth angle is the angle formed by rotating clockwise from the north direction to the beam direction.
- FIG5 is an exemplary schematic diagram of an azimuth angle and elevation angle of a beam direction provided in an embodiment of the present application.
- the ellipse in FIG5 is a beam of the antenna array, the angle between the beam direction and the horizontal plane is the elevation angle ⁇ , and the angle between the true north direction and the beam direction is the azimuth angle
- the base station can number the beams in the beam set in the order of elevation angle first and azimuth angle second according to the beam direction of each beam in the beam set. That is, the beams in each beam direction are first sorted in the order of elevation angle from small to large, and then for beams with the same elevation angle, the azimuth angle is sorted in the order of small to large.
- the base station can number the beams in order according to the sorting results.
- the horizontal axis in FIG6 is the azimuth angle
- the vertical axis is the elevation angle
- the units of the azimuth angle and the elevation angle are both degrees.
- Each shaded rectangular area in FIG6 represents a beam direction
- the elevation angles of the four beams in the bottom row in FIG6 are the same, and the azimuth angles increase from left to right, so the corresponding numbers are 0, 1, 2, and 3, respectively.
- the elevation angles of the beams in the middle row are greater than the elevation angles of the beams in the bottom row, and the azimuth angles increase from left to right, so the corresponding numbers are 4, 5, 6, and 7, respectively.
- the corresponding numbers of the four beams in the top row are 8, 9, 10, and 11, respectively.
- the beams in the beam set can be divided into 360*360 numbers.
- the base station can map the beam to a number according to the pitch angle and azimuth angle of the beam that needs to be converted, and use the number as the physical ID of the beam.
- the base station can determine the elevation angle and azimuth angle of the beam corresponding to the logical ID, and then use the numbers corresponding to the elevation angle and azimuth angle as the physical ID of the beam, thereby replacing the logical ID of the beam with the physical ID.
- the base station can use the same method to replace the logical ID of the working beam with the physical ID.
- the physical ID of the beam is determined based on the elevation angle and azimuth angle of the beam, the physical IDs determined by different base stations for beams in the same beam direction are the same, which can make the beam directions corresponding to the same physical ID in the training data generated by subsequent base stations the same.
- a unified data format is provided to better support the joint training of multi-scenario large models and the joint optimization of multi-vendor data. change.
- the base station can send pilot signals for beam measurement in each beam direction of the training beam set and the working beam set multiple times in a periodic, semi-periodic or non-periodic manner, and accordingly, the terminal will also feedback the first measurement report and the second measurement report multiple times.
- the base station needs to ensure that the time difference between the first measurement report and the second measurement report used to generate a set of training data and label data is within a certain range.
- the above S303, generating training data based on the first measurement result, and generating label data corresponding to the training data based on the second measurement result can be specifically implemented as follows:
- the measurement data included in the first measurement result is used as training data, and the measurement data included in the second measurement result is used as label data corresponding to the training data.
- the base station After receiving the first measurement result and the second measurement result fed back by the same terminal, the base station can determine whether the absolute value of the difference between the first measurement timestamp carried by the first measurement result and the second timestamp carried by the second measurement result is less than a preset time threshold. If so, the first measurement result and the second measurement result can be combined as a measurement result, and the measurement data included in the first measurement result can be used as training data, and the measurement data included in the second measurement result can be used as label data corresponding to the training data.
- the preset time threshold in the embodiment of the present application is a value preset based on experience, which is used to indicate whether the second measurement result can be used as label data of training data in the first measurement result.
- the value of the preset time threshold is affected by the beam change speed, terminal movement speed and environmental changes, and can be flexibly configured manually.
- the base station may also generate training data and label data in batches, that is, the above S303, generating training data based on the first measurement result, and generating label data corresponding to the training data based on the second measurement result, may be specifically implemented as follows:
- a first measurement result and a second measurement result whose absolute value of the difference between the first measurement timestamp and the second measurement timestamp is less than a preset time threshold are combined as a measurement result;
- the measurement data included in the first measurement result in the measurement result combination is used as training data, and the measurement data included in the second measurement result in the measurement result combination is used as label data.
- the first measurement result and the second measurement result are both from the same terminal.
- training data and label data can be generated according to the method.
- the base station may generate training data and label data in batches after receiving multiple first measurement results and multiple second measurement results from the same terminal.
- the base station may perform the operation of batch generating training data and label data once at fixed intervals; or perform the operation of batch generating training data and label data once each time the number of received first measurement results and second measurement results reaches a fixed number.
- the base station may first convert the logical IDs included in the first measurement result and the second measurement result into physical IDs, and then generate training data and label data.
- FIG. 7 is an exemplary schematic diagram of a method for constructing a training data set during a training data collection process provided in an embodiment of the present application.
- Table A in FIG7 is a plurality of first measurement results, that is, a plurality of measurement results for a training beam set.
- each row in Table A represents a first measurement result
- each first measurement result includes a first measurement timestamp of a training beam set and measurement data of the training beam set
- the measurement data of the training beam set includes measurement data of a plurality of beams in the training beam set
- the measurement data of each beam of the training beam set can be identified by the logical ID of the beam
- the measurement data can specifically be L1-RSRP.
- logical IDs and measurement data of multiple beams in a training beam set are simply represented as "logical ID/measurement data".
- a first measurement timestamp corresponds to the logical ID and measurement data of each beam in the training beam set. For example, if the training beam set includes 4 beams, namely beams 1-4, then the measurement data corresponding to a first measurement timestamp includes: beam 1/ Measurement data 1, beam 2/measurement data 2, beam 3/measurement data 3, and beam 4/measurement data 4.
- Table B in Figure 7 is a plurality of second measurement results, that is, a plurality of measurement results for the working beam set.
- each row in Table B represents a second measurement result
- each second measurement result includes a second measurement timestamp of a working beam set and measurement data of the working beam set
- the measurement data of the working beam set includes measurement data of a preset number of working beams
- the measurement data of each working beam can be identified by the logical ID of the working beam.
- one second measurement timestamp corresponds to the logical ID and measurement data of each working beam in the working beam set.
- the working beam set includes 5 working beams, namely beams 7, 10, 11, 14, and 20
- the measurement data corresponding to one second measurement timestamp include: beam 7/measurement data 7, beam 10/measurement data 10, beam 11/measurement data 11, beam 14/measurement data 14, and beam 20/measurement data 20.
- T11, T12 to T1N in Table A and Table C of FIG. 7 represent N first measurement timestamps, and T21, T22 to T2M in Table B and Table D represent M second measurement timestamps.
- the base station may convert each logical ID in Table A into a physical ID, so that Table A is converted into Table C; and may convert each logical ID in Table B into a physical ID, so that Table B is converted into Table D.
- the base station may search Table C and Table D for a first measurement result and a second measurement result in which the absolute value of the difference between the first measurement timestamp and the second measurement timestamp is less than a preset time threshold.
- the first measurement timestamp may be represented as T1i
- the second measurement timestamp may be represented as T2j
- the preset time threshold may be represented as ReportTgap (reporting time threshold)
- the base station may search for a measurement result combination satisfying
- the measurement timestamp in Table E of FIG. 7 is the first measurement timestamp of each training data in Table C, and the training data and label data in the same row in Table E come from the same measurement result combination.
- measurement timestamps in Table E of FIG. 7 are used to identify different groups of training data and label data.
- other forms of IDs may also be used to distinguish different groups of training data and label data, and the embodiment of the present application is not limited to this.
- the base station can report the training data set to the operation administration and maintenance (OAM) equipment.
- the reporting format can be [measurement timestamp, training data, label data].
- the measurement timestamp can also be replaced by other forms of ID.
- Each base station reports the training data set to the OAM device in a unified format, which can make the training data set more universal.
- OAM does not need to perform operations such as format unification on the received training data set, which can improve the efficiency of OAM training beam prediction model.
- the OAM device may be used to trigger the base station to start the beam measurement process in the above embodiment. Based on this, before the base station sends a pilot signal for beam measurement to the terminal in each beam direction of the training beam set, the method includes:
- An inter-system message sent by an OAM device is received, where the inter-system message includes a first field and a second field, where the first field is used to indicate starting beam measurement, and the second field is used to carry a preset time threshold.
- the preset time threshold is pre-configured on the OAM device manually based on experience. After the manual configuration is completed, the OAM device can be manually triggered to send the inter-system message to the base station, or the OAM device can trigger the operation of sending the inter-system message to the base station by itself.
- the inter-system message sent by the OAM device can be in the RESTful HyperText Transfer Protocol POST (HTTP POST) message format.
- HTTP POST HyperText Transfer Protocol
- RESTful is a design style and development method for network applications.
- the notifyBeamTraining field in the message can be used to indicate the start of the above beam measurement process, and the ReportTgap The field indicates the preset time threshold.
- Table 2 The specific configuration is shown in Table 2:
- the "S" in the last column of Table 2 indicates the optional type of the parameter.
- M it indicates that the corresponding parameter is a required parameter.
- O it indicates that the corresponding parameter is an optional parameter.
- the base station After receiving the above inter-system message, the base station sends a first measurement control message to the terminal in response to the inter-system message, where the first measurement control message is used to instruct the terminal to perform beam measurement on the training beam set at the first specified time-frequency position. Then, a first confirmation message replied by the terminal is received, where the first confirmation message is used to indicate that the terminal supports beam measurement on the training beam set.
- the beam measurement in the embodiment of the present application is used to collect training data.
- a preset indication parameter can be added to the first measurement control message. That is, the first measurement control message includes a preset indication parameter.
- the preset indication parameter is used to indicate the collection of training data for the beam prediction model, that is, the first measurement control message instructs the base station to perform beam measurement and use the beam measurement result as the training data for the beam prediction model.
- the first measurement control message may be carried in a channel state information measurement configuration (CSI-MeasConfig) in an RRC message of the 3rd Generation Partnership Project (3GPP).
- the indication parameter added in the CSI-MeasConfig may be a Boolean variable.
- the name of the Boolean variable may specifically be ML-TrainingData (machine learning training data). When the value of ML-TrainingData is True, it indicates that this measurement is used for collecting training data for machine learning.
- the terminal when the terminal recognizes that the value of the indication field of the first measurement control message is True, it can determine whether it supports measuring the training data used for the beam prediction model. If supported, it will reply with a first confirmation message to the base station, and then the base station can send pilot signals for beam measurement in each beam direction of the working beam set according to the specified time-frequency resources.
- the method further includes:
- a second measurement control message is sent to the terminal, where the second measurement control message is used to instruct the terminal to perform beam measurement on the working beam set at a second specified time-frequency position.
- the receiving terminal replies with a second confirmation message, where the second confirmation message is used to indicate that the terminal supports beam measurement of the working beam set.
- the second measurement control message also includes a preset indication parameter.
- the preset indication parameter When the value of the preset indication parameter is the first value, the preset indication parameter The number is used to indicate the training data of the acquisition beam prediction model. As an example, the first value may be "True".
- the second measurement control message has the same structure as the first measurement control message, and reference may be made to the relevant description of the first measurement control message in the above embodiment.
- the base station can send a pilot signal for beam measurement in each beam direction of the working beam set according to the second specified time-frequency position.
- the base station sends pilot signals in each beam direction of the training beam set and in each beam direction of the working beam set.
- the order of sending depends on the time and frequency resources of the training beam set and the working beam set.
- the scale of the working beam set can be dynamically adjusted during the training data acquisition process, and the measurement period of the training beam set and the working beam set can be dynamically adjusted.
- the training data acquisition method provided by the embodiment of the present application is shown in FIG8 , which is illustrated by taking the communication between the base station and the terminal 1 and the terminal 2 as an example, and the method includes:
- the base station obtains the measurement result of terminal 1 on the training beam set and the working beam set.
- the base station obtains the measurement result of terminal 2 on the training beam set and the working beam set.
- the method for the base station to obtain the measurement results of the training beam set and the working beam set fed back by each terminal can refer to the relevant introduction in the above embodiment.
- the base station may adjust the number of working beams that terminal 1 needs to measure and the measurement period of the training beam set and the working beam set based on the measurement result reported by terminal 1.
- the base station may adjust the number of working beams that terminal 2 needs to measure and the measurement period of the training beam set and the working beam set based on the measurement result reported by terminal 2.
- FIG8 takes reducing the number of working beams that terminal 1 needs to measure, extending the measurement period of terminal 1 for the training beam set and the working beam set, increasing the number of working beams that terminal 2 needs to measure, and reducing the measurement period of terminal 2 for the training beam set and the working beam set as an example for explanation.
- S804 Configure terminal 1 to measure fewer working beams, and extend the measurement period of terminal 1 for the training beam set and the working beam set.
- the black rectangle under terminal 1 represents one measurement cycle configured for terminal 1
- the three black rectangles under terminal 2 represent three measurement cycles configured for terminal 2. It can be seen that the duration of a single measurement cycle configured for terminal 1 is longer, and the duration of a single measurement cycle configured for terminal 2 is shorter.
- Terminal 2 reports the measurement result after each measurement cycle ends.
- Terminal 1 reports the measurement result after each measurement cycle ends.
- the following introduces a method in which the base station dynamically adjusts the number of working beams that the terminal needs to measure during the training data collection process and a method in which the base station dynamically adjusts the measurement period of the terminal for the working beam set and the training beam set.
- the base station when the base station measures each beam in the working beam set for the first time, the working beam set includes all working beams. After receiving the first measurement result of the training beam set and the second measurement result of the working beam set sent by the terminal, the base station can update the beams included in the working beam set, and the specific method is:
- the beams included in the working beam set are updated to beams within a specified range, where the specified range is the working beams within a specified elevation angle range and a specified azimuth angle range.
- the specified pitch angle range is the pitch angle range obtained by expanding the pitch angle range of a preset number of working beams according to a first preset step size
- the specified azimuth angle range is the azimuth angle range obtained by expanding the azimuth angle range of a preset number of working beams according to a second preset step size.
- the method may specifically include the following steps:
- S902 Obtain a second measurement result of the terminal for the working beam set and a moving speed of the terminal.
- the second measurement result includes measurement data of a preset number of working beams
- the pitch angle range ⁇ corresponding to the second measurement result refers to the pitch angle range of the preset number of working beams
- the azimuth angle range refers to the azimuth range of a preset number of working beams.
- the moving speed threshold value Vth may be configured by the base station according to the environment where the base station is located and the performance requirements of the base station.
- the working beam set can be updated to include ⁇ + ⁇ step and Working beam within range.
- ⁇ step and are the first preset step length and the second preset step length, respectively, and can be configured by the base station according to the environment and the performance requirements of the base station. For example, if the base station is indoors, due to the large number of indoor obstructions, the base station can set ⁇ step and Set a larger value; if the base station is outdoors, since there are fewer obstructions that can block the beam outdoors, the base station can set ⁇ step and Set to a smaller value.
- the base station can set ⁇ step and Set to a larger value so that the base station can obtain more working beam measurement results to improve the performance of the beam prediction model; if the base station performance requirements are lower, the base station can set ⁇ step and Set to a smaller value to reduce pilot overhead.
- the working beam set is not updated.
- the base station can flexibly adjust the working beam set according to the moving speed of the terminal and the pitch angle range and azimuth angle range corresponding to the preset number of working beams. Since the preset number of working beams are beams with better signal quality measured by the terminal, when the terminal moves slowly, it is highly likely that the working beam with the best signal quality measured by the terminal next time is still within the above pitch angle and azimuth angle range. Therefore, through the above adjustment, unnecessary measurement overhead can be reduced on the basis of ensuring measurement accuracy.
- the following introduces a method for adjusting the measurement period of the training beam set and the working beam set during the training data collection process. After receiving the first measurement result of the training beam set and the second measurement result of the working beam set sent by the terminal, the method further includes:
- the pitch angle change rate and azimuth angle change rate of the terminal are calculated, and then the first change period of the base station beam in the pitch angle dimension and the second change period in the azimuth dimension are calculated, and then the smaller one of the first change period and the second change period and the ratio of the preset coefficient is used as the measurement period after the training beam set and the working beam set are updated.
- the first change period is the ratio of the beam width in the pitch angle dimension to the pitch angle change rate
- the second change period is the ratio of the beam width in the azimuth angle dimension to the azimuth angle change rate
- the method specifically comprises the following steps:
- the geometric position relationship between the base station and the terminal can be determined. Based on the geometric position relationship and the moving speed of the terminal, the pitch angle change rate ⁇ and the azimuth angle change rate can be calculated using the existing mathematical algorithm.
- the left side of FIG10b is the base station and the right side is the terminal.
- the position of the base station is (x, y) and the height of the base station is h.
- the position of the terminal at time t1 is ( x1 , y1 ) and the height of the terminal is h1 .
- the pitch angle of the terminal is ⁇ 1 .
- ⁇ 1 can be calculated by the following formula:
- the pitch angle of the terminal is ⁇ 1, which can be calculated by the following formula:
- the pitch angle ⁇ 2 of the terminal at time t2 can be calculated.
- the pitch angle ⁇ 2 of the terminal can be calculated by the following formula:
- the elevation angle ⁇ 2 of the terminal can be calculated by the following formula:
- the base station can calculate the pitch angle change rate ⁇ ' based on the calculated ⁇ 1 and ⁇ 2 :
- ⁇ t is the difference between t2 and t1 .
- the base station calculates the azimuth change rate of the terminal Still taking FIG. 10b as an example, the position of the base station is represented by (x, y), and the height of the base station is h; the position of the terminal at time t1 is ( x1 , y1 ), and the height of the terminal is h1 , then the azimuth of the terminal at time t1 is It can be calculated by the following formula:
- the position of the terminal at time t 2 is (x 2 , y 2 ), and the height of the terminal is h 2 , then the azimuth of the terminal at time t 2 is It can be calculated by the following formula:
- the base station calculates and The azimuth change rate can be calculated
- ⁇ t is the difference between t2 and t1 .
- the pitch angle change rate of the terminal is 0, and the angular velocity of the terminal may be equal to the azimuth angle change rate of the terminal.
- the beam width ⁇ max of the base station beam in the pitch angle dimension is the angle between the two half-power points of the main lobe of each working beam in the pitch direction
- the beam width of the base station beam in the azimuth dimension is It is the angle between the two half-power angles of the main lobe of each working beam in the azimuth direction.
- S1007 Determine the measurement period of the training beam set and the working beam set as T/m.
- the base station can configure a preset coefficient m according to the environment and performance requirements. For example, when the performance requirements are high, m can be configured to a larger value to make the final measurement period smaller and the measurement more precise.
- the base station can configure the updated measurement period in the channel state information reporting configuration (Channel State Information-Report configuration, CSI-ReportConfig) of the RRC message of 3GPP, and specifically can be configured in the period and offset of the channel state information reporting (Channel State Information-ReportPeriodicityAndOffset, CSI-ReportPeriodicit yAndOffset).
- the channel state information reporting configuration Channel State Information-Report configuration, CSI-ReportConfig
- CSI-ReportConfig Channel State Information-ReportConfig
- the first change period and the second change period calculated by the base station can reflect the speed of beam change, determine the smaller value of the first change period and the second change period, and use the ratio of the smaller value of the determined first change period and the second change period to the preset coefficient as the measurement period.
- the calculated measurement period is more precise and controllable, and more in line with actual measurement needs. It can reduce measurement overhead while ensuring that the measured measurement results are accurate and rich, and can avoid collecting a large amount of training data with small differences, which can improve the quality of subsequently generated training data.
- the embodiment of the present application also provides a training data collection method, which is applied to a terminal, as shown in FIG11 , and includes:
- S1102. Send a first measurement result of a training beam set and a second measurement result of a working beam set to a base station.
- the first measurement result is the data source of the training data
- the second measurement result is the data source of the label data of the training data
- the terminal performs beam measurement on the pilot signal sent by the base station in each beam direction of the training beam set and the pilot signal sent in each beam direction of the working beam set, and the terminal feeds back the first measurement result of the training beam set and the second measurement result of the working beam set to the base station.
- the first measurement result and the second measurement result are both obtained by actual measurement of the terminal, so the training data and label data generated by this method are consistent with the actual communication situation between the base station and the terminal, that is, the method can accurately obtain the training data used to train the beam prediction model.
- the method before performing beam measurement on the pilot signal sent by the base station in each beam direction of the training beam set, the method further includes:
- a first measurement control message sent by the base station is received. If the terminal supports beam measurement of the training beam set, a first confirmation message is replied to the base station.
- the first measurement control message is used to instruct the terminal to perform beam measurement on the training beam set at a first specified time-frequency position.
- the method before performing beam measurement on the pilot signal sent by the base station in each beam direction of the working beam set, the method further includes:
- a second measurement control message sent by the base station is received, where the second measurement control message is used to instruct the terminal to perform beam measurement on the working beam set at a second specified time-frequency position. If the terminal supports beam measurement on the working beam set, a second confirmation message is replied to the base station.
- both the first measurement control message and the second measurement control message include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data of the beam prediction model.
- the terminal After the terminal completes the beam measurement, it needs to send the measurement result to the base station.
- the terminal sends the first measurement result of the training beam set to the base station. and a second measurement result of the working beam set, including the following two implementation methods:
- Method 1 sending a first measurement report message and a second measurement report message to the base station through physical layer signaling, where the first measurement report message includes the first measurement result, and the second measurement report message includes the second measurement result.
- Method 2 sending a first RRC message and a second RRC message to the base station through high-layer signaling, wherein the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the format of the first RRC message and the second RRC message and the specific contents included can be referred to the relevant description in the above embodiment, and will not be repeated here.
- the method includes:
- the OAM device sends an inter-system message to a base station.
- the inter-system message is used to configure the measurement parameters of the beam prediction model.
- the base station sends a first measurement control message to the terminal.
- the first measurement control message is used to instruct the terminal to perform beam measurement on the training beam set at the first specified time-frequency position.
- first measurement control message reference may be made to the description in the above embodiment, which will not be repeated here.
- the terminal feeds back a first confirmation message to the base station.
- the first confirmation message is used to indicate that the terminal supports beam measurement of the training beam set.
- the embodiment of the present application does not specifically limit the form of the first confirmation message.
- the base station sends a pilot signal for beam measurement to the terminal in each beam direction of the training beam set.
- S1205 The terminal measures the training beam set.
- the terminal feeds back the first measurement result to the base station.
- the terminal can feedback the first measurement result to the base station through a first measurement report message sent by physical layer signaling, and can also feedback the first measurement result through a first RRC message sent by high-level signaling.
- the message format of the first measurement report message and the first RRC can refer to the relevant description in the above embodiments, and will not be repeated here.
- the base station sends a second measurement control message to the terminal.
- the second measurement control message is used to instruct the terminal to perform beam measurement on the working beam set at a second specified time-frequency position, and the message format of the second measurement control message is the same as that of the first measurement control message.
- the terminal feeds back a second confirmation message to the base station.
- the second confirmation message is used to indicate that the terminal supports beam measurement of the working beam set.
- the embodiment of the present application does not specifically limit the form of the second confirmation message.
- the base station sends a pilot signal for beam measurement to the terminal in each beam direction of the working beam set.
- the terminal measures the working beam.
- the terminal feeds back a second measurement result to the base station.
- the terminal can feedback the second measurement result to the base station through a second measurement report message sent by physical layer signaling, and can also feedback the second measurement result through a second RRC message sent by high-level signaling.
- the specific formats of the second measurement report message and the second RRC message refer to the above embodiment and are not repeated here.
- the base station converts the logical ID into a physical ID and labels each beam in the training beam set.
- the base station sends the labeled training data set to the OAM device.
- the OAM device trains the model based on the labeled training data set to obtain a beam prediction model.
- the beam prediction model can be used to determine the optimal beam. As shown in FIG. 13, the method includes:
- the base station performs sparse beam scanning.
- Sparse beam scanning means that the base station sends a pilot signal to the terminal in the beam direction of the sparse beam.
- the sparse beam refers to a part of the full working beam, and the number of the sparse beams is the same as the number of measurement data that needs to be input to the input end of the beam prediction model. For example, if the input end of the beam prediction model needs to be input with measurement data of 16 beams, the sparse beam includes 16 beams.
- the 16 small black-filled squares in FIG. 13 are the 16 beams of this scan.
- the terminal feeds back signal strength to the base station.
- the terminal may specifically feed back the signal strength of each pilot signal, that is, the measurement result of each beam, to the base station.
- the signal strength may specifically be L-RSRP.
- the base station performs beam inference based on the beam prediction model.
- the beam prediction model can predict the physical IDs and signal strengths of the top K beams with the best performance among all beams based on the signal strengths of fewer beams.
- the base station performs a channel-state information reference signal (CSI-RS) beam scan of the Top-K beam to the terminal.
- CSI-RS channel-state information reference signal
- the Top-K beams are the top K beams with the best performance.
- K may be 5, and the Top-K beams are the 5 beams filled in the square array in the middle of FIG. 13 .
- the terminal determines the best CSI-RS beam.
- the best CSI-RS refers to the beam with the best channel state measured by the terminal.
- the quality of the channel state can be measured by signal strength.
- the terminal feeds back the best CSI-RS ID to the base station.
- the best CSI-RS ID refers to the CSI-RS ID corresponding to the beam with the best channel state measured by the terminal.
- the base station uses the best CSI-RS beam for transmission.
- the best CSI-RS ID refers to the beam corresponding to the best CSI-RS ID.
- the filled small squares represent the best CSI-RS beam.
- an embodiment of the present application provides a training data acquisition device, which is applied to a base station. As shown in FIG14 , the device includes:
- a sending module 1401 is used to send a pilot signal for beam measurement to the terminal in each beam direction of the training beam set, and send a pilot signal for beam measurement to the terminal in each beam direction of the working beam set;
- the receiving module 1402 is configured to receive a first measurement result of the training beam set and a second measurement result of the working beam set sent by the terminal;
- the generating module 1403 is configured to generate training data based on the first measurement result, and generate label data corresponding to the training data based on the second measurement result.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the first measurement timestamp is a starting timestamp for the terminal to measure the training beam set
- the second measurement timestamp is a starting timestamp for the terminal to measure the working beam set
- the generating module 1403 is specifically used for:
- the measurement data included in the first measurement result is used as training data, and the measurement data included in the second measurement result is used as label data corresponding to the training data.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the first measurement timestamp is a starting timestamp for the terminal to measure the training beam set
- the second measurement timestamp is a starting timestamp for the terminal to measure the working beam set
- the generating module 1403 is specifically used for:
- a first measurement result and a second measurement result whose absolute value of the difference between the first measurement timestamp and the second measurement timestamp is less than a preset time threshold are combined as a measurement result;
- the measurement data included in the first measurement result in the measurement result combination is used as training data, and the measurement data included in the second measurement result in the measurement result combination is used as label data.
- the device further comprises:
- a conversion module is used to convert the logical ID of each beam included in the first measurement result into a physical ID, and convert the logical ID of each beam included in the second measurement result into a physical ID.
- the physical ID of each beam of the base station is determined by the base station based on the pitch angle and azimuth angle of each beam.
- the sending module 1401 is specifically configured to:
- a cell common pilot signal for beam measurement is sent in each beam direction of the training beam set, and a terminal-specific pilot signal for beam measurement is sent to the terminal in each beam direction of the working beam.
- the receiving module 1402 is specifically configured to:
- a first RRC message and a second RRC message are sent by a receiving terminal through high-layer signaling, where the first RRC message includes a first measurement result, and the second RRC message includes a second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate the beam in the training beam set or the working beam set, and the measurement time field is used to indicate the start timestamp of the beam measurement;
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set, and the value of the beam number field in the second RRC message is a preset number, which is smaller than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- the receiving module is further used to receive an inter-system message sent by the OAM device, where the inter-system message includes a first field and a second field, where the first field is used to indicate the start of beam measurement, and the second field is used to carry a preset time threshold;
- a sending module 1401 is configured to send a first measurement control message to a terminal in response to receiving an inter-system message sent by an OAM device, where the first measurement control message is used to instruct the terminal to perform beam measurement on a training beam set at a first specified time-frequency position;
- the receiving module 1402 is used to receive a first confirmation message replied by the terminal, where the first confirmation message is used to indicate that the terminal supports beam measurement of the training beam set.
- the sending module 1401 is further configured to send a second measurement control message to the terminal in response to the inter-system message, wherein the second measurement control message The control message is used to instruct the terminal to perform beam measurement on the working beam set at the second specified time-frequency position;
- the receiving module 1402 is further used to receive a second confirmation message replied by the terminal, where the second confirmation message is used to indicate that the terminal supports beam measurement of the working beam set.
- both the first measurement control message and the second measurement control message include a preset indication field.
- the preset indication field is used to indicate the acquisition of training data of the beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set;
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of a preset number of working beams, where the preset number of working beams are the first preset number of working beams selected from the working beam set in order of signal quality from high to low.
- the device further comprises:
- An updating module configured to update the beams included in the working beam set to beams within a specified range if the moving speed of the terminal is less than or equal to a preset speed threshold, wherein the specified range is a working beam within a specified elevation angle range and a specified azimuth angle range;
- the specified pitch angle range is the pitch angle range obtained by expanding the pitch angle range of a preset number of working beams according to a first preset step size
- the specified azimuth angle range is the azimuth angle range obtained by expanding the azimuth angle range of a preset number of working beams according to a second preset step size.
- the device further comprises:
- a calculation module used for calculating the pitch angle change rate and the azimuth angle change rate of the terminal based on the position of the base station, the position of the terminal and the moving speed of the terminal;
- the calculation module is further used to calculate a first change period of the beam of the base station in the pitch angle dimension and a second change period in the azimuth angle dimension, the first change period being the ratio of the beam width in the pitch angle dimension to the pitch angle change rate, and the second change period being the ratio of the beam width in the azimuth angle dimension to the azimuth angle change rate;
- the updating module is further used to use the ratio of the smaller one of the first change period and the second change period to the preset coefficient as the measurement period after the training beam set and the working beam set are updated.
- the embodiment of the present application also provides a training data acquisition device, which is applied to a terminal.
- the device includes:
- the measurement module 1501 is configured to perform beam measurement on the pilot signal sent by the base station to the terminal in each beam direction of the training beam set, and perform beam measurement on the pilot signal sent by the base station to the terminal in each beam direction of the working beam set;
- the sending module 1502 is used to send a first measurement result of the training beam set and a second measurement result of the working beam set to the base station, where the first measurement result is the data source of the training data and the second measurement result is the data source of the label data of the training data.
- the sending module 1502 is specifically configured to:
- a first RRC message and a second RRC message are sent to a base station through high-layer signaling, wherein the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate the beam in the training beam set or the working beam set, and the measurement time field is used to indicate the start timestamp of the beam measurement;
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set
- the value of the beam number field in the second RRC message is the number of beams included in the training beam set.
- the value of the number field is a preset number, which is smaller than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- the device further comprises:
- a receiving module used to receive a first measurement control message sent by a base station, where the first measurement control message is used to instruct the terminal to perform beam measurement on the training beam set at a first specified time-frequency position;
- the sending module 1502 replies a first confirmation message to the base station;
- the receiving module is further used to receive a second measurement control message sent by the base station, where the second measurement control message is used to instruct the terminal to perform beam measurement on the working beam set at a second specified time-frequency position;
- the sending module 1502 replies a second confirmation message to the base station.
- both the first measurement control message and the second measurement control message include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data for the beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set;
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of a preset number of working beams, where the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in order of signal quality from high to low.
- the embodiment of the present application also provides a base station, as shown in FIG16 , including: a processor 1601; a transceiver 1604;
- a machine-readable storage medium 1602 stores machine-executable instructions that can be executed by the processor 1601; the machine-executable instructions prompt the processor 1601 to perform the following steps:
- Training data is generated based on the first measurement result, and label data corresponding to the training data is generated based on the second measurement result.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the first measurement timestamp is a starting timestamp for the terminal to measure the training beam set
- the second measurement timestamp is a starting timestamp for the terminal to measure the working beam set
- the machine executable instructions further cause the processor 1601 to perform the following steps:
- the measurement data included in the first measurement result is used as training data, and the measurement data included in the second measurement result is used as label data corresponding to the training data.
- the first measurement result includes a first measurement timestamp
- the second measurement result includes a second measurement timestamp
- the machine executable instructions further cause the processor 1601 to perform the following steps:
- a first measurement result and a second measurement result whose absolute value of the difference between the first measurement timestamp and the second measurement timestamp is less than a preset time threshold are combined as a measurement result;
- the measurement data included in the first measurement result in the measurement result combination is used as training data, and the measurement data included in the second measurement result in the measurement result combination is used as label data.
- machine executable instructions further cause the processor 1601 to perform the following steps:
- the logical ID of each training included in the first measurement result is converted into a physical ID
- the logical ID of each beam included in the second measurement result is converted into a physical ID.
- the ID is converted into a physical ID.
- the physical ID of each beam of the base station is determined by the base station based on the elevation angle and azimuth angle of each beam.
- machine executable instructions cause the processor 1601 to perform the following steps:
- the cell common pilot signal for beam measurement is sent in each beam direction of the training beam set through the transceiver 1604, and the terminal dedicated pilot signal for beam measurement is sent to the terminal in each beam direction of the working beam.
- machine executable instructions further cause the processor 1601 to perform the following steps:
- a first RRC message and a second RRC message sent by the terminal through high-layer signaling are received through the transceiver 1604 , where the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate the beam in the training beam set or the working beam set, and the measurement time field is used to indicate the start timestamp of the beam measurement;
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set, and the value of the beam number field in the second RRC message is a preset number, which is smaller than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- machine executable instructions further cause the processor 1601 to perform the following steps:
- the inter-system message includes a first field and a second field, the first field is used to indicate the start of beam measurement, and the second field is used to carry a preset time threshold;
- a first confirmation message replied by the terminal is received through the transceiver 1604, where the first confirmation message is used to indicate that the terminal supports beam measurement of the training beam set.
- machine executable instructions further cause the processor 1601 to perform the following steps:
- a second confirmation message replied by the terminal is received through the transceiver 1604, where the second confirmation message is used to indicate that the terminal supports beam measurement of the working beam set.
- both the first measurement control message and the second measurement control message include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data for the beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set;
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of a preset number of working beams, where the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in order of signal quality from high to low.
- machine executable instructions further cause the processor 1601 to perform the following steps:
- the beams included in the working beam set are updated to beams in a specified range, where the specified range is working beams in a specified elevation angle range and a specified azimuth angle range;
- the specified pitch angle range is the pitch angle range obtained by expanding the pitch angle range of a preset number of working beams according to a first preset step size
- the specified azimuth angle range is the azimuth angle range obtained by expanding the azimuth angle range of a preset number of working beams according to a second preset step size.
- machine executable instructions further cause the processor 1601 to perform the following steps:
- the ratio of the smaller one of the first change period and the second change period to the preset coefficient is used as the measurement period after the training beam set and the working beam set are updated.
- a communication bus 1603 may also be included.
- the processor 1601, the machine-readable storage medium 1602, and the transceiver 1604 communicate with each other through the communication bus 1603.
- the communication bus 1603 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus.
- PCI Peripheral Component Interconnect
- EISA Extended Industry Standard Architecture
- the communication bus may be divided into an address bus, a data bus, a control bus, and the like.
- the transceiver 1604 may be a wireless communication module. Under the control of the processor 1601 , the transceiver 1604 exchanges data with other devices.
- the machine-readable storage medium 1602 may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk storage.
- the machine-readable storage medium may also be at least one storage device located away from the aforementioned processor.
- Processor 1601 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
- CPU central processing unit
- NP network processor
- DSP digital signal processor
- ASIC application specific integrated circuit
- FPGA field programmable gate array
- the embodiment of the present application provides a terminal, as shown in FIG17 , the terminal includes: a processor 1701; a transceiver 1704;
- a machine-readable storage medium 1702 wherein the machine-readable storage medium 1702 stores machine-executable instructions that can be executed by the processor 1701; the machine-executable instructions cause the processor 1701 to perform the following steps:
- the first measurement result of the training beam set and the second measurement result of the working beam set are sent to the base station through the transceiver 1704.
- the first measurement result is the data source of the training data
- the second measurement result is the data source of the label data of the training data.
- machine executable instructions cause the processor 1701 to perform the following steps:
- the transceiver 1704 sends a first RRC message and a second RRC message to the base station through high-layer signaling, where the first RRC message includes the first measurement result, and the second RRC message includes the second measurement result.
- the first RRC message and the second RRC message both include a message type field, a measurement time field, a beam number field, a logical ID of each beam, and measurement data;
- the message type field is used to indicate the beam in the training beam set or the working beam set, and the measurement time field is used to indicate the start timestamp of the beam measurement;
- the value of the beam number field in the first RRC message is the number of beams included in the training beam set, and the value of the beam number field in the second RRC message is a preset number, which is smaller than the number of beams included in the working beam set.
- the first RRC message and the second RRC message also include a terminal location field and a terminal speed field; the terminal location field is used to indicate the geographical location of the terminal, and the terminal speed field is used to indicate the moving speed of the terminal during the beam measurement process.
- machine executable instructions further cause the processor 1701 to perform the following steps:
- a first confirmation message is replied to the base station through the transceiver 1704;
- the machine executable instructions further cause the processor 1701 to perform the following steps:
- the terminal If the terminal supports beam measurement of the working beam set, it replies a second confirmation message to the base station through transceiver 1704.
- both the first measurement control message and the second measurement control message include a preset indication parameter.
- the preset indication parameter is used to indicate the acquisition of training data for the beam prediction model.
- the measurement data included in the first measurement result is a logical ID and a signal quality of each beam included in the training beam set;
- the second measurement result includes the logical IDs and signal qualities of a preset number of working beams, or includes the logical IDs of a preset number of working beams, where the preset number of working beams are the first preset number of working beams selected by the terminal from the working beam set in order of signal quality from high to low.
- a communication bus 1703 may also be included.
- the processor 1701, the machine-readable storage medium 1702, and the transceiver 1704 communicate with each other through the communication bus 1703.
- the communication bus 1703 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus.
- PCI Peripheral Component Interconnect
- EISA Extended Industry Standard Architecture
- the communication bus may be divided into an address bus, a data bus, a control bus, and the like.
- the transceiver 1704 may be a wireless communication module. Under the control of the processor 1701 , the transceiver 1704 exchanges data with other devices.
- the machine-readable storage medium 1702 may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk storage.
- the machine-readable storage medium may also be at least one storage device located away from the aforementioned processor.
- Processor 1701 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
- CPU central processing unit
- NP network processor
- DSP digital signal processor
- ASIC application specific integrated circuit
- FPGA field programmable gate array
- the embodiment of the present application further provides a machine-readable storage medium, which stores machine-executable instructions that can be executed by a processor.
- the processor is prompted by the machine-executable instructions to implement the steps of any of the above training data collection methods.
- a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute the steps of any training data collection method in the above embodiments.
- each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments.
- the segment identification determination device equipment and machine-readable storage medium.
- the machine-readable storage medium embodiment since it is basically similar to the segment identification determination method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the segment identification determination method embodiment.
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
本申请提供一种训练数据采集方法、装置、基站、终端及存储介质,涉及通信技术领域,该方法包括:在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向终端发送用于波束测量的导频信号;接收终端发送的对训练波束集的第一测量结果,以及对工作波束集的第二测量结果;基于第一测量结果生成训练数据,基于第二测量结果生成训练数据对应的标签数据。能够准确获取到训练数据和标签数据。
Description
本申请涉及通信技术领域,尤其涉及一种训练数据采集方法、装置、基站、终端及存储介质。
在移动通信场景下,用户移动与信号被随机遮挡等因素都会导致信道时变,所以需要反复进行波束测量和跟踪,从而实时准确地确定基站与终端通信的信道信息。传统的波束预测方式为基站在所有波束方向上分别发送导频信号,从而预测得到最优波束的波束方向,然而这种方式会产生巨大的导频开销。
为了降低导频开销,产生了基于人工智能和机器学习的波束预测算法,可以利用机器学习的方式训练波束预测模型,从而利用波束预测模型,通过对较少波束的测量结果预测最优波束的波束方向。如何得到用于训练波束预测模型的训练数据,是目前亟需解决的问题。
发明内容
本申请实施例的目的在于提供一种训练数据采集方法、装置、基站、终端及存储介质,以准确获取用于训练波束预测模型的训练数据,具体技术方案如下:
第一方面,本申请实施例提供一种训练数据采集方法,所述方法应用于基站,所述方法包括:
在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号;
接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果;
基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据。
在一种可能的实现方式中,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;
所述基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据,包括:
若所述第一测量时间戳和所述第二测量时间戳之间的差值绝对值小于预设时间门限,则将所述第一测量结果包括的测量数据作为训练数据,将所述第二测量结果包括的测量数据作为所述训练数据对应的标签数据。
在一种可能的实现方式中,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;
所述基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据,包括:
针对已接收的多个第一测量结果和多个第二测量结果,将第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的一个第一测量结果和一个第二测量结果作为一个测量结果组合;
针对每个测量结果组合,将该测量结果组合中的第一测量结果包括的测量数据作为训练数据,将该测量结果组合中的第二测量结果包括的测量数据作为标签数据。
在一种可能的实现方式中,在所述基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据之前,所述方法还包括:
将所述第一测量结果包括的每个波束的逻辑ID转换为物理ID,将所述第二测量结果包括的每个波
束的逻辑ID转换为物理ID,所述基站的各波束的物理ID为所述基站基于各自波束的俯仰角和方位角确定出的。
在一种可能的实现方式中,所述在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号,包括:
在所述训练波束集的各波束方向上向所述终端发送用于波束测量的终端专用导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号;或者,
在所述训练波束集的各波束方向上发送用于波束测量的小区公共导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号。
在一种可能的实现方式中,所述接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,包括:
接收所述终端通过物理层信令发送的第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,
接收所述终端通过高层信令发送的第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;
所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数量小于所述工作波束集中包括的波束数量。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
在一种可能的实现方式中,在所述在训练波束集的各波束方向上向终端发送用于波束测量的导频信号之前,所述方法包括:
接收OAM设备发送的系统间消息,所述系统间消息包括第一字段和第二字段,所述第一字段用于指示启动波束测量,所述第二字段用于承载所述预设时间门限;
响应于所述系统间消息,向所述终端发送第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;
接收所述终端回复的第一确认消息,所述第一确认消息用于表示所述终端支持对所述训练波束集进行波束测量。
在一种可能的实现方式中,在所述在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号之前,所述方法还包括:
响应于所述系统间消息,向所述终端发送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;
接收所述终端回复的第二确认消息,所述第二确认消息用于表示所述终端支持对所述工作波束集进行波束测量。
在一种可能的实现方式中,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,
当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
在一种可能的实现方式中,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
在一种可能的实现方式中,在所述接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果之后,所述方法还包括:
若所述终端的移动速度小于等于预设速度阈值,则将所述工作波束集包括的波束更新为指定范围的波束,所述指定范围为指定俯仰角范围和指定方位角范围中的工作波束;
其中,所述指定俯仰角范围为将所述预设数量的工作波束的俯仰角范围按第一预设步长扩大后得到的俯仰角范围,所述指定方位角范围为将所述预设数量的工作波束的方位角范围按第二预设步长扩大后得到的方位角范围。
在一种可能的实现方式中,在所述接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果之后,所述方法还包括:
基于所述基站的位置、所述终端的位置和所述终端的移动速度,计算所述终端的俯仰角变化速率和方位角变化速率;
计算所述基站的波束在俯仰角维度的第一变化周期和在方位角维度的第二变化周期,所述第一变化周期为所述俯仰角维度的波束宽度与所述俯仰角变化速率的比值,所述第二变化周期为所述方位角维度的波束宽度与所述方位角变化速率的比值;
将所述第一变化周期和所述第二变化周期中较小的一个变化周期与预设系数的比值,作为所述训练波束集和所述工作波束集更新后的测量周期。
第二方面,本申请实施例提供一种训练数据采集方法,所述方法应用于终端,所述方法包括:
对基站在训练波束集的各波束方向上向所述终端发送的导频信号进行波束测量,对所述基站在工作波束集的各波束方向上向所述终端发送的导频信号进行波束测量;
向所述基站发送对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,所述第一测量结果为训练数据的数据来源,所述第二测量结果为所述训练数据的标签数据的数据来源。
在一种可能的实现方式中,所述向所述基站发送对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,包括:
通过物理层信令向所述基站发送第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,
通过高层信令向所述基站发送第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;
所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC
消息中的波束数目字段的取值为预设数量,所述预设数目小于所述工作波束集中包括的波束数量。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
在一种可能的实现方式中,在所述对基站在训练波束集的各波束方向上发送的导频信号进行波束测量之前,所述方法还包括:
接收所述基站发送的第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;
若所述终端支持对所述训练波束集进行波束测量,则向所述基站回复第一确认消息;
在所述对所述基站在工作波束集的各波束方向上发送的导频信号进行波束测量之前,所述方法还包括:
接收所述基站发送的送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;
若所述终端支持对所述工作波束集进行波束测量,则向所述基站回复第二确认消息。
在一种可能的实现方式中,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
在一种可能的实现方式中,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
第三方面,本申请实施例提供一种训练数据采集装置,所述装置应用于基站,所述装置包括:
发送模块,用于在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号;
接收模块,用于接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果;
生成模块,用于基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据。
在一种可能的实现方式中,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;
所述生成模块,具体用于:
若所述第一测量时间戳和所述第二测量时间戳之间的差值绝对值小于预设时间门限,则将所述第一测量结果包括的测量数据作为训练数据,将所述第二测量结果包括的测量数据作为所述训练数据对应的标签数据。
在一种可能的实现方式中,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;
所述生成模块,具体用于:
针对已接收的多个第一测量结果和多个第二测量结果,将第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的一个第一测量结果和一个第二测量结果作为一个测量结果组合;
针对每个测量结果组合,将该测量结果组合中的第一测量结果包括的测量数据作为训练数据,将该测量结果组合中的第二测量结果包括的测量数据作为标签数据。
在一种可能的实现方式中,所述装置还包括:
转换模块,用于将所述第一测量结果包括的每个波束的逻辑ID转换为物理ID,将所述第二测量结果包括的每个波束的逻辑ID转换为物理ID,所述基站的各波束的物理ID为所述基站基于各自波束的俯仰角和方位角确定出的。
在一种可能的实现方式中,所述发送模块,具体用于:
在所述训练波束集的各波束方向上向所述终端发送用于波束测量的终端专用导频信号,在所述工作波束的各波束方向上向所述终端上发送用于波束测量的终端专用导频信号;或者,
在所述训练波束集的各波束方向上发送用于波束测量的小区公共导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号。
在一种可能的实现方式中,所述接收模块,具体用于:
接收所述终端通过物理层信令发送的第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,
接收所述终端通过高层信令发送的第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;
所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数量小于所述工作波束集中包括的波束数量。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
在一种可能的实现方式中,所述接收模块,还用于接收OAM设备发送的系统间消息,所述系统间消息包括第一字段和第二字段,所述第一字段用于指示启动波束测量,所述第二字段用于承载所述预设时间门限;
所述发送模块,用于响应于接收OAM设备发送的系统间消息,向所述终端发送第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;
所述接收模块,用于接收所述终端回复的第一确认消息,所述第一确认消息用于表示所述终端支持对所述训练波束集进行波束测量。
在一种可能的实现方式中,所述发送模块,还用于响应于所述系统间消息,向所述终端发送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;
所述接收模块,还用于接收所述终端回复的第二确认消息,所述第二确认消息用于表示所述终端支持对所述工作波束集进行波束测量。
在一种可能的实现方式中,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
在一种可能的实现方式中,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
在一种可能的实现方式中,所述装置还包括:
更新模块,用于若所述终端的移动速度小于等于预设速度阈值,则将所述工作波束集包括的波束更新为指定范围的波束,所述指定范围为指定俯仰角范围和指定方位角范围中的工作波束;
其中,所述指定俯仰角范围为将所述预设数量的工作波束的俯仰角范围按第一预设步长扩大后得到的俯仰角范围,所述指定方位角范围为将所述预设数量的工作波束的方位角范围按第二预设步长扩大后得到的方位角范围。
在一种可能的实现方式中,所述装置还包括:
计算模块,用于基于所述基站的位置、所述终端的位置和所述终端的移动速度,计算所述终端的俯仰角变化速率和方位角变化速率;
所述计算模块,还用于计算所述基站的波束在俯仰角维度的第一变化周期和在方位角维度的第二变化周期,所述第一变化周期为所述俯仰角维度的波束宽度与所述俯仰角变化速率的比值,所述第二变化周期为所述方位角维度的波束宽度与所述方位角变化速率的比值;
所述更新模块,还用于将所述第一变化周期和所述第二变化周期中较小的一个变化周期与预设系数的比值,作为所述训练波束集和所述工作波束集更新后的测量周期。
第四方面,本申请实施例提供一种训练数据采集装置,所述装置应用于终端,所述装置包括:
测量模块,用于对基站在训练波束集的各波束方向上向所述终端发送的导频信号进行波束测量,对所述基站在工作波束集的各波束方向上向所述终端发送的导频信号进行波束测量;
发送模块,用于向所述基站发送对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,所述第一测量结果为训练数据的数据来源,所述第二测量结果为所述训练数据的标签数据的数据来源。
在一种可能的实现方式中,所述发送模块,具体用于:
通过物理层信令向所述基站发送第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,
通过高层信令向所述基站发送第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;
所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数目小于所述工作波束集中包括的波束数量。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
在一种可能的实现方式中,所述装置还包括:
接收模块,用于接收所述基站发送的第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;
若所述终端支持对所述训练波束集进行波束测量,则所述发送模块向所述基站回复第一确认消息;
所述接收模块,还用于接收所述基站发送的送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;
若所述终端支持对所述工作波束集进行波束测量,则所述发送模块向所述基站回复第二确认消息。
在一种可能的实现方式中,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
在一种可能的实现方式中,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
第五方面,本申请实施例提供一种基站,包括:处理器;收发器;
机器可读存储介质,所述机器可读存储介质存储有能够被所述处理器执行的机器可执行指令;所述机器可执行指令促使所述处理器执行以下步骤:
通过所述收发器在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号;
通过所述收发器接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果;
基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据。
在一种可能的实现方式中,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;
所述机器可执行指令还促使所述处理器执行以下步骤:
若所述第一测量时间戳和所述第二测量时间戳之间的差值绝对值小于预设时间门限,则将所述第一测量结果包括的测量数据作为训练数据,将所述第二测量结果包括的测量数据作为所述训练数据对应的标签数据。
在一种可能的实现方式中,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;
所述机器可执行指令还促使所述处理器执行以下步骤:
针对已接收的多个第一测量结果和多个第二测量结果,将第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的一个第一测量结果和一个第二测量结果作为一个测量结果组合;
针对每个测量结果组合,将该测量结果组合中的第一测量结果包括的测量数据作为训练数据,将该测量结果组合中的第二测量结果包括的测量数据作为标签数据。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
将所述第一测量结果包括的每个波束的逻辑ID转换为物理ID,将所述第二测量结果包括的每个波束的逻辑ID转换为物理ID,所述基站的各波束的物理ID为所述基站基于各自波束的俯仰角和方位角确定出的。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
通过所述收发器在所述训练波束集的各波束方向上向所述终端发送用于波束测量的终端专用导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号;或者,
通过所述收发器在所述训练波束集的各波束方向上发送用于波束测量的小区公共导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
通过所述收发器接收所述终端通过物理层信令发送的第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,
通过所述收发器接收所述终端通过高层信令发送的第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;
所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数量小于所述工作波束集中包括的波束数量。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
通过所述收发器接收OAM设备发送的系统间消息,所述系统间消息包括第一字段和第二字段,所述第一字段用于指示启动波束测量,所述第二字段用于承载所述预设时间门限;
响应于所述系统间消息,通过所述收发器向所述终端发送第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;
通过所述收发器接收所述终端回复的第一确认消息,所述第一确认消息用于表示所述终端支持对所述训练波束集进行波束测量。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
响应于所述系统间消息,通过所述收发器向所述终端发送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;
通过所述收发器接收所述终端回复的第二确认消息,所述第二确认消息用于表示所述终端支持对所
述工作波束集进行波束测量。
在一种可能的实现方式中,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
在一种可能的实现方式中,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
若所述终端的移动速度小于等于预设速度阈值,则将所述工作波束集包括的波束更新为指定范围的波束,所述指定范围为指定俯仰角范围和指定方位角范围中的工作波束;
其中,所述指定俯仰角范围为将所述预设数量的工作波束的俯仰角范围按第一预设步长扩大后得到的俯仰角范围,所述指定方位角范围为将所述预设数量的工作波束的方位角范围按第二预设步长扩大后得到的方位角范围。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
基于所述基站的位置、所述终端的位置和所述终端的移动速度,计算所述终端的俯仰角变化速率和方位角变化速率;
计算所述基站的波束在俯仰角维度的第一变化周期和在方位角维度的第二变化周期,所述第一变化周期为所述俯仰角维度的波束宽度与所述俯仰角变化速率的比值,所述第二变化周期为所述方位角维度的波束宽度与所述方位角变化速率的比值;
将所述第一变化周期和所述第二变化周期中较小的一个变化周期与预设系数的比值,作为所述训练波束集和所述工作波束集更新后的测量周期。
第六方面,本申请实施例提供一种终端,包括:处理器;收发器;
机器可读存储介质,所述机器可读存储介质存储有能够被所述处理器执行的机器可执行指令;所述机器可执行指令促使所述处理器执行以下步骤:
对基站在训练波束集的各波束方向上向所述终端发送的导频信号进行波束测量,对所述基站在工作波束集的各波束方向上向所述终端发送的导频信号进行波束测量;
通过所述收发器向所述基站发送对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,所述第一测量结果为训练数据的数据来源,所述第二测量结果为所述训练数据的标签数据的数据来源。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
通过所述收发器通过物理层信令向所述基站发送第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,
通过所述收发器通过高层信令向所述基站发送第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测
量的起始时间戳;
所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数目小于所述工作波束集中包括的波束数量。
在一种可能的实现方式中,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
在一种可能的实现方式中,所述机器可执行指令还促使所述处理器执行以下步骤:
通过所述收发器接收所述基站发送的第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;
若所述终端支持对所述训练波束集进行波束测量,则通过所述收发器向所述基站回复第一确认消息;
所述机器可执行指令还促使所述处理器执行以下步骤:
通过所述收发器接收所述基站发送的送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;
若所述终端支持对所述工作波束集进行波束测量,则通过所述收发器向所述基站回复第二确认消息。
在一种可能的实现方式中,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
在一种可能的实现方式中,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
第七方面,本申请实施例提供一种机器可读存储介质,存储有机器可执行指令,在被处理器调用和执行时,所述机器可执行指令促使所述处理器:实现第一方面或第二方面任一所述的方法步骤。
第八方面,本申请实施例提供一种计算机程序产品,所述计算机程序产品促使所述处理器:实现第一方面或第二方面任一所述的方法步骤。
采用该上述技术方案,基站可以在训练波束集和工作波束集的各波束方向上向终端发送用于波束测量的导频信号,基于终端对训练波束集的第一测量结果生成训练数据,并基于终端对工作波束集的第二测量结果生成训练数据对应的标签数据。第一测量结果和第二测量结果均为终端实际测量得到的,所以采用该方法生成的训练数据和标签数据均符合基站和终端之间的实际通信情况,即采用该方法可以准确地获取到用于训练波束预测模型的训练数据。
此处所说明的附图用来提供对本申请的进一步理解,构成本申请的一部分,本申请的示意性实施例及其说明用于解释本申请,并不构成对本申请的不当限定。
图1为本申请实施例提供的一种移动通信系统的架构图;
图2为本申请实施例提供的一种基于AI/ML的波束预测结构的示例性示意图;
图3为本申请实施例提供的一种训练数据采集方法的流程图;
图4为本申请实施例提供的一种波束的时频资源与逻辑ID的映射关系的示例性示意图;
图5为本申请实施例提供的波束的方位角与俯仰角的示例性示意图;
图6为本申请实施例提供的一种物理ID的编号结果的示例性示意图;
图7为本申请实施例提供的一种训练数据采集过程中构建训练数据集的方法的示例性示意图;
图8为本申请实施例提供的另一种训练数据采集方法的流程图;
图9为本申请实施例提供的一种训练数据采集过程中调整工作波束集的方法流程图;
图10a为本申请实施例提供的一种训练数据采集过程中调整测量周期的方法流程图;
图10b为本申请实施例提供的一种终端与基站的相对位置关系的示例性示意图;
图11为本申请实施例提供的另一种训练数据采集方法的流程图;
图12为本申请实施例提供的另一种训练数据采集方法的流程图;
图13为本申请实施例提供的一种基站确定最佳波束的方法的流程图;
图14为本申请实施例提供的一种训练数据采集装置的结构示意图;
图15为本申请实施例提供的另一种训练数据采集装置的结构示意图;
图16为本申请实施例提供的一种基站的结构示意图;
图17为本申请实施例提供的一种终端的结构示意图。
为使本发明的目的、技术方案、及优点更加清楚明白,以下参照附图并举实施例,对本发明进一步详细说明。显然,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员所获得的所有其他实施例,都属于本发明保护的范围。
本申请实施例提供的方法可以应用于移动通信系统中,如图1所示,移动通信系统中包括基站101和基站101覆盖范围内的多个终端102,基站101可使用多个不同指向的波束覆盖自身服务的小区。波束预测的目的是分别确定基站101与每个终端102通信时所能使用的最优发射波束。
传统的波束预测方法为,基站进行全波束扫描,即基站分别通过每个方向的波束向终端发送导频信号,终端测量各导频信号,并向基站上报针对每个导频信号的测量结果,进而基站基于测量结果为该终端选择最优发射波束。
为了降低导频开销,产生了基于人工智能/机器学习(Artificial Intelligence/Machine Learning,AI/ML)的波束预测方法,即利用波束预测模型进行波束预测,可以实现通过对少量的波束(比如16个)进行测量,从而预测全量工作波束(例如64个)中的较优波束,然后从16个较优波束中选择最优发射波束,可以达到较高的波束预测效果,并且可以大幅降低导频开销。
如图2所示,图2是本申请实施例提供的一种基于AI/ML的波束预测结构的示例性示意图,包括数据采集模块201、模型训练模块202、模型推理模块203以及执行模块204。
数据采集模块201可以被部署在基站中,或者部分功能被部署在基站中,部分功能被部署在其他网络设备中。
模型训练模块202可以被部署在基站中,或者被部署在其他网络设备中。
模型推理模块203和执行模块204均被部署在基站中。
其中,数据采集模块201用于采集训练数据、训练数据对应的标签数据和推理数据,并对训练数据和推理数据进行标识、清洗和归一化处理,以使处理后的训练数据和推理数据的格式和质量满足模型训练的要求。训练数据和推理数据均可以为各波束的层1参考信号接收功率(Layer1 Reference Signal Receiving Power,L1-RSRP),L1-RSRP可用于表示物理层导频信号强度。
训练数据和训练数据对应的标签数据用于训练波束预测模型,推理数据为在训练好的波束预测模型被部署之后,需输入至波束预测模型中的数据。
在模型训练阶段,数据采集模块201可以将训练数据和训练数据对应的标签数据输入模型训练模块202,模型训练模块202用于根据数据采集模块201发送的训练数据进行模型训练。具体可以采用深度神经网络(Deep Neural Network,DNN)或长短期记忆网络(Deep Neural Network,LSTM)等机器学习算法进行训练,以得到波束预测模型。例如,模型训练模块202可以将训练数据输入模型中,获取模型的输出结果,基于模型的输出结果和训练数据对应的标签数据计算损失函数值,然后基于损失函数值调整模型参数,直至模型收敛时,将训练得到的模型作为波束预测模型。
进而,模型训练模块202可以在模型推理模块203中进行模型部署,也就是将训练好的波束预测模型部署在模型推理模块203中。
在模型应用阶段,数据采集模块201可以将推理数据输入模型推理模块203,模型推理模块203可以利用波束预测模型处理推理数据,从而获取波束预测模型输出的推理结果,并将推理结果发送给执行模块204。该推理结果可以为波束预测模型预测出的波束。
执行模块204为基站中用于与终端通信的模块,执行模块204用于利用发射波束向终端发送信号,并利用接收波束接收终端发送的信号。
例如,执行模块204可以按照波束预测模型预测出的发射波束向终端发送信号。
并且执行模块204还可以利用部分发射波束向终端发射导频信号,并接收终端反馈的每个发射波束对应的L1-RSRP。进而执行模块204可以向数据采集模块201进行数据反馈,也就是将终端反馈的每个发射波束对应的L1-RSRP反馈给数据采集模块201。数据采集模块可将获取到的每个发射波束对应的L1-RSRP作为推理数据。
模型推理模块203还可向模型训练模块202进行模型性能反馈,具体可以反馈波束预测模型预测的波束的准确率,若该准确率小于预设准确率阈值,则模型训练模块202可再次对波束预测模型进行训练,以优化波束预测模型的参数,并基于优化后的参数对模型推理模块203中的波束预测模型进行模型更新。
在上述过程中,为了训练得到能够准确进行波束预测的波束预测模型,需要采集大量的训练数据,目前对于如何准确高效地采集训练数据还没有定义。
为了解决上述问题,本申请实施例提供了一种训练数据采集方法,该方法应用于基站,如图3所示,该方法包括:
S301、在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向终端发送用于波束测量的导频信号。
其中,工作波束集中可以包括全量工作波束,全量工作波束是指基站中正常工作的波束,例如工作波束集中包括64个工作波束。
训练波束集中包括用于波束预测的波束,训练波束集中包括的波束数量与波束预测模型单次需被输入的训练数据的数量相同。
例如,若波束预测模型的输入为16个波束的L1-RSRP,则训练波束集中包括的波束数量为16,该波束预测模型可以利用训练波束集中的16个波束的L1-RSRP预测所有工作波束中的较优波束。
S302、接收终端发送的对训练波束集的第一测量结果,以及对工作波束集的第二测量结果。
终端可以对基站发送的每个导频信号进行测量,将对导频信号的测量结果,作为该导频信号对应的波束的测量结果。
其中,第一测量结果中包括终端对训练波束集中每个波束的测量结果,第二测量结果中包括终端对工作波束集中信号质量排名前预设数量的工作波束的测量结果。
该预设数量与波束预测模型预测的波束数量相同,若波束预测模型用于预测5个较优波束,即信号质量排名靠前的5个波束,则该预设数量为5。
S303、基于第一测量结果生成训练数据,基于第二测量结果生成训练数据对应的标签数据。
采用该方法,基站可以在训练波束集和工作波束集的各波束方向上向终端发送用于波束测量的导频信号,基于终端对训练波束集的第一测量结果生成训练数据,并基于终端对工作波束集的第二测量结果生成训练数据对应的标签数据。第一测量结果和第二测量结果均为终端实际测量得到的,所以采用该方法生成的训练数据和标签数据均符合基站和终端之间的实际通信情况,即采用该方法可以准确地获取到用于训练波束预测模型的训练数据。
在本申请的一些实施例中,上述S301、在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向终端发送用于波束测量的导频信号,具体包括以下两种实现方式:
方式一、在训练波束集的各波束方向上向终端发送用于波束测量的终端专用导频信号,在工作波束的各波束方向上向终端发送用于波束测量的终端专用导频信号。
其中,基站可以根据预先配置的时频位置,向多个终端发送用于波束测量的导频信号。即基站可以在训练波束集的各波束方向上分别向每个终端发送终端专用导频信号,并在工作波束的各波束方向上分别向每个终端发送终端专用导频信号。进而,每个接收到终端专用导频信号的终端均可以向基站反馈第一测量结果和第二测量结果。如此,通过终端专用导频信号可以使得终端更准确地对训练波束集的各波束和工作波束进行测量。
方式二、在训练波束集的各波束方向上发送用于波束测量的小区公共导频信号,在工作波束的各波束方向上向终端发送用于波束测量的终端专用导频信号。
其中,小区公共导频信号为基站覆盖区域内,所有终端均可以接收到的导频信号,小区公共导频信号可以为同步广播块信号(Synchronization Signal Block,SSB)。
基站在训练波束集的各波束方向上发送小区公共导频信号后,基站覆盖范围内的所有终端均可以反馈针对各小区公共导频信号的第一测量结果。同样的,基站在工作波束的各波束方向上分别向每个终端发送终端专用导频信号后,每个终端均可以反馈自身的终端专用导频信号对应的第二测量结果。如此,向终端发送小区公共导频信号,相比与分别向每个终端发送终端专用导频信号,可以节省导频开销,提高测量效率。
在本申请的一些实施例中,S302、接收终端发送的对训练波束集的第一测量结果,以及对工作波束集的第二测量结果,具体包括以下两种实现方式:
方式一、接收终端通过物理层信令发送的第一测量报告消息和第二测量报告消息。
其中,该物理层信令具体为层1(Layer1,L1)信令,第一测量报告消息中包括第一测量结果,第二测量报告消息中包括第二测量结果。第一测量报告消息和第二测量报告消息的格式遵循无线资源控制(Radio Resource Control,RRC)协议的规定,此处不再赘述。
第一测量结果中具体可以包括第一测量时间戳以及终端对训练波束集中各波束的测量数据。第一测量时间戳为终端对训练波束集测量的起始时间戳。
第二测量结果中具体可以包括第二测量时间戳以及终端对工作波束集中预设数量个工作波束的测
量数据,第二测量时间戳为终端对工作波束集测量的起始时间戳。
需要说明的是,终端可周期性对训练波束集和工作波束集进行测量。终端在每个周期对训练波束集进行波束测量后,均会向基站反馈一次第一测量结果,该第一测量结果中包括的第一测量时间戳为在该周期内,对训练波束集测量的起始时间戳。同理,终端在每个周期对工作波束集进行波束测量后,均会向基站反馈一次第二测量结果,该第二测量结果中包括的第二测量时间戳为在该周期内,对工作波束集测量测量的起始时间戳。
采用物理层信令传输第一测量报告消息和第二测量报告消息的传输时延较低,可以使得波束测量的时效性更高。
方式二、接收终端通过高层信令发送的第一RRC消息和第二RRC消息,第一RRC消息中包括第一测量结果,第二RRC消息中包括第二测量结果。
由于常规波束测量对时效性要求较高,所以可以采用时延较低的物理层信令的方式传输测量结果,而本申请实施例中的训练数据用于对波束预测模型进行训练,对时延要求不高,所以第一测量结果和第二测量结果均可采用高层信令进行传输,如此可以减少对物理层控制信道的容量需求,且高层信令的灵活性和扩展性较高,可以按照训练数据的需求设置高层信令传输的RRC消息格式,利用高层信令有助于提升波束预测算法的性能。
本申请实施例定义了利用高层信令传输的RRC消息的格式,以下对第一RRC消息和第二RRC消息的消息格式进行介绍。
第一RRC消息和第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑身份标识(Identity,ID)和测量数据。
其中,消息类型字段用于指示被测波束的类型,即用于指示训练波束集或工作波束集包括的各波束,例如,若消息类型字段的取值为1,则表示携带有训练波束集包括的各波束的测量结果;若消息类型的取值为2,则表示携带有工作波束集包括的各波束的测量结果。第一RRC消息中的消息类型字段的取值为1,第二RRC消息中的消息类型字段的取值为2。
测量时间字段用于表示波束测量的起始时间戳,该起始时间戳具体为终端对训练波束集或工作波束集开始测量的起始时间戳,该起始时间戳可通过系统帧号(System Frame Number,SFN)编号或者协调世界时(Coordinated Universal Time,UTC)表示。
第一RRC消息中的波束数目字段用于表示此次对训练波束集包括的波束的上报数目,该波束数目字段的取值为训练波束集中包括的波束数量,相应地,第一RRC消息中携带训练波束集中各波束的逻辑ID和测量数据。
第二RRC消息中的波束数目字段用于表示此次对工作波束的上报数目,该波束数目字段的取值为预设数量,相应地,第二RRC消息中携带工作波束集中预设数量个波束的逻辑ID和测量数据。其中,预设数量小于工作波束集中包括的波束数量,该预设数量与波束预测模型预测的波束数量相同,若波束预测模型用于预测5个较优波束,则该预设数量为5。
可选地,第一RRC消息和第二RRC消息中还包括终端位置字段和终端速度字段。
终端位置字段用于表示终端所处的地理位置,终端可通过全球导航卫星系统(Global Navigation Satellite System,GNSS)定位或其它定位方法获取自身的地理位置。
终端速度字段用于表示终端在波束测量过程中的移动速度。可选地,第一RRC消息中的终端速度字段用于表示终端对训练波束集中的各波束测量过程中的平均移动速度,第二RRC消息中的终端速度
字段用于表示终端对工作波束集中的各波束测量过程中的平均移动速度。移动速度包括终端在水平和垂直两个方向上的移动速度。
第一RRC消息和第二RRC消息中包括的各字段以及各字段的含义如表1所示。
表1
在上述实施例的基础上,可以理解的是,第一测量结果中包括第一测量时间戳和测量数据,第二测量结果中包括第二测量时间戳和测量数据。第一测量时间戳为训练波束集被进行波束测量的起始时间戳,例如,第一测量时间戳具体为第一RRC消息中携带的测量时间字段的值。第二测量时间戳为工作波束集被进行波束测量的起始时间戳,例如第二测量时间戳具体为第二RRC消息中携带的测量时间字段的值。
第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量,信号质量具体可通过L1-RSRP表示。
第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括预设数量个工作波束的逻辑ID。其中,预设数量个工作波束为终端按照信号质量从高到低的顺序,从工作波束集中选择的前预设数量个工作波束。
换言之,第二测量结果中包括信号质量TOP-K的工作波束的逻辑ID和信号质量,或者第二测量结果中包括信号质量TOP-K的工作波束的逻辑ID。
其中K为预设数量,假设预设数量为5,则第二测量结果中包括64个工作波束中,信号质量排前5的5个工作波束的逻辑ID和信号质量,或包括信号质量排前5的5个工作波束的逻辑ID。
第二测量结果中是否包括信号质量取决于波束预测模型的输出是否包括信号质量。若波束预测模型的输出为预设数量个波束的ID和信号质量,则第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量;若波束预测模型的输出为预设数量个波束的ID,则第二测量结果中包括预设数量个工作波束的逻辑ID。
在上述实施例中,波束的逻辑ID也可称为波束的索引号,由于不同基站为波束设置逻辑ID的规
则不同,所以不同基站可能对相同波束方向的波束设置不同的逻辑ID。
作为示例,如图4所示,图4左侧的横轴表示时间,纵轴表示频率,从左到右4列被填充的位置分别表示4个波束的时频资源,这4个波束的物理ID分别为索引(index)1、索引2、索引3和索引4。图4右侧每个波束占用的时频资源为图4左侧具有相同填充位置处表示的时频资源。
图4右侧基站1的4个波束的逻辑ID分别为索引1、索引2、索引3和索引4,但基站2中的4个波束的逻辑ID分别为索引3、索引4、索引1和索引2。对比可知,基站1中索引1对应的波束与基站2中索引1对应的波束方向不同,基站1与基站2中索引2对应的波束方向不同,索引3和索引4同理。
因波束预测模型的训练数据来自多个基站,为了避免训练数据中使用的各波束的ID不统一,本申请实施例提供了一种将波束的逻辑上的逻辑ID与真实的物理波束匹配的方法,即基站在基于第一测量结果生成训练数据,基于第二测量结果生成训练数据对应的标签数据之前,可以将第一测量结果包括的每个波束的逻辑ID转换为物理ID,将第二测量结果包括的每个波束的逻辑ID转换为物理ID。
其中,基站的各波束的物理ID基于波束的俯仰角和方位角确定。具体可以基于各波束在笛卡尔坐标系中的俯仰角和方位角进行编号,将编号作为波束的物理ID。
俯仰角是波束指向和水平面的夹角,方位角是正北方向沿顺时针方向旋转至波束方向所形成的的夹角。
例如,如图5所示,图5是本申请实施例提供的一种波束方向的方位角和俯仰角的示例性示意图。图5中的椭圆形为天线阵面的一个波束,该波束方向和水平面的夹角为俯仰角θ,正北方向与波束方向之间的夹角为方位角
具体的,基站对于自身任意天线的波束集,可以依据波束集中各波束的波束方向,按照先俯仰角,再方位角的顺序依次对波束集中的各波束进行编号。即先按照俯仰角从小到大的顺序,对各波束方向的波束进行排序,然后对于俯仰角相同的波束,按照方位角从小到大的顺序。基站可以按照排序结果依次对各波束进行编号。
作为示例,如图6所示,图6中的横轴为方位角,纵轴为俯仰角,方位角和俯仰角的单位均为度。图6中每个阴影矩形区域代表一个波束方向,图6中最下面一行的4个波束的俯仰角相同,从左到右方位角依次增大,所以对应的编号分别为0、1、2和3。然后中间一行波束的俯仰角大于最下面一行波束的俯仰角,从左到右方位角依次增大,所以对应的编号分别为4、5、6和7。同理,最上面一行的4个波束对应的编号分别为8、9、10和11。
图6仅为一种示例,在实际实现中,若以1度为单位,可以将波束集中的波束划分为360*360个编号,基站可以根据需要进行转换的波束的俯仰角和方位角,将该波束映射至一个编号上,将该编号作为该波束的物理ID。
具体的,对于第一测量结果包括的训练波束集中各波束的逻辑ID,基站可确定该逻辑ID对应的波束的俯仰角和方位角,然后将该俯仰角和方位角对应的编号作为该波束的物理ID,从而将该波束的逻辑ID替换为物理ID。
同理,对于第二测量结果中包括的每个工作波束的逻辑ID,基站可利用相同的方法,将该工作波束的逻辑ID替换为物理ID。
如此,由于波束的物理ID是根据波束的俯仰角和方位角确定的,所以不同基站对相同波束方向的波束确定出的物理ID相同,进而可以使得后续各基站生成的训练数据中相同物理ID对应的波束方向相同。如此,提供了统一的数据格式,能够更好地支持多场景大模型的联合训练和多厂商数据的联合优
化。
可以理解的是,基站可以采用周期、半周期或非周期等方式多次在训练波束集和工作波束集的各波束方向上发送用于波束测量的导频信号,相应地,终端也会多次反馈第一测量报告和第二测量报告。为了保证训练数据的准确性,基站需保证生成一组训练数据和标签数据所使用的第一测量报告和第二测量报告的时间差在一定范围内。
在一种实施方式中,上述S303、基于第一测量结果生成训练数据,基于第二测量结果生成训练数据对应的标签数据,具体可以实现为:
若第一测量时间戳和第二测量时间戳之间的差值绝对值小于预设时间门限,则将第一测量结果包括的测量数据作为训练数据,将第二测量结果包括的测量数据作为训练数据对应的标签数据。
基站在接收到同一终端反馈的第一测量结果和第二测量结果后,可以判断第一测量结果携带的第一测量时间戳和第二测量结果携带的第二时间戳之间的差值绝对值是否小于预设时间门限,若是,则可将该第一测量结果和第二测量结果作为一个测量结果组合,并将该第一测量结果包括的测量数据作为训练数据,将第二测量结果包括的测量数据作为训练数据对应的标签数据。
其中,本申请实施例中的预设时间门限为基于经验预先设置的值,用于指示第二测量结果能否用于作为第一测量结果中的训练数据的标签数据。预设时间门限的取值受波束变化速度、终端移动速度和环境变化影响,可由人工灵活配置。
在另一种实施方式中,基站也可以批量生成训练数据和标签数据,即上述S303、基于第一测量结果生成训练数据,基于第二测量结果生成训练数据对应的标签数据,具体可以实现为:
针对已接收的多个第一测量结果和多个第二测量结果,将第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的一个第一测量结果和一个第二测量结果作为一个测量结果组合;
针对每个测量结果组合,将该测量结果组合中的第一测量结果包括的测量数据作为训练数据,将该测量结果组合中的第二测量结果包括的测量数据作为标签数据。
其中,上述第一测量结果和第二测量结果均来自同一终端,对于来自每个终端的第一测量结果和第二测量结果,均可按照该方法生成训练数据和标签数据。
基站可以在接收到来自同一终端的多个第一测量结果和多个第二测量结果后,批量生成训练数据和标签数据。可选地,基站可以每隔固定时长执行一次批量生成训练数据和标签数据的操作;或者接收到的第一测量结果和第二测量结果的数量每达到固定数量,执行一次批量生成训练数据和标签数据的操作。
结合上述实施例,基站可以先将第一测量结果和第二测量结果中包括的逻辑ID均转换为物理ID,然后再生成训练数据和标签数据。
作为示例,如图7所示,图7是本申请实施例提供的一种训练数据采集过程中构建训练数据集的方法的示例性示意图。
图7中的表A为多个第一测量结果,即对训练波束集的多个测量结果。具体的,表A中的每一行表示一个第一测量结果,每个第一测量结果中包括一个训练波束集的第一测量时间戳和该训练波束集的测量数据,该训练波束集的测量数据中包括该训练波束集中多个波束的测量数据,训练波束集的每个波束的测量数据可通过该波束的逻辑ID进行标识,该测量数据具体可以为L1-RSRP。
需要说明的是,表A中将一个训练波束集中的多个波束的逻辑ID与测量数据简单表示为“逻辑ID/测量数据”,实际上,一个第一测量时间戳对应于训练波束集中每个波束的逻辑ID与测量数据。例如,若训练波束集中包括4个波束,分别为波束1-4,则一个第一测量时间戳对应的测量数据包括:波束1/
测量数据1、波束2/测量数据2、波束3/测量数据3以及波束4/测量数据4。
同样地,图7中表B为多个第二测量结果,即对工作波束集的多个测量结果。具体地,表B中的每一行表示一个第二测量结果,每个第二测量结果中包括一个工作波束集的第二测量时间戳和该工作波束集的测量数据,该工作波束集的测量数据中包括预设数量个工作波束的测量数据,每个工作波束的测量数据可通过该工作波束的逻辑ID进行标识。
需要说明的是,表B中也将一个工作波束集中的多个工作波束的逻辑ID与测量数据简单表示为“逻辑ID/测量数据”,实际上,一个第二测量时间戳对应于工作波束集中每个工作波束的逻辑ID与测量数据。例如,若工作波束集中包括5个工作波束,分别为波束7、10、11、14、20,则一个第二测量时间戳对应的测量数据包括:波束7/测量数据7、波束10/测量数据10、波束11/测量数据11、波束14/测量数据14以及波束20/测量数据20。
图7的表A和表C中的T11、T12至T1N表示N个第一测量时间戳,表B和表D中的T21、T22至T2M表示M个第二测量时间戳。
基站可将表A中的各逻辑ID转换为物理ID,从而表A被转换为表C;且可将表B中的各逻辑ID转换为物理ID,从而表B被转换为表D。
进而,基站可从表C和表D中查找第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的第一测量结果和个第二测量结果。
作为示例,可以将第一测量时间戳表示为T1i,将第二测量时间戳表示为T2j,将预设时间门限表示为ReportTgap(上报的时间门限),则基站可以查找满足|T1i-T2j|<ReportTgap的测量结果组合。然后可以基于每个测量结果组合生成一组训练数据和一组标签数据,从而得到具有标签的训练数据集,即图7中的表E。
可以看出,图7的表E中的测量时间戳为表C中各训练数据的第一测量时间戳,表E中位于同一行的训练数据和标签数据来自于同一测量结果组合。
需要说明的是,图7的表E中的测量时间戳用于标识不同组的训练数据和标签数据,本申请实施例中也可用其他形式的ID区分不同组的训练数据和标签数据,本申请实施例对此不作限定。
基站在构建具有标签的训练数据集后,可向操作管理维护(Operation Administration and Maintenance,OAM)设备上报该训练数据集,上报的格式可以为【测量时间戳,训练数据,标签数据】,该测量时间戳也可由其他形式的ID代替。
各基站通过统一的格式向OAM设备上报训练数据集,可以使得训练数据集的通用性更强,OAM无需对接收到的训练数据集进行格式统一等操作,可以提高OAM训练波束预测模型的效率。
在本申请的一些实施例中,可以通过OAM设备触发基站启动上述实施例中波束测量的流程,基于此,基站在训练波束集的各波束方向上向终端发送用于波束测量的导频信号之前,该方法包括:
接收OAM设备发送的系统间消息,系统间消息包括第一字段和第二字段,第一字段用于指示启动波束测量,第二字段用于承载预设时间门限。
其中,预设时间门限是人工根据经验在OAM设备上预先配置的,人工配置完成后,可以人工触发OAM设备向基站下发系统间消息的操作,也可以由OAM设备自行触发向基站下发系统间消息的操作。
其中,OAM设备下发的系统间消息可以为RESTful超文本传输协议发布(HyperText Transfer Protocol POST,HTTP POST)消息格式,RESTful是一种网络应用程序的设计风格和开发方式。可以通过该消息中的notifyBeamTraining(波束训练通知)字段表示启动上述波束测量流程,通过ReportTgap
字段表示预设时间门限。具体配置如表2所示:
表2
表2最后一列中的“S”表示参数可选类型,取值为M时表示对应的参数为必选参数,取值为O时,表示对应的参数为可选参数。
在接收到上述系统间消息后,基站响应于系统间消息,向终端发送第一测量控制消息,第一测量控制消息用于指示终端在第一指定时频位置对训练波束集进行波束测量。然后接收终端回复的第一确认消息,第一确认消息用于表示终端支持对训练波束集进行波束测量。
可以理解的是,本申请实施例中的波束测量用于采集训练数据,为了与常规的波束测量进行区分,可以在第一测量控制消息中添加一个预设指示参数。即第一测量控制消息中包括预设指示参数,当预设指示参数的取值为第一值时,该预设指示参数用于指示采集波束预测模型的训练数据,即第一测量控制消息指示基站进行波束测量,并将波束测量结果作为波束预测模型的训练数据。
可选地,该第一测量控制消息可以被携带在第三代合作伙伴计划(3rd Generation Partnership Project,3GPP)的RRC消息中的信道状态信息测量配置(Channel State Information-Measurement configuration,CSI-MeasConfig)中,在CSI-MeasConfig中添加的指示参数可以为一个布尔型(BOOLEAN)变量,该布尔型变量的名称具体可以为ML-TrainingData(机器学习训练数据),当ML-TrainingData取值为True时,表示此次测量用于机器学习的训练数据采集。
相应地,终端识别到第一测量控制消息该指示域的取值为True时,可判断自身是否支持测量用于波束预测模型的训练数据,若支持,则向基站回复第一确认消息,进而基站可按照指定时频资源在工作波束集的各波束方向上发送用于波束测量的导频信号。
类似的,基站在工作波束集的各波束方向上发送用于波束测量的导频信号之前,该方法还包括:
响应于系统间消息,向终端发送第二测量控制消息,第二测量控制消息用于指示终端在第二指定时频位置对工作波束集进行波束测量。
接收终端回复的第二确认消息,第二确认消息用于表示终端支持对工作波束集进行波束测量。
其中,第二测量控制消息中也包括预设指示参数,当预设指示参数的取值为第一值时,预设指示参
数用于指示采集波束预测模型的训练数据。作为示例,该第一值可以为“True”。第二测量控制消息与第一测量控制消息的结构相同,可参考上述实施例中对第一测量控制消息的相关描述。
可以理解的是,基站接收到第二确认消息后,即可按照第二指定时频位置,在工作波束集的各波束方向上发送用于波束测量的导频信号。
需要说明的是,本申请实施例中对基站在训练波束集的各波束方向上发送导频信号,与在工作波束集的各波束方向上发送导频信号的先后顺序不作具体限定,发送的先后顺序取决于训练波束集和工作波束集的时频资源。
可选地,在上述实施例的基础上,为了提高获取训练数据的效率和获取到的训练数据的质量,可以在训练数据采集过程中对工作波束集的规模进行动态调整,且可以对训练波束集和工作波束集的测量周期进行动态调整。基于此,本申请实施例提供的训练数据采集方法如图8所示,图8中以基站与终端1和终端2通信为例进行说明,该方法包括:
S801、基站获取终端1对训练波束集和工作波束集的测量结果。
S802、基站获取终端2对训练波束集和工作波束集的测量结果。
需要说明的是,基站获取各终端反馈的训练波束集和工作波束集的测量结果的方法可参考上述实施例中的相关介绍。
S803、根据接收到的测量结果,确定终端1的移动速度和波束变化情况以及终端2的移动速度和波束变化情况。
基站可基于终端1上报的测量结果,调整终端1需测量的工作波束的数量以及对训练波束集和工作波束集的测量周期。
并且,基站可基于终端2上报的测量结果,调整终端2需测量的工作波束的数量以及对训练波束集和工作波束集的测量周期。
具体调整方法将在后续实施例中进行详细介绍,图8中以减少终端1需测量的工作波束数量,延长终端1对训练波束集和工作波束集的测量周期,且增加终端2需测量的工作波束数量,减少终端2对训练波束集和工作波束集的测量周期为例进行说明。
S804、配置终端1测量较少的工作波束,且延长终端1对训练波束集和工作波束集的测量周期。
S805、配置终端2测量较多的工作波束,且减少终端2对训练波束集和工作波束集的测量周期。
图8中终端1下方的黑色矩形代表为终端1配置的1个测量周期,终端2下方的3个黑色矩形代表为终端2配置的3个测量周期,可以看出,为终端1配置的单个测量周期的时长较长,为终端2配置的单个测量周期的时长较短。
S806、终端2在每个测量周期结束后上报测量结果。
S807、终端1在每个测量周期结束后上报测量结果。
以下分别对训练数据采集过程中基站动态调整终端需测量的工作波束的数量的方法以及基站动态调整终端对工作波束集和训练波束集的测量周期的方法进行介绍。
在本申请实施例中,基站首次对工作波束集中的各波束进行测量时,工作波束集中包括全量工作波束。在接收到终端发送的对训练波束集的第一测量结果以及对工作波束集的第二测量结果后,基站可对工作波束集中包括的波束进行更新,具体方法为:
若终端的移动速度小于等于预设速度阈值,则将工作波束集包括的波束更新为指定范围的波束,指定范围为指定俯仰角范围和指定方位角范围中的工作波束。
其中,指定俯仰角范围为将预设数量的工作波束的俯仰角范围按第一预设步长扩大后得到的俯仰角范围,指定方位角范围为将预设数量的工作波束的方位角范围按第二预设步长扩大后得到的方位角范围。
如图9所示,该方法具体可以包括以下步骤:
S901、启动。
S902、获取终端对工作波束集的第二测量结果和终端移动速度。
S903、获取第二测量结果对应的俯仰角范围Δθ和方位角范围
结合上述实施例中的介绍,第二测量结果中包括预设数量个工作波束的测量数据,第二测量结果对应的俯仰角范围Δθ是指预设数量个工作波束的俯仰角范围,方位角范围是指预设数量个工作波束的方位角范围。
S904、判断移动速度是否大于门限值Vth。
若是,则执行S906;若否,则执行S905。
其中,移动速度门限值Vth可以由基站根据基站所在的环境和基站的性能要求进行配置。
S905、配置终端测量Δθ+θstep和范围内的工作波束。
也就是说,可以将工作波束集更新为包括Δθ+θstep和范围内的工作波束。
其中,θstep和分别为第一预设步长和第二预设步长,可以由基站根据环境和基站的性能要求进行配置,例如,若基站在室内,由于室内遮挡物较多,基站可以将θstep和设置较大的值;若基站在室外,由于室外可以对波束进行遮挡的遮挡物较少,基站可以将θstep和设置为较小的值。
再例如,若基站性能要求较高,基站可以将θstep和设置为较大的值,以使基站获取更多的工作波束测量结果,以提高波束预测模型性能;若基站性能要求较低,基站可以将θstep和设置为较小的值,以降低导频开销。
S906、配置终端测量所有的工作波束。
也就是说,不对工作波束集进行更新。
S907、结束。
采用该方法,基站可根据终端的移动速度和预设数量个工作波束对应的俯仰角范围和方位角范围灵活调整工作波束集,由于预设数量个工作波束是终端测量得到的信号质量较好的波束,所以在终端移动速度较慢的情况下,大概率下一次终端测量得到的信号质量最好的工作波束还在上述俯仰角和方位角范围内。所以通过上述调整可以在保证测量准确性的基础上,减少不必要的测量开销。
以下对训练数据采集过程中调整训练波束集和工作波束集的测量周期的方法进行介绍,在接收终端发送的对训练波束集的第一测量结果,以及对工作波束集的第二测量结果之后,该方法还包括:
基于基站的位置、终端的位置和终端的移动速度,计算终端的俯仰角变化速率和方位角变化速率,然后计算基站的波束在俯仰角维度的第一变化周期和在方位角维度的第二变化周期,进而将第一变化周期和第二变化周期中较小的一个变化周期与预设系数的比值,作为训练波束集和工作波束集更新后的测量周期。
其中,第一变化周期为俯仰角维度的波束宽度与俯仰角变化速率的比值,第二变化周期为方位角维度的波束宽度与方位角变化速率的比值。
如图10a所示,该方法具体包括以下步骤:
S1001、启动。
S1002、获取终端位置和移动速度。
S1003、根据终端位置和基站位置,终端的移动速度v,计算终端的俯仰角变化速率θ’和方位角变化速率
根据终端位置和基站位置,可以确定基站与终端之间的几何位置关系,基于该几何位置关系和终端的移动速度,可以利用已有的数学算法计算出俯仰角变化速率θ,和方位角变化速率
作为示例,如图10b所示,图10b左侧为基站,右侧为终端,假设基站的位置为(x,y),基站高度为h,终端在t1时刻的位置为(x1,y1),终端高度为h1。
在t1时刻终端的俯仰角为θ1,当h大于h1时,终端的俯仰角为θ1,θ1可以通过以下公式计算:
当h小于h1时,终端的俯仰角为θ1可以通过以下公式计算:
同理,若终端在t2时刻的位置为(x2,y2),终端高度为h2,可计算终端在t2时刻的俯仰角θ2,当h大于h2时,则终端的俯仰角θ2可以通过以下公式计算:
当h小于h2时,终端的俯仰角θ2可以通过以下公式计算:
则基站根据计算得到的θ1和θ2,可以计算得到俯仰角变化速率θ’:
θ,=|θ2-θ1|/Δt
其中,Δt为t2和t1的差值。
进而,基站计算终端的方位角变化速率仍以图10b为例,基站的位置以(x,y)表示,基站的高度为h;t1时刻终端的位置为(x1,y1),终端的高度为h1,则终端在t1时刻的方位角可以通过以下公式计算:
同理,t2时刻终端的位置为(x2,y2),终端的高度为h2,则终端在t2时刻的方位角可以通过以下公式计算:
则基站根据计算得到的和可以计算得到方位角变化速率
其中,Δt为t2和t1的差值。
作为示例,若确定终端在以基站为圆心围基站转圈,则终端的俯仰角变化率为0,此时终端的角速度可等同于终端的方位角变化速率。
S1004、获取基站的波束在俯仰角维度的波束宽度θmax和方位角维度的波束宽度
其中,基站的波束在俯仰角维度的波束宽度θmax为各工作波束在俯仰方向的波束主瓣两个半功率点间的夹角,基站的波束在方位角维度的波束宽度为各工作波束的在方位方向的波束主瓣两个半功率角间的夹角。
S1005、计算第一变化周期第二变化周期Tθ=θmax/θ’。
S1006、选择Tθ和中较小的值作为T。
其中,为了满足实际业务需求,所以将Tθ和中较小的值作为T,以使选择的T可以同时适用于俯仰角变化和方位角变化。
S1007、将训练波束集和工作波束集的测量周期确定为T/m。
其中,为了进行更细的时间粒度的测量,基站可以根据环境和性能要求,配置预设系数m。例如,性能要求较高时,可以将m配置为较大的值,以使最终测量周期更小,测量更精细。
需要说明的是,基站可以将更新后的测量周期配置在3GPP的RRC消息的信道状态信息上报的配置(Channel State Information-Report configuration,CSI-ReportConfig)中,具体可以配置在信道状态信息上报的周期和偏移(Channel State Information-ReportPeriodicityAndOffset,CSI-ReportPeriodicit yAndOffset)中。
S1008、结束。
采用该方法,基站计算得到的第一变化周期和第二变化周期可以反应波束变化的快慢,确定第一变化周期和第二变化周期中较小的值,利用确定的第一变化周期和第二变化周期中较小的值和预设系数的比值作为测量周期,如此,使计算得到的测量周期更加精细可控,且更加贴合实际测量需求,可以在保证测量得到的测量结果准确丰富的前提下,降低测量开销,且可以避免采集大量差异较小的训练数据,可以提高后续生成的训练数据的质量。
基于相同的技术构思,本申请实施例还提供一种训练数据采集方法,该方法应用于终端,如图11所示,该方法包括:
S1101、对基站在训练波束集的各波束方向上向终端发送的导频信号进行波束测量,对基站在工作波束集的各波束方向上向终端发送的导频信号进行波束测量。
S1102、向基站发送对训练波束集的第一测量结果,以及对工作波束集的第二测量结果。
其中,第一测量结果为训练数据的数据来源,第二测量结果为训练数据的标签数据的数据来源。
采用该方法,终端分别对基站在训练波束集的各波束方向上发送的导频信号进行波束测量,和在工作波束集的各波束方向上发送的导频信号进行波束测量,终端向基站反馈训练波束集的第一测量结果,以及对工作波束集的第二测量结果。第一测量结果和第二测量结果均为终端实际测量得到的,所以采用该方法生成的训练数据和标签数据均符合基站和终端之间的实际通信情况,即采用该方法可以准确地获取到用于训练波束预测模型的训练数据。
在上述实施例的基础上,在对基站在训练波束集的各波束方向上发送的导频信号进行波束测量之前,该方法还包括:
接收基站发送的第一测量控制消息。若终端支持对训练波束集进行波束测量,则向基站回复第一确认消息。
其中,第一测量控制消息用于指示终端在第一指定时频位置对训练波束集进行波束测量。
类似的,在对基站在工作波束集的各波束方向上发送的导频信号进行波束测量之前,该方法还包括:
接收基站发送的第二测量控制消息,第二测量控制消息用于指示终端在第二指定时频位置对工作波束集进行波束测量。若终端支持对工作波束集进行波束测量,则向基站回复第二确认消息。
其中,第一测量控制消息和第二测量控制消息中均包括预设指示参数,当预设指示参数的取值为第一值时,预设指示参数用于指示采集波束预测模型的训练数据。
在终端完成波束测量后,需要向基站发送测量结果,终端向基站发送对训练波束集的第一测量结果,
以及对工作波束集的第二测量结果,包括以下两种实现方式:
方式一、通过物理层信令向基站发送第一测量报告消息和第二测量报告消息,第一测量报告消息中包括第一测量结果,第二测量报告消息中包括第二测量结果。
方式二、通过高层信令向基站发送第一RRC消息和第二RRC消息,第一RRC消息中包括第一测量结果,第二RRC消息中包括第二测量结果。
其中,第一RRC消息和第二RRC消息的格式以及包括的具体内容可参考上述实施例中的相关描述,此处不再赘述。
以下以各设备间交互的角度对本申请实施例提供的训练数据采集方法进行介绍,如图12所示,该方法包括:
S1201、OAM设备向基站发送系统间消息。
其中,该系统间消息用于配置波束预测模型的测量参数。
S1202、基站向终端发送第一测量控制消息。
其中,第一测量控制消息用于指示终端在第一指定时频位置对训练波束集进行波束测量。关于第一测量控制消息可参考上述实施例中的描述,此处不再赘述。
S1203、终端向基站反馈第一确认消息。
其中,第一确认消息用于表示终端支持对训练波束集进行波束测量,本申请实施例对第一确认消息的形式不做具体限定。
S1204、基站向终端在训练波束集的各波束方向上发送用于波束测量的导频信号。
S1205、终端对训练波束集进行测量。
S1206、终端向基站反馈第一测量结果。
其中,终端可以通过物理层信令发送的第一测量报告消息向基站反馈第一测量结果,也可以通过高层信令发送的第一RRC消息反馈第一测量结果,第一测量报告消息和第一RRC的消息格式可参考上述实施例中的相关描述,此处不再赘述。
S1207、基站向终端发送第二测量控制消息。
其中,第二测量控制消息用于指示终端在第二指定时频位置对工作波束集进行波束测量,第二测量控制消息和第一测量控制消息的消息格式相同。
S1208、终端向基站反馈第二确认消息。
其中,第二确认消息用于表示终端支持对工作波束集进行波束测量,本申请实施例对第二确认消息的形式不做具体限定。
S1209、基站向终端在工作波束集的各波束方向上发送用于波束测量的导频信号。
S1210、终端对工作波束进行测量。
S1211、终端向基站反馈第二测量结果。
其中,终端可以通过物理层信令发送的第二测量报告消息向基站反馈第二测量结果,也可以通过高层信令发送的第二RRC消息反馈第二测量结果,具体第二测量报告消息和第二RRC消息格式参考上述实施例,此处不再赘述。
S1212、基站将逻辑ID转换为物理ID,对训练波束集的各波束打标签。
将训练波束集的各波束的逻辑ID转换为物理ID的方法以及对训练波束集的各波束打标签的方法参考上述实施例,此处不再赘述。
S1213、基站向OAM设备发送打完标签的训练数据集。
S1214、OAM设备基于打完标签的训练数据集对模型进行训练,得到波束预测模型。
进而,在利用本申请实施例的训练数据采集方法采集得到的训练数据得到波束预测模型后,可以利用该波束预测模型确定最佳波束,如图13所示,该方法包括:
S1301、基站进行稀疏波束扫描。
稀疏波束扫描是指基站在稀疏波束的波束方向上向终端发送导频信号。
稀疏波束是指全量工作波束中的部分波束,稀疏波束的数量与波束预测模型的输入端需被输入的测量数据的数量相同。例如,波束预测模型的输入端需被输入16个波束的测量数据,则稀疏波束包括16个波束。
图13中16个具有黑色填充的小方格即为此次扫描的16个波束。
S1302、终端向基站反馈信号强度。
终端具体可向基站反馈每个导频信号的信号强度,即对每个波束的测量结果。该信号强度具体可以为L-RSRP。
S1303、基站基于波束预测模型进行波束推理。
其中,波束预测模型可以根据较少的波束的信号强度,在所有的波束中,预测性能最佳的前K个波束的物理ID和信号强度。
S1304、基站向终端进行Top-K波束的信道状态信息参考信号(Channel-state information Reference Signal,CSI-RS)波束扫描。
其中,Top-K波束即为上述性能最佳的前K个波束。
作为示例,K可以为5,Top-K波束即为图13中间的方格阵列中具有填充的5个波束。
S1305、终端确定最佳CSI-RS波束。
最佳CSI-RS是指终端测量得到的信道状态最好的波束,该信道状态的好坏可通过信号强度衡量。
S1306、终端向基站反馈最佳CSI-RS ID。
其中,最佳CSI-RS ID是指终端测量到的信道状态最好的波束对应的CSI-RS ID。
S1307、基站利用最佳CSI-RS波束传输。
其中,最佳CSI-RS ID是指最佳CSI-RS ID对应的波束。
作为示例,图13中最下方的方格阵列中,具有填充的小方块表示最佳CSI-RS波束。
基于相同的技术构思,本申请实施例提供一种训练数据采集装置,该装置应用于基站,如图14所示,该装置包括:
发送模块1401,用于在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向终端发送用于波束测量的导频信号;
接收模块1402,用于接收终端发送的对训练波束集的第一测量结果,以及对工作波束集的第二测量结果;
生成模块1403,用于基于第一测量结果生成训练数据,基于第二测量结果生成训练数据对应的标签数据。
可选的,第一测量结果中包括第一测量时间戳,第二测量结果中包括第二测量时间戳;第一测量时间戳为终端对训练波束集测量的起始时间戳,第二测量时间戳为终端对工作波束集测量的起始时间戳;
生成模块1403,具体用于:
若第一测量时间戳和第二测量时间戳之间的差值绝对值小于预设时间门限,则将第一测量结果包括的测量数据作为训练数据,将第二测量结果包括的测量数据作为训练数据对应的标签数据。
可选的,第一测量结果中包括第一测量时间戳,第二测量结果中包括第二测量时间戳,第一测量时间戳为终端对训练波束集测量的起始时间戳,第二测量时间戳为终端对工作波束集测量的起始时间戳;
生成模块1403,具体用于:
针对已接收的多个第一测量结果和多个第二测量结果,将第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的一个第一测量结果和一个第二测量结果作为一个测量结果组合;
针对每个测量结果组合,将该测量结果组合中的第一测量结果包括的测量数据作为训练数据,将该测量结果组合中的第二测量结果包括的测量数据作为标签数据。
可选的,该装置还包括:
转换模块,用于将第一测量结果包括的每个波束的逻辑ID转换为物理ID,将第二测量结果包括的每个波束的逻辑ID转换为物理ID,基站的各波束的物理ID为基站基于各自波束的俯仰角和方位角确定出的。
可选的,发送模块1401,具体用于:
在训练波束集的各波束方向上向终端发送用于波束测量的终端专用导频信号,在工作波束的各波束方向上向终端上发送用于波束测量的终端专用导频信号;或者,
在训练波束集的各波束方向上发送用于波束测量的小区公共导频信号,在工作波束的各波束方向上向终端发送用于波束测量的终端专用导频信号。
可选的,接收模块1402,具体用于:
接收终端通过物理层信令发送的第一测量报告消息和第二测量报告消息,第一测量报告消息中包括第一测量结果,第二测量报告消息中包括第二测量结果;或者,
接收终端通过高层信令发送的第一RRC消息和第二RRC消息,第一RRC消息中包括第一测量结果,第二RRC消息中包括第二测量结果。
可选的,第一RRC消息和第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
消息类型字段用于指示训练波束集或工作波束集中的波束,测量时间字段用于表示波束测量的起始时间戳;
第一RRC消息中的波束数目字段的取值为训练波束集中包括的波束数量,第二RRC消息中的波束数目字段的取值为预设数量,预设数量小于工作波束集中包括的波束数量。
可选的,第一RRC消息和第二RRC消息中还包括终端位置字段和终端速度字段;终端位置字段用于表示终端所处的地理位置,终端速度字段用于表示终端在波束测量过程中的移动速度。
可选的,接收模块,还用于接收OAM设备发送的系统间消息,系统间消息包括第一字段和第二字段,第一字段用于指示启动波束测量,第二字段用于承载预设时间门限;
发送模块1401,用于响应于接收OAM设备发送的系统间消息,向终端发送第一测量控制消息,第一测量控制消息用于指示终端在第一指定时频位置对训练波束集进行波束测量;
接收模块1402,用于接收终端回复的第一确认消息,第一确认消息用于表示终端支持对训练波束集进行波束测量。
可选的,发送模块1401,还用于响应于系统间消息,向终端发送第二测量控制消息,第二测量控
制消息用于指示终端在第二指定时频位置对工作波束集进行波束测量;
接收模块1402,还用于接收终端回复的第二确认消息,第二确认消息用于表示终端支持对工作波束集进行波束测量。
可选的,第一测量控制消息和第二测量控制消息中均包括预设指示域,当预设指示域的取值为第一值时,预设指示域用于指示采集波束预测模型的训练数据。
可选的,第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括预设数量个工作波束的逻辑ID,预设数量个工作波束为按照信号质量从高到低的顺序,从工作波束集中选择的前预设数量个工作波束。
可选的,装置还包括:
更新模块,用于若终端的移动速度小于等于预设速度阈值,则将工作波束集包括的波束更新为指定范围的波束,指定范围为指定俯仰角范围和指定方位角范围中的工作波束;
其中,指定俯仰角范围为将预设数量的工作波束的俯仰角范围按第一预设步长扩大后得到的俯仰角范围,指定方位角范围为将预设数量的工作波束的方位角范围按第二预设步长扩大后得到的方位角范围。
可选的,该装置还包括:
计算模块,用于基于基站的位置、终端的位置和终端的移动速度,计算终端的俯仰角变化速率和方位角变化速率;
计算模块,还用于计算基站的波束在俯仰角维度的第一变化周期和在方位角维度的第二变化周期,第一变化周期为俯仰角维度的波束宽度与俯仰角变化速率的比值,第二变化周期为方位角维度的波束宽度与方位角变化速率的比值;
更新模块,还用于将第一变化周期和第二变化周期中较小的一个变化周期与预设系数的比值,作为训练波束集和工作波束集更新后的测量周期。
基于相同的技术构思,本申请实施例还提供一种训练数据采集装置,该装置应用于终端,如图15所示,该装置包括:
测量模块1501,用于对基站在训练波束集的各波束方向上向终端发送的导频信号进行波束测量,对基站在工作波束集的各波束方向上向终端发送的导频信号进行波束测量;
发送模块1502,用于向基站发送对训练波束集的第一测量结果,以及对工作波束集的第二测量结果,第一测量结果为训练数据的数据来源,第二测量结果为训练数据的标签数据的数据来源。
可选的,发送模块1502,具体用于:
通过物理层信令向基站发送第一测量报告消息和第二测量报告消息,第一测量报告消息中包括第一测量结果,第二测量报告消息中包括第二测量结果;或者,
通过高层信令向基站发送第一RRC消息和第二RRC消息,第一RRC消息中包括第一测量结果,第二RRC消息中包括第二测量结果。
可选的,第一RRC消息和第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
消息类型字段用于指示训练波束集或工作波束集中的波束,测量时间字段用于表示波束测量的起始时间戳;
第一RRC消息中的波束数目字段的取值为训练波束集中包括的波束数量,第二RRC消息中的波束
数目字段的取值为预设数量,预设数目小于工作波束集中包括的波束数量。
可选的,第一RRC消息和第二RRC消息中还包括终端位置字段和终端速度字段;终端位置字段用于表示终端所处的地理位置,终端速度字段用于表示终端在波束测量过程中的移动速度。
可选的,该装置还包括:
接收模块,用于接收基站发送的第一测量控制消息,第一测量控制消息用于指示终端在第一指定时频位置对训练波束集进行波束测量;
若终端支持对训练波束集进行波束测量,则发送模块1502向基站回复第一确认消息;
接收模块,还用于接收基站发送的送第二测量控制消息,第二测量控制消息用于指示终端在第二指定时频位置对工作波束集进行波束测量;
若终端支持对工作波束集进行波束测量,则发送模块1502向基站回复第二确认消息。
可选的,第一测量控制消息和第二测量控制消息中均包括预设指示参数,当预设指示参数的取值为第一值时,预设指示参数用于指示采集波束预测模型的训练数据。
可选的,第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括预设数量个工作波束的逻辑ID,预设数量个工作波束为终端按照信号质量从高到低的顺序,从工作波束集中选择的前预设数量个工作波束。
基于相同的技术构思,本申请实施例还提供一种基站,如图16所示,包括:处理器1601;收发器1604;
机器可读存储介质1602,机器可读存储介质1602存储有能够被处理器1601执行的机器可执行指令;机器可执行指令促使处理器1601执行以下步骤:
通过收发器1604在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向终端发送用于波束测量的导频信号;
通过收发器1604接收终端发送的对训练波束集的第一测量结果,以及对工作波束集的第二测量结果;
基于第一测量结果生成训练数据,基于第二测量结果生成训练数据对应的标签数据。
可选的,第一测量结果中包括第一测量时间戳,第二测量结果中包括第二测量时间戳,第一测量时间戳为终端对训练波束集测量的起始时间戳,第二测量时间戳为终端对工作波束集测量的起始时间戳;
机器可执行指令还促使处理器1601执行以下步骤:
若第一测量时间戳和第二测量时间戳之间的差值绝对值小于预设时间门限,则将第一测量结果包括的测量数据作为训练数据,将第二测量结果包括的测量数据作为训练数据对应的标签数据。
可选的,第一测量结果中包括第一测量时间戳,第二测量结果中包括第二测量时间戳;
机器可执行指令还促使处理器1601执行以下步骤:
针对已接收的多个第一测量结果和多个第二测量结果,将第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的一个第一测量结果和一个第二测量结果作为一个测量结果组合;
针对每个测量结果组合,将该测量结果组合中的第一测量结果包括的测量数据作为训练数据,将该测量结果组合中的第二测量结果包括的测量数据作为标签数据。
可选的,机器可执行指令还促使处理器1601执行以下步骤:
将第一测量结果包括的每个训练的逻辑ID转换为物理ID,将第二测量结果包括的每个波束的逻辑
ID转换为物理ID,基站的各波束的物理ID为基站基于各自波束的俯仰角和方位角确定出的。
可选的,机器可执行指令促使处理器1601执行以下步骤:
通过收发器1604在训练波束集的各波束方向上向终端发送用于波束测量的终端专用导频信号,在工作波束的各波束方向上向终端发送用于波束测量的终端专用导频信号;或者,
通过收发器1604在训练波束集的各波束方向上发送用于波束测量的小区公共导频信号,在工作波束的各波束方向上向终端发送用于波束测量的终端专用导频信号。
可选的,机器可执行指令还促使处理器1601执行以下步骤:
通过收发器1604接收终端通过物理层信令发送的第一测量报告消息和第二测量报告消息,第一测量报告消息中包括第一测量结果,第二测量报告消息中包括第二测量结果;或者,
通过收发器1604接收终端通过高层信令发送的第一RRC消息和第二RRC消息,第一RRC消息中包括第一测量结果,第二RRC消息中包括第二测量结果。
可选的,第一RRC消息和第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
消息类型字段用于指示训练波束集或工作波束集中的波束,测量时间字段用于表示波束测量的起始时间戳;
第一RRC消息中的波束数目字段的取值为训练波束集中包括的波束数量,第二RRC消息中的波束数目字段的取值为预设数量,预设数量小于工作波束集中包括的波束数量。
可选的,第一RRC消息和第二RRC消息中还包括终端位置字段和终端速度字段;终端位置字段用于表示终端所处的地理位置,终端速度字段用于表示终端在波束测量过程中的移动速度。
可选的,机器可执行指令还促使处理器1601执行以下步骤:
通过收发器1604接收OAM设备发送的系统间消息,系统间消息包括第一字段和第二字段,第一字段用于指示启动波束测量,第二字段用于承载预设时间门限;
响应于系统间消息,通过收发器1604向终端发送第一测量控制消息,第一测量控制消息用于指示终端在第一指定时频位置对训练波束集进行波束测量;
通过收发器1604接收终端回复的第一确认消息,第一确认消息用于表示终端支持对训练波束集进行波束测量。
可选的,机器可执行指令还促使处理器1601执行以下步骤:
响应于系统间消息,通过收发器1604向终端发送第二测量控制消息,第二测量控制消息用于指示终端在第二指定时频位置对工作波束集进行波束测量;
通过收发器1604接收终端回复的第二确认消息,第二确认消息用于表示终端支持对工作波束集进行波束测量。
可选的,第一测量控制消息和第二测量控制消息中均包括预设指示参数,当预设指示参数的取值为第一值时,预设指示参数用于指示采集波束预测模型的训练数据。
可选的,第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括预设数量个工作波束的逻辑ID,预设数量个工作波束为终端按照信号质量从高到低的顺序,从工作波束集中选择的前预设数量个工作波束。
可选的,机器可执行指令还促使处理器1601执行以下步骤:
若终端的移动速度小于等于预设速度阈值,则将工作波束集包括的波束更新为指定范围的波束,指定范围为指定俯仰角范围和指定方位角范围中的工作波束;
其中,指定俯仰角范围为将预设数量的工作波束的俯仰角范围按第一预设步长扩大后得到的俯仰角范围,指定方位角范围为将预设数量的工作波束的方位角范围按第二预设步长扩大后得到的方位角范围。
可选的,机器可执行指令还促使处理器1601执行以下步骤:
基于基站的位置、终端的位置和终端的移动速度,计算终端的俯仰角变化速率和方位角变化速率;
计算基站的波束在俯仰角维度的第一变化周期和在方位角维度的第二变化周期,第一变化周期为俯仰角维度的波束宽度与俯仰角变化速率的比值,第二变化周期为方位角维度的波束宽度与方位角变化速率的比值;
将第一变化周期和第二变化周期中较小的一个变化周期与预设系数的比值,作为训练波束集和工作波束集更新后的测量周期。
在图16中,还可以包括通信总线1603。处理器1601、机器可读存储介质1602及收发器1604之间通过通信总线1603完成相互间的通信,通信总线1603可以是外设部件互连标准(Peripheral Component Interconnect,PCI)总线或扩展工业标准结构(Extended Industry Standard Architecture,EISA)总线等。该通信总线可以分为地址总线、数据总线、控制总线等。
收发器1604可以为无线通信模块,收发器1604在处理器1601的控制下,与其他设备进行数据交互。
机器可读存储介质1602可以包括随机存取存储器(Random Access Memory,RAM),也可以包括非易失性存储器(Non-Volatile Memory,NVM),例如至少一个磁盘存储器。另外,机器可读存储介质还可以是至少一个位于远离前述处理器的存储装置。
处理器1601可以是通用处理器,包括中央处理器(Central Processing Unit,CPU)、网络处理器(Network Processor,NP)等;还可以是数字信号处理器(Digital Signal Processing,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现场可编程门阵列(Field-Programmable Gate Array,FPGA)或其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
本申请实施例提供一种终端,如图17所示,该终端包括:处理器1701;收发器1704;
机器可读存储介质1702,机器可读存储介质1702存储有能够被处理器1701执行的机器可执行指令;机器可执行指令促使处理器1701执行以下步骤:
对基站在训练波束集的各波束方向上向终端发送的导频信号进行波束测量,对基站在工作波束集的各波束方向上向终端发送的导频信号进行波束测量;
通过收发器1704向基站发送对训练波束集的第一测量结果,以及对工作波束集的第二测量结果,第一测量结果为训练数据的数据来源,第二测量结果为训练数据的标签数据的数据来源。
可选的,机器可执行指令促使处理器1701执行以下步骤:
通过收发器1704通过物理层信令向基站发送第一测量报告消息和第二测量报告消息,第一测量报告消息中包括第一测量结果,第二测量报告消息中包括第二测量结果;或者,
通过收发器1704通过高层信令向基站发送第一RRC消息和第二RRC消息,第一RRC消息中包括第一测量结果,第二RRC消息中包括第二测量结果。
可选的,第一RRC消息和第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;
消息类型字段用于指示训练波束集或工作波束集中的波束,测量时间字段用于表示波束测量的起始时间戳;
第一RRC消息中的波束数目字段的取值为训练波束集中包括的波束数量,第二RRC消息中的波束数目字段的取值为预设数量,预设数目小于工作波束集中包括的波束数量。
可选的,第一RRC消息和第二RRC消息中还包括终端位置字段和终端速度字段;终端位置字段用于表示终端所处的地理位置,终端速度字段用于表示终端在波束测量过程中的移动速度。
可选的,机器可执行指令还促使处理器1701执行以下步骤:
通过收发器1704接收基站发送的第一测量控制消息,第一测量控制消息用于指示终端在第一指定时频位置对训练波束集进行波束测量;
若终端支持对训练波束集进行波束测量,则通过收发器1704向基站回复第一确认消息;
机器可执行指令还促使处理器1701执行以下步骤:
通过收发器1704接收基站发送的送第二测量控制消息,第二测量控制消息用于指示终端在第二指定时频位置对工作波束集进行波束测量;
若终端支持对工作波束集进行波束测量,则通过收发器1704向基站回复第二确认消息。
可选的,第一测量控制消息和第二测量控制消息中均包括预设指示参数,当预设指示参数的取值为第一值时,预设指示参数用于指示采集波束预测模型的训练数据。
可选的,第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;
第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括预设数量个工作波束的逻辑ID,预设数量个工作波束为终端按照信号质量从高到低的顺序,从工作波束集中选择的前预设数量个工作波束。
在图17中,还可以包括通信总线1703。处理器1701、机器可读存储介质1702及收发器1704之间通过通信总线1703完成相互间的通信,通信总线1703可以是外设部件互连标准(Peripheral Component Interconnect,PCI)总线或扩展工业标准结构(Extended Industry Standard Architecture,EISA)总线等。该通信总线可以分为地址总线、数据总线、控制总线等。
收发器1704可以为无线通信模块,收发器1704在处理器1701的控制下,与其他设备进行数据交互。
机器可读存储介质1702可以包括随机存取存储器(Random Access Memory,RAM),也可以包括非易失性存储器(Non-Volatile Memory,NVM),例如至少一个磁盘存储器。另外,机器可读存储介质还可以是至少一个位于远离前述处理器的存储装置。
处理器1701可以是通用处理器,包括中央处理器(Central Processing Unit,CPU)、网络处理器(Network Processor,NP)等;还可以是数字信号处理器(Digital Signal Processing,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现场可编程门阵列(Field-Programmable Gate Array,FPGA)或其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
基于同一种发明构思,根据上述本申请实施例提供的训练数据采集方法,本申请实施例还提供了一种机器可读存储介质,机器可读存储介质存储有能够被处理器执行的机器可执行指令。处理器被机器可执行指令促使实现上述任一训练数据采集方法的步骤。
在本申请提供的又一实施例中,还提供了一种包含指令的计算机程序产品,当其在计算机上运行时,使得计算机执行上述实施例中任一训练数据采集方法的步骤。
需要说明的是,在本文中,诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来,而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且,术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
本说明书中的各个实施例均采用相关的方式描述,各个实施例之间相同相似的部分互相参见即可,每个实施例重点说明的都是与其他实施例的不同之处。尤其,对于段标识确定装置、设备及机器可读存储介质。机器可读存储介质实施例而言,由于其基本相似于段标识确定方法实施例,所以描述的比较简单,相关之处参见段标识确定方法实施例的部分说明即可。
以上所述仅为本发明的较佳实施例,并不用以限制本发明,凡在本发明的精神和原则之内,所做的任何修改、等同替换、改进等,均应包含在本发明保护的范围之内。
Claims (65)
- 一种训练数据采集方法,其特征在于,所述方法应用于基站,所述方法包括:在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号;接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果;基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据。
- 根据权利要求1所述的方法,其特征在于,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;所述基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据,包括:若所述第一测量时间戳和所述第二测量时间戳之间的差值绝对值小于预设时间门限,则将所述第一测量结果包括的测量数据作为训练数据,将所述第二测量结果包括的测量数据作为所述训练数据对应的标签数据。
- 根据权利要求1所述的方法,其特征在于,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;所述基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据,包括:针对已接收的多个第一测量结果和多个第二测量结果,将第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的一个第一测量结果和一个第二测量结果作为一个测量结果组合;针对每个测量结果组合,将该测量结果组合中的第一测量结果包括的测量数据作为训练数据,将该测量结果组合中的第二测量结果包括的测量数据作为标签数据。
- 根据权利要求1所述的方法,其特征在于,在所述基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据之前,所述方法还包括:将所述第一测量结果包括的每个波束的逻辑ID转换为物理ID,将所述第二测量结果包括的每个波束的逻辑ID转换为物理ID,所述基站的各波束的物理ID为所述基站基于各自波束的俯仰角和方位角确定出的。
- 根据权利要求1所述的方法,其特征在于,所述在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号,包括:在所述训练波束集的各波束方向上向所述终端发送用于波束测量的终端专用导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号;或者,在所述训练波束集的各波束方向上发送用于波束测量的小区公共导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号。
- 根据权利要求1或5所述的方法,其特征在于,所述接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,包括:接收所述终端通过物理层信令发送的第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,接收所述终端通过高层信令发送的第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
- 根据权利要求6所述的方法,其特征在于,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数量小于所述工作波束集中包括的波束数量。
- 根据权利要求7所述的方法,其特征在于,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
- 根据权利要求2所述的方法,其特征在于,在所述在训练波束集的各波束方向上向终端发送用于波束测量的导频信号之前,所述方法包括:接收OAM设备发送的系统间消息,所述系统间消息包括第一字段和第二字段,所述第一字段用于指示启动波束测量,所述第二字段用于承载所述预设时间门限;响应于所述系统间消息,向所述终端发送第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;接收所述终端回复的第一确认消息,所述第一确认消息用于表示所述终端支持对所述训练波束集进行波束测量。
- 根据权利要求9所述的方法,其特征在于,在所述在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号之前,所述方法还包括:响应于所述系统间消息,向所述终端发送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;接收所述终端回复的第二确认消息,所述第二确认消息用于表示所述终端支持对所述工作波束集进行波束测量。
- 根据权利要求10所述的方法,其特征在于,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
- 根据权利要求2或3所述的方法,其特征在于,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
- 根据权利要求8所述的方法,其特征在于,在所述接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果之后,所述方法还包括:若所述终端的移动速度小于等于预设速度阈值,则将所述工作波束集包括的波束更新为指定范围的波束,所述指定范围为指定俯仰角范围和指定方位角范围中的工作波束;其中,所述指定俯仰角范围为将所述预设数量的工作波束的俯仰角范围按第一预设步长扩大后得到 的俯仰角范围,所述指定方位角范围为将所述预设数量的工作波束的方位角范围按第二预设步长扩大后得到的方位角范围。
- 根据权利要求8所述的方法,其特征在于,在所述接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果之后,所述方法还包括:基于所述基站的位置、所述终端的位置和所述终端的移动速度,计算所述终端的俯仰角变化速率和方位角变化速率;计算所述基站的波束在俯仰角维度的第一变化周期和在方位角维度的第二变化周期,所述第一变化周期为所述俯仰角维度的波束宽度与所述俯仰角变化速率的比值,所述第二变化周期为所述方位角维度的波束宽度与所述方位角变化速率的比值;将所述第一变化周期和所述第二变化周期中较小的一个变化周期与预设系数的比值,作为所述训练波束集和所述工作波束集更新后的测量周期。
- 一种训练数据采集方法,其特征在于,所述方法应用于终端,所述方法包括:对基站在训练波束集的各波束方向上向所述终端发送的导频信号进行波束测量,对所述基站在工作波束集的各波束方向上向所述终端发送的导频信号进行波束测量;向所述基站发送对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,所述第一测量结果为训练数据的数据来源,所述第二测量结果为所述训练数据的标签数据的数据来源。
- 根据权利要求15所述的方法,其特征在于,所述向所述基站发送对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,包括:通过物理层信令向所述基站发送第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,通过高层信令向所述基站发送第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
- 根据权利要求16所述的方法,其特征在于,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数目小于所述工作波束集中包括的波束数量。
- 根据权利要求17所述的方法,其特征在于,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
- 根据权利要求15所述的方法,其特征在于,在所述对基站在训练波束集的各波束方向上发送的导频信号进行波束测量之前,所述方法还包括:接收所述基站发送的第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;若所述终端支持对所述训练波束集进行波束测量,则向所述基站回复第一确认消息;在所述对所述基站在工作波束集的各波束方向上发送的导频信号进行波束测量之前,所述方法还包括:接收所述基站发送的送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;若所述终端支持对所述工作波束集进行波束测量,则向所述基站回复第二确认消息。
- 根据权利要求19所述的方法,其特征在于,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示用于采集波束预测模型的训练数据。
- 根据权利要求15-20任一项所述的方法,其特征在于,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
- 一种训练数据采集装置,其特征在于,所述装置应用于基站,所述装置包括:发送模块,用于在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号;接收模块,用于接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果;生成模块,用于基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据。
- 根据权利要求22所述的装置,其特征在于,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;所述生成模块,具体用于:若所述第一测量时间戳和所述第二测量时间戳之间的差值绝对值小于预设时间门限,则将所述第一测量结果包括的测量数据作为训练数据,将所述第二测量结果包括的测量数据作为所述训练数据对应的标签数据。
- 根据权利要求22所述的装置,其特征在于,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;所述生成模块,具体用于:针对已接收的多个第一测量结果和多个第二测量结果,将第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的一个第一测量结果和一个第二测量结果作为一个测量结果组合;针对每个测量结果组合,将该测量结果组合中的第一测量结果包括的测量数据作为训练数据,将该测量结果组合中的第二测量结果包括的测量数据作为标签数据。
- 根据权利要求22所述的装置,所述装置还包括:转换模块,用于将所述第一测量结果包括的每个波束的逻辑ID转换为物理ID,将所述第二测量结果包括的每个波束的逻辑ID转换为物理ID,所述基站的各波束的物理ID为所述基站基于各自波束的俯仰角和方位角确定出的。
- 根据权利要求22所述的装置,其特征在于,所述发送模块,具体用于:在所述训练波束集的各波束方向上向所述终端发送用于波束测量的终端专用导频信号,在所述工作波束的各波束方向上向所述终端上发送用于波束测量的终端专用导频信号;或者,在所述训练波束集的各波束方向上发送用于波束测量的小区公共导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号。
- 根据权利要求22或26所述的装置,其特征在于,所述接收模块,具体用于:接收所述终端通过物理层信令发送的第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,接收所述终端通过高层信令发送的第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
- 根据权利要求27所述的装置,其特征在于,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数量小于所述工作波束集中包括的波束数量。
- 根据权利要求28所述的装置,其特征在于,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
- 根据权利要求23所述的装置,其特征在于,所述接收模块,还用于接收OAM设备发送的系统间消息,所述系统间消息包括第一字段和第二字段,所述第一字段用于指示启动波束测量,所述第二字段用于承载所述预设时间门限;所述发送模块,用于响应于接收OAM设备发送的系统间消息,向所述终端发送第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;所述接收模块,用于接收所述终端回复的第一确认消息,所述第一确认消息用于表示所述终端支持对所述训练波束集进行波束测量。
- 根据权利要求30所述的装置,其特征在于,所述发送模块,还用于响应于所述系统间消息,向所述终端发送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;所述接收模块,还用于接收所述终端回复的第二确认消息,所述第二确认消息用于表示所述终端支持对所述工作波束集进行波束测量。
- 根据权利要求31所述的装置,其特征在于,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
- 根据权利要求23或24所述的装置,其特征在于,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
- 根据权利要求29所述的装置,其特征在于,所述装置还包括:更新模块,用于若所述终端的移动速度小于等于预设速度阈值,则将所述工作波束集包括的波束更新为指定范围的波束,所述指定范围为指定俯仰角范围和指定方位角范围中的工作波束;其中,所述指定俯仰角范围为将所述预设数量的工作波束的俯仰角范围按第一预设步长扩大后得到的俯仰角范围,所述指定方位角范围为将所述预设数量的工作波束的方位角范围按第二预设步长扩大后得到的方位角范围。
- 根据权利要求29所述的装置,其特征在于,所述装置还包括:计算模块,用于基于所述基站的位置、所述终端的位置和所述终端的移动速度,计算所述终端的俯仰角变化速率和方位角变化速率;所述计算模块,还用于计算所述基站的波束在俯仰角维度的第一变化周期和在方位角维度的第二变化周期,所述第一变化周期为所述俯仰角维度的波束宽度与所述俯仰角变化速率的比值,所述第二变化周期为所述方位角维度的波束宽度与所述方位角变化速率的比值;所述更新模块,还用于将所述第一变化周期和所述第二变化周期中较小的一个变化周期与预设系数的比值,作为所述训练波束集和所述工作波束集更新后的测量周期。
- 一种训练数据采集装置,其特征在于,所述装置应用于终端,所述装置包括:测量模块,用于对基站在训练波束集的各波束方向上向所述终端发送的导频信号进行波束测量,对所述基站在工作波束集的各波束方向上向所述终端发送的导频信号进行波束测量;发送模块,用于向所述基站发送对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,所述第一测量结果为训练数据的数据来源,所述第二测量结果为所述训练数据的标签数据的数据来源。
- 根据权利要求36所述的装置,其特征在于,所述发送模块,具体用于:通过物理层信令向所述基站发送第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,通过高层信令向所述基站发送第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
- 根据权利要求37所述的装置,其特征在于,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数目小于所述工作波束集中包括的波束数量。
- 根据权利要求38所述的装置,其特征在于,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
- 根据权利要求36所述的装置,其特征在于,所述装置还包括:接收模块,用于接收所述基站发送的第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;若所述终端支持对所述训练波束集进行波束测量,则所述发送模块向所述基站回复第一确认消息;所述接收模块,还用于接收所述基站发送的送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;若所述终端支持对所述工作波束集进行波束测量,则所述发送模块向所述基站回复第二确认消息。
- 根据权利要求40所述的装置,其特征在于,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
- 根据权利要求36-41任一项所述的装置,其特征在于,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
- 一种基站,其特征在于,包括:处理器;收发器;机器可读存储介质,所述机器可读存储介质存储有能够被所述处理器执行的机器可执行指令;所述机器可执行指令促使所述处理器执行以下步骤:通过所述收发器在训练波束集的各波束方向上向终端发送用于波束测量的导频信号,在工作波束集的各波束方向上向所述终端发送用于波束测量的导频信号;通过所述收发器接收终端发送的对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果;基于所述第一测量结果生成训练数据,基于所述第二测量结果生成所述训练数据对应的标签数据。
- 根据权利要求43所述的基站,其特征在于,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;所述机器可执行指令还促使所述处理器执行以下步骤:若所述第一测量时间戳和所述第二测量时间戳之间的差值绝对值小于预设时间门限,则将所述第一测量结果包括的测量数据作为训练数据,将所述第二测量结果包括的测量数据作为所述训练数据对应的标签数据。
- 根据权利要求43所述的基站,其特征在于,所述第一测量结果中包括第一测量时间戳,所述第二测量结果中包括第二测量时间戳,所述第一测量时间戳为所述终端对所述训练波束集测量的起始时间戳,所述第二测量时间戳为所述终端对所述工作波束集测量的起始时间戳;所述机器可执行指令还促使所述处理器执行以下步骤:针对已接收的多个第一测量结果和多个第二测量结果,将第一测量时间戳与第二测量时间戳之间的差值绝对值小于预设时间门限的一个第一测量结果和一个第二测量结果作为一个测量结果组合;针对每个测量结果组合,将该测量结果组合中的第一测量结果包括的测量数据作为训练数据,将该测量结果组合中的第二测量结果包括的测量数据作为标签数据。
- 根据权利要求43所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:将所述第一测量结果包括的每个波束的逻辑ID转换为物理ID,将所述第二测量结果包括的每个波束的逻辑ID转换为物理ID,所述基站的各波束的物理ID为所述基站基于各自波束的俯仰角和方位角确定出的。
- 根据权利要求43所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:通过所述收发器在所述训练波束集的各波束方向上向所述终端发送用于波束测量的终端专用导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号;或者,通过所述收发器在所述训练波束集的各波束方向上发送用于波束测量的小区公共导频信号,在所述工作波束的各波束方向上向所述终端发送用于波束测量的终端专用导频信号。
- 根据权利要求43或47所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:通过所述收发器接收所述终端通过物理层信令发送的第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,通过所述收发器接收所述终端通过高层信令发送的第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
- 根据权利要求48所述的基站,其特征在于,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数量小于所述工作波束集中包括的波束数量。
- 根据权利要求49所述的基站,其特征在于,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
- 根据权利要求44所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:通过所述收发器接收OAM设备发送的系统间消息,所述系统间消息包括第一字段和第二字段,所述第一字段用于指示启动波束测量,所述第二字段用于承载所述预设时间门限;响应于所述系统间消息,通过所述收发器向所述终端发送第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;通过所述收发器接收所述终端回复的第一确认消息,所述第一确认消息用于表示所述终端支持对所述训练波束集进行波束测量。
- 根据权利要求51所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:响应于所述系统间消息,通过所述收发器向所述终端发送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;通过所述收发器接收所述终端回复的第二确认消息,所述第二确认消息用于表示所述终端支持对所述工作波束集进行波束测量。
- 根据权利要求52所述的基站,其特征在于,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
- 根据权利要求44或45所述的基站,其特征在于,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
- 根据权利要求50所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:若所述终端的移动速度小于等于预设速度阈值,则将所述工作波束集包括的波束更新为指定范围的波束,所述指定范围为指定俯仰角范围和指定方位角范围中的工作波束;其中,所述指定俯仰角范围为将所述预设数量的工作波束的俯仰角范围按第一预设步长扩大后得到的俯仰角范围,所述指定方位角范围为将所述预设数量的工作波束的方位角范围按第二预设步长扩大后得到的方位角范围。
- 根据权利要求50所述的基站,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:基于所述基站的位置、所述终端的位置和所述终端的移动速度,计算所述终端的俯仰角变化速率和方位角变化速率;计算所述基站的波束在俯仰角维度的第一变化周期和在方位角维度的第二变化周期,所述第一变化周期为所述俯仰角维度的波束宽度与所述俯仰角变化速率的比值,所述第二变化周期为所述方位角维度的波束宽度与所述方位角变化速率的比值;将所述第一变化周期和所述第二变化周期中较小的一个变化周期与预设系数的比值,作为所述训练波束集和所述工作波束集更新后的测量周期。
- 一种终端,其特征在于,包括:处理器;收发器;机器可读存储介质,所述机器可读存储介质存储有能够被所述处理器执行的机器可执行指令;所述机器可执行指令促使所述处理器执行以下步骤:对基站在训练波束集的各波束方向上向所述终端发送的导频信号进行波束测量,对所述基站在工作波束集的各波束方向上向所述终端发送的导频信号进行波束测量;通过所述收发器向所述基站发送对所述训练波束集的第一测量结果,以及对所述工作波束集的第二测量结果,所述第一测量结果为训练数据的数据来源,所述第二测量结果为所述训练数据的标签数据的数据来源。
- 根据权利要求57所述的终端,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:通过所述收发器通过物理层信令向所述基站发送第一测量报告消息和第二测量报告消息,所述第一测量报告消息中包括所述第一测量结果,所述第二测量报告消息中包括所述第二测量结果;或者,通过所述收发器通过高层信令向所述基站发送第一RRC消息和第二RRC消息,所述第一RRC消息中包括所述第一测量结果,所述第二RRC消息中包括所述第二测量结果。
- 根据权利要求58所述的终端,其特征在于,所述第一RRC消息和所述第二RRC消息中均包括消息类型字段、测量时间字段、波束数目字段、各波束的逻辑ID和测量数据;所述消息类型字段用于指示训练波束集或工作波束集中的波束,所述测量时间字段用于表示波束测量的起始时间戳;所述第一RRC消息中的波束数目字段的取值为所述训练波束集中包括的波束数量,所述第二RRC消息中的波束数目字段的取值为预设数量,所述预设数目小于所述工作波束集中包括的波束数量。
- 根据权利要求59所述的终端,其特征在于,所述第一RRC消息和所述第二RRC消息中还包括终端位置字段和终端速度字段;所述终端位置字段用于表示所述终端所处的地理位置,所述终端速度字段用于表示所述终端在波束测量过程中的移动速度。
- 根据权利要求57所述的终端,其特征在于,所述机器可执行指令还促使所述处理器执行以下步骤:通过所述收发器接收所述基站发送的第一测量控制消息,所述第一测量控制消息用于指示所述终端在第一指定时频位置对所述训练波束集进行波束测量;若所述终端支持对所述训练波束集进行波束测量,则通过所述收发器向所述基站回复第一确认消息;所述机器可执行指令还促使所述处理器执行以下步骤:通过所述收发器接收所述基站发送的送第二测量控制消息,所述第二测量控制消息用于指示所述终端在第二指定时频位置对所述工作波束集进行波束测量;若所述终端支持对所述工作波束集进行波束测量,则通过所述收发器向所述基站回复第二确认消息。
- 根据权利要求61所述的终端,其特征在于,所述第一测量控制消息和所述第二测量控制消息中均包括预设指示参数,当所述预设指示参数的取值为第一值时,所述预设指示参数用于指示采集波束预测模型的训练数据。
- 根据权利要求57-62任一项所述的终端,其特征在于,所述第一测量结果中包括的测量数据为训练波束集包括的各波束的逻辑ID和信号质量;所述第二测量结果中包括预设数量个工作波束的逻辑ID和信号质量,或者包括所述预设数量个工作波束的逻辑ID,所述预设数量个工作波束为所述终端按照信号质量从高到低的顺序,从所述工作波束集中选择的前预设数量个工作波束。
- 一种机器可读存储介质,其特征在于,存储有机器可执行指令,在被处理器调用和执行时,所述机器可执行指令促使所述处理器:实现权利要求1-14或15-21任一所述的方法步骤。
- 一种计算机程序产品,其特征在于,所述计算机程序产品促使所述处理器:实现权利要求1-14或15-21任一所述的方法步骤。
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/CN2023/074151 WO2024159458A1 (zh) | 2023-02-01 | 2023-02-01 | 训练数据采集方法、装置、基站、终端及存储介质 |
| CN202380008076.6A CN118786696A (zh) | 2023-02-01 | 2023-02-01 | 训练数据采集方法、装置、基站、终端及存储介质 |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/CN2023/074151 WO2024159458A1 (zh) | 2023-02-01 | 2023-02-01 | 训练数据采集方法、装置、基站、终端及存储介质 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2024159458A1 true WO2024159458A1 (zh) | 2024-08-08 |
Family
ID=92145607
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2023/074151 Ceased WO2024159458A1 (zh) | 2023-02-01 | 2023-02-01 | 训练数据采集方法、装置、基站、终端及存储介质 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN118786696A (zh) |
| WO (1) | WO2024159458A1 (zh) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9853702B1 (en) * | 2016-01-13 | 2017-12-26 | Keysight Technologies, Inc. | Methods for channel estimation in OFDMA based hybrid beamforming (HBF) systems |
| CN113810920A (zh) * | 2020-06-11 | 2021-12-17 | 中兴通讯股份有限公司 | 波束训练方法、网络设备、终端、系统和存储介质 |
| CN115622596A (zh) * | 2022-12-12 | 2023-01-17 | 深圳大学 | 一种基于多任务学习的快速波束对齐方法 |
-
2023
- 2023-02-01 CN CN202380008076.6A patent/CN118786696A/zh active Pending
- 2023-02-01 WO PCT/CN2023/074151 patent/WO2024159458A1/zh not_active Ceased
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9853702B1 (en) * | 2016-01-13 | 2017-12-26 | Keysight Technologies, Inc. | Methods for channel estimation in OFDMA based hybrid beamforming (HBF) systems |
| CN113810920A (zh) * | 2020-06-11 | 2021-12-17 | 中兴通讯股份有限公司 | 波束训练方法、网络设备、终端、系统和存储介质 |
| CN115622596A (zh) * | 2022-12-12 | 2023-01-17 | 深圳大学 | 一种基于多任务学习的快速波束对齐方法 |
Non-Patent Citations (1)
| Title |
|---|
| NOKIA, NOKIA SHANGHAI BELL: "Other aspects on ML for beam management", 3GPP TSG RAN WG1 #109, R1-2204574, 29 April 2022 (2022-04-29), XP052153596 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118786696A (zh) | 2024-10-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113259999B (zh) | 一种网络资源配置的方法和装置 | |
| WO2020042081A1 (en) | Method and apparatus for location services | |
| US9426044B2 (en) | Radio access network geographic information system with multiple format | |
| CN103237291B (zh) | 移动终端的整合定位方法和信息服务主动推荐方法 | |
| WO2024012331A1 (zh) | 一种确定人工智能ai模型的方法及装置 | |
| CN112866904A (zh) | 基于波束索引地图的免信道训练大维通信波束对齐方法 | |
| CN115150937B (zh) | 一种通信方法和装置 | |
| WO2025077215A1 (zh) | 通信方法和装置、以及存储介质 | |
| CN113453334B (zh) | 定位方法及定位装置 | |
| WO2024159458A1 (zh) | 训练数据采集方法、装置、基站、终端及存储介质 | |
| WO2018121439A1 (zh) | 一种直射径判断方法及装置 | |
| WO2024199150A1 (zh) | 一种通信方法、通信设备、介质及程序产品 | |
| WO2025097782A1 (zh) | 定位方法、网络设备、终端设备和通信系统 | |
| WO2024207978A1 (zh) | 通信方法和通信装置 | |
| WO2024125510A1 (zh) | 模型优化方法、装置、设备及存储介质 | |
| CN117639974A (zh) | 信息处理方法、装置、节点设备及介质 | |
| CN117528772B (zh) | 终端定位方法、装置、设备及存储介质 | |
| Zhang et al. | Demo Abstract: PlanIt and DQ-N for low-power wide-area networks | |
| CN120475508B (zh) | 通信方法、装置、通信芯片、存储介质及计算机程序产品 | |
| US20260012825A1 (en) | Channel information determination method and apparatus, and medium | |
| CN119697762A (zh) | 定位方法、装置、通信设备、可读存储介质和程序产品 | |
| WO2024229681A1 (zh) | 波束预测模型训练方法、波束预测方法及装置 | |
| WO2025167215A1 (zh) | 一种通信方法及装置 | |
| CN121604000A (zh) | 一种训练感知模型的方法、用户设备、基站及服务器 | |
| WO2025016043A1 (zh) | 模型获取方法、装置及系统 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| WWE | Wipo information: entry into national phase |
Ref document number: 202380008076.6 Country of ref document: CN |
|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 23919036 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 23919036 Country of ref document: EP Kind code of ref document: A1 |