Data processing method, computer readable storage medium and related device

Abstract

The invention provides a data processing method, a computer readable storage medium and a related device, and the method comprises the steps that first equipment obtains first data information, the first data information comprises N pieces of first data, the N pieces of first data correspond to N positions, and N is an integer greater than 1; selecting M target positions from the N positions based on a first position selection mode, wherein M is an integer less than N and greater than 0; obtaining K pieces of second data based on a first data extraction mode and the M target positions, wherein K is an integer smaller than N and larger than 0; and sending the K pieces of second data and the position information of the M target positions to second equipment. According to the embodiment of the invention, the compression benefit of the data information (such as a radio frequency map) can be improved, and the transmission overhead of the data information is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication, and particularly relates to a data processing method, a computer readable storage medium and related devices. BACKGROUND

[0002] In a 6G communication and sensing integrated scene, a network device (such as a base station), a terminal device, and the like can obtain a radio frequency map (RF map) of a signal propagation environment, and then can assist a communication task based on the radio frequency map, or perform resource management and resource allocation based on the radio frequency map, for example, can perform waveform management, MIMO (multiple in multiple out) management, and the like based on the radio frequency map.

[0003] Generally, radio frequency map transmission usually needs to occupy more transmission resources, and therefore, how to reduce the transmission overhead of the radio frequency map is a problem that a person skilled in the art is concerned about. SUMMARY

[0004] Embodiments of the present application disclose a data processing method, a computer readable storage medium and related devices, which can improve the compression efficiency of the radio frequency map and reduce the transmission overhead of the radio frequency map.

[0005] A first aspect discloses a data processing method, which can be applied to a first device, can be applied to a module (for example, a processor) in the first device, can be applied to a logic module or software capable of realizing all or part of the function of the first device. The following takes the application to the first device as an example for description, and the data processing method can comprise: obtaining first data information, the first data information comprising N first data, the N first data corresponding to N positions, N being an integer greater than 1; selecting M target positions from the N positions, M being an integer less than N and greater than 0; obtaining K second data based on the M target positions, K being an integer less than N and greater than 0; and sending the K second data and position information of the M target positions to a second device.

[0006] In the embodiments of the present application, the first device can select part target positions (M target positions) from all positions of the data information to be transmitted (first data information) based on a first position selection manner, then perform data extraction based on the first data extraction manner and the selected part target positions, and then the first device can send position information of the selected part target positions and the extracted data (K second data) to the second device. After the second device receives the K second data and the position information of the M target positions, the corresponding data information can be reconstructed based on the reconstruction model. In this way, the data sent by the first device can not include the information of the complete data information to be transmitted, but can only include the position information of the selected part positions and a small amount of data, so that the compression efficiency of the data information (such as the radio map) can be improved, the amount of data to be transmitted can be reduced, and the transmission resources can be saved.

[0007] For example, the data information can be a radio map.

[0008] In combination with the first aspect, in a possible implementation, the first data information is two-dimensional data information or three-dimensional data information.

[0009] In combination with the first aspect, in a possible implementation, the first data information is a radio map.

[0010] In combination with the first aspect, in a possible implementation, the selecting M target positions from the N positions includes: selecting M target positions from the N positions based on a first position selection manner, and the first position selection manner is one of random selection, topology selection, grouping selection, and iteration selection.

[0011] In the embodiments of the present application, various position selection manners such as random selection, topology selection, grouping selection, and iteration selection are included, and the first device can select to use according to actual conditions, and can be flexibly adapted to different scenes.

[0012] In combination with the first aspect, in a possible implementation, in the case where the first position selection manner is grouping selection, the selecting M target positions from the N positions based on the first position selection manner includes: dividing the N positions into L groups of positions based on the N first data, L is an integer greater than 1; and selecting at least one position from each group of positions in the L groups of positions to obtain M target positions.

[0013] In the embodiments of the present application, grouping selection can enable the first device to select certain positions from different groups, which can ensure that the data extracted later can be more uniform, so that the reconstruction performance can be more stable, and the accuracy can be higher.

[0014] With reference to the first aspect, in a possible implementation of the first aspect, in a case where the first position selection manner is topology selection, the selecting, by the first device, the M target positions from the N positions based on the first position selection manner comprises: selecting, by the first device, the M target positions from the N positions based on a topology of the N first data.

[0015] In the embodiments of the present application, the topology selection manner retains the correlation features between positions, so that the subsequently extracted data also contains the correlation features between positions, thereby enabling the reconstruction accuracy to be higher.

[0016] With reference to the first aspect, in a possible implementation of the first aspect, the obtaining, by the first device, the K second data based on the M target positions comprises: obtaining, by the first device, the K second data based on a first data extraction manner and the M target positions, the first data extraction manner being original data extraction or weighted data extraction.

[0017] With reference to the first aspect, in a possible implementation of the first aspect, in a case where the first data extraction manner is weighted data extraction, the obtaining, by the first device, the K second data based on the first data extraction manner and the M target positions comprises: performing weighted calculation on the first data corresponding to the M target positions and all or part of positions that are within a first threshold distance from the M target positions respectively, to obtain the K second data.

[0018] In the embodiments of the present application, the weighted data extraction manner can enable the obtained second data to contain data information of the target positions and data information of the positions in the vicinity of the target positions, and can enable the generalization to be better and the reconstruction effect to be better.

[0019] With reference to the first aspect, in a possible implementation of the first aspect, the method further comprises: sending, by the first device, first information to the second device, the first information comprising identification information of the first position selection manner and / or identification information of the first data extraction manner.

[0020] In the embodiments of the present application, different position selection manners and / or data extraction manners can correspond to different reconstruction models, and therefore the first device can send the first information to the second device, so that the second device can adopt the corresponding reconstruction model to perform data information reconstruction, which can improve the reconstruction accuracy and reduce the reconstruction error.

[0021] With reference to the first aspect, in a possible implementation of the first aspect, the method further comprises: selecting, by the first device, J target positions from the N positions, the J target positions being different from the target positions that have been sent, J being an integer less than N and greater than 0; obtaining, by the first device, H second data based on the J target positions, H being an integer less than N and greater than 0; and sending, by the first device, the H second data and position information of the J target positions to the second device.

[0022] In the embodiments of the present application, the first device can send multiple groups of second data and the position information of the corresponding target positions in multiple times, so as to ensure the reconstruction effect in the scenario where the transmission resource is limited, so as to meet the error requirement (such as the reconstruction error is less than a second threshold) of the second device.

[0023] In combination with the first aspect, in a possible implementation, the method further includes: obtaining multiple data information; processing the multiple data information based on the first position selection manner and the first data extraction manner respectively to obtain multiple groups of model input data; training an initial first reconstruction model based on the multiple groups of model input data and the multiple data information to obtain a trained first reconstruction model; and sending the trained first reconstruction model to the second device.

[0024] In the embodiments of the present application, the first device can first obtain multiple data information, and then can process the multiple data information based on a first position selection manner and a first data extraction manner to obtain multiple groups of model input data. Then, the first device can input the multiple groups of model data into a first reconstruction model to obtain corresponding reconstructed data information. Then, the first device can optimize the parameters of the first reconstruction model based on the error between the reconstructed data information and the original data information to obtain a trained first reconstruction model. Then, the first device can send the trained first reconstruction model to the second device, so that the second device can perform data information reconstruction based on the trained first reconstruction model.

[0025] In combination with the first aspect, in a possible implementation, the multiple groups of model input data are obtained based on different position selection rates, and the position selection rate is used to indicate the proportion of the number of selected positions to the total number of positions in the data information.

[0026] In the embodiments of the present application, the multiple groups of model input data used for training the first reconstruction model can be obtained based on different position selection rates, so that the trained first reconstruction model can be applicable to different position selection rates, and the flexibility of the first reconstruction model can be improved.

[0027] The second aspect discloses a data processing method, which can be applied to a second device, a module (for example, a processor) in the second device, or a logic module or software capable of realizing all or part of the functions of the second device. The data processing method is described below by taking the application to the second device as an example. The data processing method can include: receiving K second data and position information of M target positions from a first device; reconstructing data information corresponding to the K second data and the position information of the M target positions based on a first reconstruction model to obtain second data information, the second data information including N first data; N is an integer greater than 1, and M and K are integers less than N and greater than 0.

[0028] With reference to the second aspect, in a possible implementation of the second aspect, the first data information is two-dimensional data information or three-dimensional data information.

[0029] With reference to the second aspect, in a possible implementation of the second aspect, the first data information is a radio map.

[0030] With reference to the second aspect, in a possible implementation of the second aspect, before the reconstructing, by using the first reconstruction model, the data information corresponding to the K second data and the position information of the M target positions to obtain the second data information, the method further includes: receiving first information from the first device, the first information including identification information of a first position selection manner and / or identification information of a first data extraction manner; determining the first reconstruction model corresponding to the first position selection manner and / or the first data extraction manner; and wherein different position selection manners and / or data extraction manners correspond to different reconstruction models.

[0031] With reference to the second aspect, in a possible implementation of the second aspect, the first position selection manner is one of random selection, topological selection, grouping selection, and iterative selection, and the first data extraction manner is one of raw data extraction and weighted data extraction.

[0032] With reference to the second aspect, in a possible implementation of the second aspect, the method further includes: receiving H second data and position information of J target positions from the first device, the J target positions being different from the target position that has been received, and H and J being integers less than N and greater than 0; and reconstructing third data information based on the second data and the position information of the target position that have been received and the H second data and the position information of the J target positions, the third data information including N first data.

[0033] With reference to the second aspect, in a possible implementation of the second aspect, before the reconstructing, by using the first reconstruction model, the data information corresponding to the K second data and the position information of the M target positions to obtain the second data information, the method further includes: receiving the trained first reconstruction model from the first device.

[0034] With reference to the second aspect, in a possible implementation of the second aspect, before the reconstructing, by using the first reconstruction model, the data information corresponding to the K second data and the position information of the M target positions to obtain the second data information, the method further includes: obtaining a plurality of data information; processing the plurality of data information based on the first position selection manner and the first data extraction manner to obtain a plurality of groups of model input data; and training an initial first reconstruction model based on the plurality of groups of model input data and the plurality of data information to obtain the trained first reconstruction model.

[0035] With reference to the second aspect, in a possible implementation, the multiple sets of model input data are obtained based on different position selection rates, where the position selection rate is used to indicate a ratio of a number of selected positions to a total number of positions in the data information.

[0036] It should be noted that the technical solutions of the second aspect of the present application can correspond to the solutions of the first aspect, and the related beneficial effects can refer to the beneficial effects of the first aspect.

[0037] The third aspect discloses a data processing apparatus, which can be the first device or a module (for example, a processor, a communication module) in the first device. The data processing apparatus comprises:

[0038] a processing unit, configured to obtain first data information, the first data information comprising N first data, the N first data corresponding to N positions, N being an integer greater than 1;

[0039] The processing unit is further configured to select M target positions from the N positions, M being an integer less than N and greater than 0;

[0040] The processing unit is further configured to obtain K second data based on the M target positions, K being an integer less than N and greater than 0;

[0041] a sending unit, configured to send the K second data and position information of the M target positions to a second device.

[0042] With reference to the third aspect, in a possible implementation, the first data information is two-dimensional data information or three-dimensional data information.

[0043] With reference to the third aspect, in a possible implementation, the first data information is a radio frequency map.

[0044] With reference to the third aspect, in a possible implementation, the processing unit selecting the M target positions from the N positions comprises: selecting the M target positions from the N positions based on a first position selection manner, the first position selection manner being one of random selection, topology selection, grouping selection, and iteration selection.

[0045] With reference to the third aspect, in a possible implementation, when the first position selection manner is grouping selection, the processing unit selecting the M target positions from the N positions based on the first position selection manner comprises: dividing the N positions into L groups of positions based on the N first data, L being an integer greater than 1; and selecting at least one position from each group of positions in the L groups of positions to obtain the M target positions.

[0046] In a possible implementation of the third aspect, in the case that the first position selection manner is topology selection, the processing unit selects M target positions from the N positions based on the first position selection manner, including: selecting M target positions from the N positions based on the topology of the N first data.

[0047] In a possible implementation of the third aspect, the processing unit obtains K second data based on the M target positions, including: obtaining K second data based on the first data extraction manner and the M target positions, the first data extraction manner being original data extraction or weighted data extraction.

[0048] In a possible implementation of the third aspect, in the case that the first data extraction manner is weighted data extraction, the processing unit obtains K second data based on the first data extraction manner and the M target positions, including: performing weighted calculation on the first data corresponding to the M target positions and positions with a distance less than a first threshold from the M target positions respectively, to obtain K second data.

[0049] In a possible implementation of the third aspect, the sending unit is further configured to send, to the second device, first information, the first information including identification information of the first position selection manner and / or identification information of the first data extraction manner.

[0050] In a possible implementation of the third aspect, the processing unit is further configured to select J target positions from the N positions, the J target positions being different from the target positions that have been sent, J being an integer less than N and greater than 0; the processing unit is further configured to obtain H second data based on the J target positions, H being an integer less than N and greater than 0; and the sending unit is further configured to send, to the second device, the H second data and position information of the J target positions.

[0051] In a possible implementation of the third aspect, the processing unit is further configured to obtain a plurality of data information; the processing unit is further configured to process the plurality of data information based on the first position selection manner and the first data extraction manner, to obtain a plurality of groups of model input data; the processing unit is further configured to train an initial first reconstruction model based on the plurality of groups of model input data and the plurality of data information, to obtain a trained first reconstruction model; and the sending unit is further configured to send, to the second device, the trained first reconstruction model.

[0052] In a possible implementation of the third aspect, the plurality of groups of model input data are obtained based on different position selection rates, the position selection rate being used to indicate a proportion of a number of selected positions to a total number of positions in data information.

[0053] A fourth aspect discloses a data processing apparatus, which can be the second device or a module (e.g., a processor, a communication module) in the second device. The data processing apparatus comprises:

[0054] a receiving unit configured to receive K second data and position information of M target positions from the first device;

[0055] a processing unit configured to reconstruct data information corresponding to the K second data and the position information of the M target positions based on a first reconstruction model to obtain second data information, the second data information comprising N first data; N is an integer greater than 1, and M and K are integers less than N and greater than 0.

[0056] With reference to the fourth aspect, in a possible implementation, the first data information is two-dimensional data information or three-dimensional data information.

[0057] With reference to the fourth aspect, in a possible implementation, the first data information is a radio frequency map.

[0058] With reference to the fourth aspect, in a possible implementation, before the processing unit reconstructs the data information corresponding to the K second data and the position information of the M target positions based on the first reconstruction model to obtain the second data information, the receiving unit is further configured to receive first information from the first device, the first information comprising identification information of a first position selection manner and / or identification information of a first data extraction manner; and the processing unit is further configured to determine the first reconstruction model corresponding to the first position selection manner and / or the first data extraction manner; wherein different position selection manners and / or data extraction manners correspond to different reconstruction models.

[0059] With reference to the fourth aspect, in a possible implementation, the first position selection manner is one of random selection, topology selection, grouping selection, and iteration selection, and the first data extraction manner is original data extraction or weighted data extraction.

[0060] With reference to the fourth aspect, in a possible implementation, the receiving unit is further configured to receive H second data and position information of J target positions from the first device, the J target positions being different from the target positions that have been received, and H and J are integers less than N and greater than 0; and the processing unit is further configured to reconstruct third data information based on the second data and the position information of the target positions that have been received, and the H second data and the position information of the J target positions, the third data information comprising N first data.

[0061] In combination with the fourth aspect, in a possible implementation, before the processing unit reconstructs the second data information from the K second data and the data information corresponding to the position information of the M target positions based on the first reconstruction model, the receiving unit is further configured to receive the trained first reconstruction model from the first device.

[0062] In combination with the fourth aspect, in a possible implementation, before the processing unit reconstructs the second data information from the K second data and the data information corresponding to the position information of the M target positions based on the first reconstruction model, the processing unit is further configured to obtain a plurality of data information; the processing unit is further configured to process the plurality of data information based on the first position selection manner and the first data extraction manner respectively to obtain a plurality of groups of model input data; and the processing unit is further configured to train the initial first reconstruction model based on the plurality of groups of model input data and the plurality of data information to obtain the trained first reconstruction model.

[0063] In combination with the fourth aspect, in a possible implementation, the plurality of groups of model input data are obtained based on different position selection rates, and the position selection rate is used to indicate a proportion of a number of selected positions to a total number of positions in data information.

[0064] The fifth aspect discloses a data processing apparatus, including a processor and a communication interface; the communication interface is configured to receive and / or send data; and the processor is configured to invoke a computer program or computer instructions stored in a memory to implement the method provided in the first aspect and any possible implementation of the first aspect.

[0065] The sixth aspect discloses a data processing apparatus, including a processor and a communication interface; the communication interface is configured to receive and / or send data; and the processor is configured to invoke a computer program or computer instructions stored in a memory to implement the method provided in the second aspect and any possible implementation of the second aspect.

[0066] The seventh aspect discloses a communication system, including a first device and a second device; the first device is configured to implement the method provided in the first aspect and any possible implementation of the first aspect; and the second device is configured to implement the method provided in the second aspect and any possible implementation of the second aspect.

[0067] The eighth aspect discloses a computer readable storage medium, and the computer readable storage medium stores a computer program or computer instructions, and the computer program or computer instructions is configured to implement the method provided in the first aspect and any possible implementation of the first aspect when the computer program or computer instructions is executed.

[0068] The ninth aspect discloses a computer readable storage medium, which stores a computer program or computer instructions, and when the computer program or computer instructions are executed, the method provided in the above second aspect and any possible implementation manner of the second aspect is implemented.

[0069] The tenth aspect discloses a chip, which comprises a processor for executing a program stored in a memory, and when the program is executed, the chip executes the method provided in the above first aspect and any possible implementation manner of the first aspect.

[0070] The eleventh aspect discloses a chip, which comprises a processor for executing a program stored in a memory, and when the program is executed, the chip executes the method provided in the above second aspect and any possible implementation manner of the second aspect.

[0071] As a possible implementation manner, the memory is located outside the chip.

[0072] The twelfth aspect discloses a computer program product, which comprises computer program codes, and when the computer program codes are executed, the method provided in the above first aspect and any possible implementation manner of the first aspect is executed, or the method provided in the above second aspect and any possible implementation manner of the second aspect is executed. It should be understood that the implementation and beneficial effects of the above aspects or any possible implementation manner of the present application can be mutually referred. BRIEF DESCRIPTION OF DRAWINGS

[0073] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0074] Figure 1A is a schematic diagram of a 2D radio frequency map disclosed by the embodiments of the present application;

[0075] Figure 1B is a schematic diagram of a 3D radio frequency map disclosed by the embodiments of the present application;

[0076] Figure 2 is a schematic diagram of a network architecture disclosed by the embodiments of the present application;

[0077] Figure 3 is a schematic diagram of a radio frequency map compression and reconstruction processing flow disclosed by the embodiments of the present application;

[0078] Figure 4is a flowchart of a data processing method disclosed by an embodiment of the present application;

[0079] Figure 5 is a schematic diagram of a packet disclosed by an embodiment of the present application;

[0080] Figure 6 is a schematic diagram of a weighting calculation disclosed by an embodiment of the present application;

[0081] Figure 7 is a schematic diagram of a 2D position mapping disclosed by an embodiment of the present application;

[0082] Figure 8 is a schematic diagram of generating reconstruction model input data disclosed by an embodiment of the present application;

[0083] Figure 9 is a flowchart of a reconstruction model training method disclosed by an embodiment of the present application;

[0084] Figure 10 is a schematic diagram of model training disclosed by an embodiment of the present application;

[0085] Figure 11 is a flowchart of another reconstruction model training method disclosed by an embodiment of the present application;

[0086] Figure 12 is a flowchart of another data processing method disclosed by an embodiment of the present application;

[0087] Figure 13 is a flowchart of layer increment transmission disclosed by an embodiment of the present application;

[0088] Figure 14 is a schematic diagram of reconstruction error under different conditions disclosed by an embodiment of the present application;

[0089] Figure 15 is a comparative schematic diagram of radio frequency map reconstruction disclosed by an embodiment of the present application;

[0090] Figure 16 is another comparative schematic diagram of radio frequency map reconstruction disclosed by an embodiment of the present application;

[0091] Figure 17 is a structural schematic diagram of a data processing apparatus disclosed by an embodiment of the present application;

[0092] Figure 18 is a structural schematic diagram of another data processing apparatus disclosed by an embodiment of the present application;

[0093] Figure 19 is a hardware structural schematic diagram of a data processing apparatus disclosed by an embodiment of the present application. DETAILED DESCRIPTION

[0094] The embodiment of the present application discloses a data processing method, a computer readable storage medium and related devices, which can improve the compression efficiency of a radio frequency map and reduce the transmission cost of the radio frequency map. The technical solutions in the embodiment of the present application will be clearly and completely described below with reference to the drawings in the embodiment of the present application.

[0095] A radio frequency map (RF map) can include electromagnetic information (such as signal strength, power spectral density, received signal strength indication, other electromagnetic information related to a propagation environment, etc.) in a geographic area (such as A city B district C street D area), which can be used to understand the signal propagation environment of the geographic area. Generally, a radio frequency map can be divided into two types, one is a 2-dimensional (dimension, D) radio frequency map, and the other is a 3D radio frequency map. The difference between the 2D radio frequency map and the 3D radio frequency map is that the 3D radio frequency map includes multiple frequency dimensions compared with the 2D radio frequency map.

[0096] For example, please refer to Figure 1A and Figure 1B , Figure 1A is a schematic diagram of a 2D radio frequency map disclosed by the embodiment of the present application, Figure 1B is a schematic diagram of a 3D radio frequency map disclosed by the embodiment of the present application, as shown in Figure 1A , a 2D radio frequency map can be represented as a matrix of a*b( Figure 1A for example, 6*5), and the matrix stores electromagnetic elements such as signal strength, power spectral density (PSD), received signal strength indication (RSSI), etc. Wherein, the matrix can be associated with a geographic area, and each element in the matrix can be associated with a sub-area in the geographic area. For example, as shown in Figure 1A , the 6*5 matrix is associated with area a, and area a can include multiple sub-areas, wherein S1 can be associated with sub-area 1 in area a, S2 can be associated with sub-area 2 in area a, S3 can be associated with sub-area 3 in area a, S4 can be associated with sub-area 4 in area a, S5 can be associated with sub-area 5 in area a, etc. As shown in Figure 1B , a 3D radio frequency map can be represented as a matrix of a*b*c( Figure 1BThe matrix stores electromagnetic elements such as signal strength, power spectral density, received signal strength indication, etc. The 3D radio map can be associated with a geographical area, and can include 2D radio maps corresponding to multiple frequencies in the geographical area, for example, can include 2D radio maps corresponding to frequencies f1, f2, f3, f4, f5 and f6. That is, the 3D radio map can present 2D radio maps corresponding to different frequencies respectively. It should be understood that for each 2D radio map corresponding to a frequency, each element in the 2D radio map can be associated with a sub-region in the corresponding geographical area. For example, assuming that a certain 2D radio map and a certain 3D radio map are associated with the same geographical area, for the 3D radio map, for each sub-region in the geographical area, electromagnetic data under multiple frequencies corresponding to the sub-region can be included. For the 2D radio map, for each sub-region in the geographical area, one electromagnetic data corresponding to the sub-region can be included, which can be an average of electromagnetic data under multiple frequencies corresponding to the sub-region, or can be other electromagnetic data, which is not limited here.

[0097] In actual scenarios, in order to reduce communication overhead when transmitting a radio map, it is usually necessary to compress the radio map before transmission. A common compression method is to quantize and compress the data of each sub-region (position) in the radio map, that is, to compress the data information of all sub-regions in the radio map.

[0098] In the embodiments of the present application, in order to improve the compression efficiency of the radio map and further reduce the amount of data to be transmitted, a compression and reconstruction scheme of the radio map is proposed. Specifically, in the present scheme, when performing data compression, the first device can select part of the sub-regions from the radio map, and then extract corresponding data (such as original electromagnetic data or feature data) based on the selected part of the sub-regions. Then the first device can send the extracted data and the position information of the selected part of the sub-regions to the second device. After receiving the data and the position information of the selected part of the sub-regions sent by the first device, the second device can input the received data and the position information of the selected part of the sub-regions into a reconstruction model, and obtain a radio map reconstructed by the reconstruction model. In the above manner, for the radio map to be transmitted, the first device can only transmit the data and the position information corresponding to the selected part of the sub-regions, which can save transmission resources.

[0099] In order to better understand the embodiments of the present application, the network architecture of the embodiments of the present application will be described first.

[0100] Please refer to Figure 2 , Figure 2 is a network architecture diagram disclosed by the embodiments of the present application. As Figure 2As shown, the network architecture can include a first device and a second device. The first device and the second device can communicate with each other. For example, the first device can be configured with one or more antennas for transmitting and receiving data / information, and the second device can be configured with multiple antennas for transmitting and receiving data / information. It should be understood that the first device and the second device can also include various components (e.g., a processor, a modulator, a multiplexer, a demodulator, a demultiplexer, etc.) related to the transmission and reception of data / information.

[0101] In the embodiments of the present application, the first device and the second device are not specifically limited. For example, the first device can be a user equipment, a network device (e.g., an access network device, a core network device, etc.), or other devices. The second device can be a user equipment, a network device (e.g., an access network device, a core network device, etc.), or other devices.

[0102] The user equipment (UE), also known as terminal device, terminal, mobile station (MS), mobile terminal (MT), customer premise equipment (CPE), etc., is a device with wireless communication function, which can provide voice and / or data connectivity services for users. The terminal device can be a handheld terminal, a notebook computer, an RSU (road side unit), a subscriber unit, a cellular phone, a smart phone, a wireless data card, a personal digital assistant (PDA) computer, a tablet computer, a tag, a wireless modem, other processing devices connected to the wireless modem, a handheld device, a laptop computer, a cordless phone, or a wireless local loop (WLL) station, a machine type communication (MTC) terminal, a wearable device (such as a smart watch, a smart bracelet, a pedometer, etc.), a vehicle-mounted device (such as a car, a bicycle, an electric vehicle, an airplane, a ship, a train, a high-speed rail, etc.), a virtual reality (VR) device, an augmented reality (AR) device, a wireless terminal in industrial control, a smart home device (such as a refrigerator, a television, an air conditioner, an electric meter, etc.), a smart robot, a workshop device, a wireless terminal in self driving, a wireless terminal in remote medical surgery, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, or a wireless terminal in smart home, a flight device (such as a smart robot, a hot air balloon, a drone, an airplane, etc.) or other devices that can access a network. The terminal device can be fixed or mobile, and can be deployed on land, including indoors or outdoors, handheld, wearable or vehicle-mounted; can also be deployed on the water surface (such as a ship, etc.); and can also be deployed in the air (such as an airplane, a balloon and a satellite, etc.).

[0103] The access network device can be a device providing access for a terminal device, and can include a radio access network (RAN) device and an access node (AN) device. The RAN device can include various forms of base stations, such as macro base stations, micro base stations (also referred to as small stations), relay stations, access points, balloon stations, and the like. In systems employing different radio access technologies, the names of the radio access network devices can be different. For example, a base transceiver station (BTS) in a global system for mobile communication (GSM) or code division multiple access (CDMA) network, a NB (NodeB) in wideband code division multiple access (WCDMA), an evolved NodeB (eNB or eNodeB) in long term evolution (LTE), a next generation NodeB (gNB) in a 5th generation (5G) mobile communication system, an ng-eNB (4G base station accessing a 5G core network). The radio access network device can also be a radio controller in a cloud radio access network (CRAN) scenario, a base station device in a future network (such as 6G, 7G, etc.), a radio access network device in a future evolved public land mobile network (PLMN) network, a wearable device, a vehicle-mounted device, a transmission and reception point (TRP), a radio network controller (RNC), a home base station (for example, a home evolved NodeB, or a home NodeB, HNB), a baseband unit (BBU), an access point (AP) in a wireless fidelity (WiFi) system, and the like.

[0104] In some deployments, an access network device (e.g., gNB) can include a central / centralized unit (CU) and a distributed unit (DU), etc. The access network device can also include a radio unit (RU), which can be specifically referred to as Figure 2 . As shown in Figure 2 , the access network device can communicate with a core network (CN) through a backhaul and communicate with a terminal device (UE) through an air interface (e.g., Uu interface). Specifically, a baseband unit (BBU) in the access network device can communicate with the core network through the backhaul, and a radio unit in the access network device can communicate with the terminal device through the air interface. Moreover, the BBU can communicate with the RU through a front-haul, and the BBU and the RU can be co-located or not. The BBU can include at least one central unit (CU) and at least one distributed unit (DU), and the CU and the DU can communicate through a mid-haul. The CU can implement part of the functions of the access network device, the DU can implement part of the functions of the access network device, and the CU can be used to control the operation of one or more DUs. For example, the CU can implement the functions of the radio resource control (RRC) and the packet data convergence protocol (PDCP) layer, and can also implement the functions of the service data adaptation protocol (SDAP) layer, the DU implements the functions of the radio link control (RLC) and the media access control (MAC) layer, and can also implement part of the functions of the physical (PHY) layer or all the functions of the physical layer. For specific descriptions of the above protocol layers, please refer to the relevant technical specifications of the 3rd generation partnership project (3GPP). It can be understood that, in some possible implementation manners, the access network device can be a CU node or a DU node or a device including a CU node and a DU node.

[0105] It can be understood that, when one of the first device and the second device is a user equipment and the other is an access network device, the first device and the second device can communicate through an air interface (e.g., Uu interface).

[0106] It should be noted that, Figure 2The illustrated architecture is only exemplary and does not limit the architecture. For example, in some possible implementations, Figure 2 More or fewer devices can be included in the illustrated architecture.

[0107] It should be understood that the technical solutions provided by the embodiments of the present application can be applied to communication systems of various radio access technologies (RATs), such as a fifth generation (5G) system, a network of multiple system fusion, a wireless local area network (WLAN), and the like; of course, it can also be a future communication system, such as a sixth generation (6G) system, a seventh generation (7G) system, an eighth generation (8G) system, and the like.

[0108] It should be noted that the system architecture, network architecture, and business scenarios (or application scenarios) described in the embodiments of the present application are to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the evolution of communication network architecture and the appearance of new business scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0109] In order to better understand the embodiments of the present application, the contents of the embodiments of the present application will be introduced as a whole.

[0110] In the embodiments of the present application, the first device mainly selects the position for compression when compressing the data information, so that a small amount of data corresponding to the selected position (such as the original data of the selected position or the weighted data corresponding to the selected position) can be obtained, thereby saving transmission resources and reducing storage overhead. When the second device reconstructs (restores) the data information, it can be reconstructed based on an artificial intelligence (AI) model trained in advance. The data information can be one-dimensional data information, two-dimensional data information, three-dimensional data information, or higher-dimensional data information (such as four-dimensional data information, five-dimensional data information, etc.).

[0111] In the embodiments of the present application, the data information can be a radio frequency map or other data information, which is not limited. In order to facilitate the understanding of the embodiments of the present application, the following mainly takes the data information as a radio frequency map for illustration. For example, two-dimensional data information can be a two-dimensional radio frequency map, and three-dimensional data information can be a three-dimensional radio frequency map.

[0112] Exemplarily, when the first device performs the radio map compression, the compression is mainly based on the position (sub-region) selection. In this way, a small amount of data corresponding to the selected position (such as the original data of the selected position or the weighted data corresponding to the selected position) can be obtained, so that the transmission resource can be saved and the storage overhead can be reduced. When the second device performs the radio map reconstruction, the reconstruction can be performed based on the pre-trained artificial intelligence model. In the embodiments of the present application, the sub-region can also be referred to as the position, that is, the position in the radio map.

[0113] Specifically, please refer to Figure 3 , Figure 3 is a processing flow diagram of radio map compression and reconstruction disclosed in the embodiments of the present application. As shown in Figure 3 , after the first device obtains the radio map to be transmitted, the position selection can be performed, and part of the positions in the radio map to be transmitted are selected, and then the data extraction can be performed based on the selected part of the positions. After the first device performs the data extraction, the data (such as the original data, the weighted data, etc.) corresponding to the selected positions can be obtained, and then the first device can perform the data compression on the selected position information and the data corresponding to the selected positions, and then the compressed data can be sent to the second device. It can be understood that in some embodiments, the first device can not perform the data compression, and can directly send the selected position information and the data corresponding to the selected positions to the second device. In the above-mentioned manner, by sending the selected position information and the data corresponding to the selected positions, the amount of data to be sent can be reduced, so as to realize the data compression.

[0114] After the second device receives the compressed data sent by the first device, the data decompression can be performed to obtain the selected position information and the data corresponding to the selected positions. Then, the second device can perform the position mapping, that is, the received data is mapped to the positions corresponding to the selected position information, or the received data is associated with the corresponding positions. Then, the second device can perform the radio map reconstruction through the reconstruction model to obtain the reconstructed radio map. In some embodiments, the second device can also directly input the selected position information and the data corresponding to the selected positions into the reconstruction model to perform the radio map reconstruction. In this case, the position mapping can be completed by the reconstruction model itself.

[0115] It can be understood that in order to make the reconstructed radio map obtained by the reconstruction to be within a certain range of error (such as the error being less than a second threshold value) from the original radio map, the reconstruction model can be trained based on a large number of radio maps in advance. In this way, the accuracy of the radio map reconstructed by the reconstruction model can be high, and the error from the original radio map can be small.

[0116] It should be noted that the embodiments of the present application can be applied to all RF map compression scenarios. The compression scheme provided by the embodiments of the present application can improve the compression efficiency of the radio frequency map, thereby saving transmission resources and reducing storage overhead.

[0117] The technical solutions provided by the embodiments of the present application will be described in detail below.

[0118] Based on the above network architecture, please refer to Figure 4 , Figure 4 is a flowchart of a data processing method disclosed by the embodiments of the present application. As shown in Figure 4 , the data processing method can include but is not limited to the following steps:

[0119] 401. The first device acquires a first radio frequency map, the first radio frequency map including N first data, the N first data corresponding to N positions, N being an integer greater than 1.

[0120] Wherein, the first radio frequency map is associated with the first geographic area. In the embodiments of the present application, the way in which the first device acquires the first radio frequency map is not limited, which can be obtained by the first device itself sensing the first geographic area, can be obtained by the first device from other devices, or can be obtained by other means.

[0121] The first radio frequency map is associated with the first geographic area, which can be understood as the first radio frequency map being the radio frequency map corresponding to the first geographic area. Moreover, the first geographic area can include a plurality of sub-areas, and correspondingly, the first radio frequency map can include electromagnetic data corresponding to the plurality of sub-areas.

[0122] The first radio frequency map can be a 2D radio frequency map or a 3D radio frequency map. In the case of the first radio frequency map being a 2D radio frequency map, the N positions included in the first radio frequency map can correspond to N sub-regions. Taking a 2D matrix of m*n (m*n=N) as an example of the first radio frequency map, the N positions are N matrix positions in the 2D matrix, and each matrix position can correspond to first data. The first data can be a single electromagnetic data, that is, each of the N sub-regions can correspond to an electromagnetic data such as signal strength, RSSI or PSD of the sub-region. In the case of the first radio frequency map being a 3D radio frequency map, in addition to the sub-regions, a frequency dimension can also be included, that is, for a sub-region, first data corresponding to multiple frequencies can be included. Taking a 3D matrix of m*n*k (m*n*k=N) as an example of the first radio frequency map, k can represent k frequencies. In this case, the N positions are N matrix positions in the 2D matrix, and each matrix position can correspond to first data. The number of sub-regions can be m*n, and each sub-region can correspond to k first data. The first data can be a single electromagnetic data, that is, each of the N positions can correspond to an electromagnetic data such as signal strength, RSSI or PSD corresponding to the position, that is, signal strength, RSSI or PSD corresponding to the sub-region associated with the position at a certain specific frequency.

[0123] It should be noted that the embodiments of the present application do not make specific limitations on the first radio frequency map and the first data corresponding to each position in the first radio frequency map. For example, in some embodiments, the first data corresponding to a position can include multiple different types of electromagnetic data such as signal strength, RSSI and PSD, and in this case, the first data corresponding to a position can be an electromagnetic data vector.

[0124] 402. The first device selects M target positions from the N positions, M being an integer less than N and greater than 0.

[0125] In the embodiments of the present application, in order to save transmission resources, for a radio frequency map to be transmitted (such as the first radio frequency map), the first device can select part of target positions (such as selecting M target positions from the N positions based on the first position selection mode) from the multiple positions corresponding to the radio frequency map to be transmitted, and then perform data extraction based on the selected part of target positions.

[0126] In some embodiments, the number (M) of target positions selected by the first device can be preconfigured or determined according to actual conditions (such as the size of the to-be-transmitted radio map, available transmission resources, position selection manner, etc.). For example, the second device or other devices (such as a core network device) can configure a position selection rate or a target position number selection table for the first device. The position selection rate can be used to indicate the proportion of the number of target positions selected from the radio map to the total number of positions in the radio map, and the target position number selection table can include the number of target positions to be selected corresponding to different radio map sizes. For example, assuming that the position selection rate is 0.1, in the case of a total number of positions in the to-be-transmitted radio map being 40, 5 target positions can be selected, and in the case of a total number of positions in the to-be-transmitted radio map being 100, 10 target positions can be selected. For another example, in the case of not considering resources, the position selection rate is 0.1, and the total number of positions in the to-be-transmitted radio map is 100, 10 target positions can be selected, but since the currently available transmission resources can support a maximum of 5 target positions, in order to ensure data transmission, the first device can only select 5 target positions.

[0127] In the embodiments of the present application, the manner in which the first device selects part of the target positions from the to-be-transmitted radio map can include various manners such as random selection, topology selection, grouping selection, etc. Further, since the first radio map can be a 2D radio map or a 3D radio map, in order to more clearly understand the embodiments of the present application, the above-mentioned selection manners will be introduced respectively for the two cases.

[0128] First, the case of the first radio map being a 2D radio map will be introduced, taking the size of the first radio map as m*n (m*n=N) as an example.

[0129] Random selection, that is, randomly selecting M target positions from the N positions corresponding to the first radio map. It can be understood that, in the case of meeting a certain reconstruction accuracy, this random selection manner is simple and fast.

[0130] The topology selection is selecting M target positions from the N positions corresponding to the first radio map according to the topology of the first radio map (the topology of the N first data). The topology can include the contour feature, the edge feature, the gradient feature of the first radio map, and other features associated with the topology of the first radio map. For example, the first device can extract the edge information / edge feature in the first radio map by a feature extraction algorithm (such as the canny algorithm), and then the positions corresponding to the extracted edge information can be selected as the target positions. For example, assuming that the first device determines that there is edge information in the positions 3, 4, 5, 10, and 11 of the first radio map after processing the first radio map by the canny algorithm, the first device can select the positions 3, 4, 5, 10, and 11 of the first radio map as the M target positions. For another example, the first device can calculate the gradient size of each position in the first radio map by a gradient extraction algorithm (such as the Sobel algorithm), and then the positions with a gradient greater than a certain threshold can be selected as the target positions. In some possible implementation, the first device can also select the M target positions from the N positions corresponding to the first radio map in combination with the topology of the first radio map and the topology of the environmental information (such as the point cloud information, the high-definition map, etc.) on the map. That is, the first device can also consider the environmental information associated with the first geographic area, such as the point cloud information, the high-definition map, etc. when selecting the target positions from the first radio map. It can be understood that, since the topology selection retains the associated features between the positions, the subsequently extracted data can also contain the associated features between the positions, so that the reconstruction accuracy can be higher.

[0131] The grouping selection is dividing the N positions corresponding to the first radio map into L groups of positions based on the N first data, and then selecting at least one position from each group of positions to obtain the M target positions. L is an integer greater than 1. For example, in some possible implementation, a plurality of value intervals can be pre-configured for the first data, and the positions in the first radio map can be divided into a plurality of groups based on the plurality of value intervals. For example, taking the signal strength as the first data, four value intervals (-100, -80), (-80, -60), (-60, -40), and (-40, -20) can be pre-configured, and the unit can be dBm (decibel milliwatt). Based on the four value intervals, the N positions corresponding to the first radio map can be divided into four groups, and then at least one position can be selected from each group of positions as the target position. Please refer to Figure 5 , Figure 5 is a schematic diagram of grouping disclosed by the embodiments of the present application. As shown in FIG. 4, the N positions corresponding to the first radio map are divided into four groups, and at least one position is selected from each group of positions as the target position. Figure 5As shown, based on the pre-configured value interval, the position (1, 1) can be divided into group 2, the position (1, 2) can be divided into group 1, the position (1, 3) can be divided into group 2, the position (1, 4) can be divided into group 3, the position (1, 5) and the position (1, 6) can be divided into group 2, the position (2, 1), the position (2, 3), the position (2, 4), the position (2, 5) and the position (2, 6) can be divided into group 1, the position (3, 1) can be divided into group 1, and the like. The first device can select at least one position from each of the group 1 to the group 4. It can be understood that the group selection can enable the first device to select certain positions from different groups, which can ensure that the subsequently extracted data can be more uniform, can avoid extracting data based on only positions corresponding to some larger signal strengths or extracting data based on only positions corresponding to some smaller signal strengths, so that the reconstruction performance can be more stable and the accuracy can be higher. It can be understood that the number of target positions selected by the embodiments of the present application in each group has multiple cases. In one case, for multiple groups, target positions can be uniformly selected in each group. For example, assuming that M is 12 and the number of groups is 4, then 3 target positions can be uniformly selected from each group. In another case, different groups correspond to different numbers of target positions to be selected. For example, some groups include positions that are more important, so these groups can be selected or selected more, while some groups include positions that are less important, so these groups can be selected less. In this way, the positions that are more important can correspond to better reconstruction effects and smaller errors. For example, the larger the signal strength, the more important the corresponding position. In this case, for the group 1 to the group 4, the number of target positions to be selected from them from small to large is group 4, group 3, group 2, and group 1, that is, the maximum number of target positions can be selected from the group 4 with the largest signal strength, and the minimum number of target positions can be selected from the group 1 with the smallest signal strength. For example, the position number selection rates of the group 1 to the group 4 can be [0.1, 0.2, 0.3, 0.4] respectively. The position number selection rate of a group is the proportion of the number of target positions to be selected from the group to the total number of target positions to be selected (M). Assuming that the total number of target positions to be selected is 10, then 1 target position can be selected from the group 1, 2 target positions can be selected from the group 2, 3 target positions can be selected from the group 3, and 4 target positions can be selected from the group 4.

[0132] The following introduces the position selection manner for one group. For each group, the corresponding target positions can be selected from the group in a random selection manner, or selected from the group based on a certain specific rule (such as an equal interval rule, a maximum distance rule, etc.). For example, assuming that a group needs to select z (such as 3) target positions, then z target positions can be randomly selected from the positions included in the group. The equal interval rule, that is, selecting target positions from a group in an equal interval manner. For example, assuming that a group needs to select z (such as 3) target positions, then z target positions can be selected from the positions included in the group in an equal interval manner. For example, assuming that a group includes 9 positions, and the number of target positions to be selected is 3, then 3 target positions can be selected in an interval of 2 or 3. When the interval is 3, the 1st position, the 5th position and the 9th position in the group can be selected; when the interval is 2, the 1st position, the 4th position and the 7th position in the group can be selected, or the 2nd position, the 5th position and the 8th position in the group can be selected, or the 3rd position, the 6th position and the 9th position in the group can be selected. When the equal interval manner is used, the size of the interval can be determined based on the total number of positions in a group and the number of positions to be selected, so that the selected multiple target positions are more dispersed. For example, the interval can be rounded down. It should be noted that the positions in a group can be sorted by rows or by columns, which is not limited herein. The maximum distance rule, that is, selecting target positions from a group in a maximum distance manner, that is, making the sum of distances between the multiple target positions selected from a group larger, so that the multiple target positions selected can be as dispersed as possible, so that the data extracted later can include more information of local areas. For example, assuming that the size of a radio frequency map is 5*6, a group includes positions (1, 1), (1, 2), (1, 5), (1, 6), (5, 1), (5, 2), (5, 5), (5, 6), and when 4 target positions need to be selected from the 8 positions included in the group, one can be selected from (1, 1) and (1, 2), one can be selected from (1, 5) and (1, 6), one can be selected from (5, 1) and (5, 2), and one can be selected from (5, 5) and (5, 6), so that the sum of distances between the selected target positions can be larger, and the multiple target positions selected can be more dispersed.

[0133] Secondly, the case that the first radio map is a 3D radio map is introduced. Taking the size of the first radio map as m*n*k (m*n*k=N) as an example, k can indicate that there are k different frequencies. In this case, one processing manner can regard the first radio map as k m*n 2D radio maps. Therefore, for each 2D radio map corresponding to the first radio map, the above-mentioned random selection, topological selection, grouping selection and the like can be adopted for position selection, and the corresponding target position can be selected in each 2D radio map. That is, in this case, the first device can independently perform position selection based on the 2D radio map corresponding to the k different frequencies. Moreover, the number of target positions selected by the first device in the 2D radio map corresponding to each frequency can be the same or different. The same case is that M / k target positions are uniformly selected in the 2D radio map corresponding to each frequency. The different case is that because the importance of the 2D radio map corresponding to the k frequencies is different, the number of target positions can be allocated according to the importance degree, and the more important the frequency is, the more target positions can be selected from the corresponding 2D radio map, and the less important the frequency is, the fewer target positions can be selected from the corresponding 2D radio map. For example, assuming that M is 10, the first radio map includes f1, f2 and f3 frequencies, f1 is less important than f2, and f2 is less important than f3. In this case, the number of target positions selected can be f3, f2 and f1 in turn from more to less, for example, 5 target positions can be selected from the 2D radio map corresponding to f3, 3 target positions can be selected from the 2D radio map corresponding to f2, and 2 target positions can be selected from the 2D radio map corresponding to f1. It should be understood that this manner of allocating the number of target positions based on the importance of the frequency can make the 2D radio map corresponding to the more important frequency have a better reconstruction effect when reconstructed later. The random selection, topological selection, grouping selection and the like in the case of the 2D radio map can be referred to the above-mentioned related description for detailed description.

[0134] For the case that the first radio frequency map is a 3D radio frequency map, in addition to the above-mentioned manner of splitting into k 2D radio frequency maps to select target positions respectively, M target positions can also be selected as a whole. For example, for the random selection manner, the first device can randomly select M target positions from the N positions corresponding to the first radio frequency map, for example, assuming that m is 5, n is 6, k is 6, and M is 10, the first device can randomly select 10 target positions from the m*n*k positions, such as position (1, 2, 1), position (3, 4, 2), position (4, 3, 2), position (2, 3, 3), position (3, 5, 3), position (4, 2, 4), position (4, 4, 4), position (3, 4, 4), position (3, 3, 5), and position (2, 4, 6). For example, for the topological selection manner, the first device can select M target positions from the N positions corresponding to the first radio frequency map according to the topological structure of the first radio frequency map (the topological structure of the N first data). The topological structure can include the contour feature, edge feature, gradient feature, and other features associated with the topological structure of the first radio frequency map, etc. For example, the first device can extract the edge information / edge feature in the first radio frequency map through a feature extraction algorithm (such as the canny algorithm), and then the positions corresponding to the extracted edge information can be used as target positions. For another example, the first device can calculate the gradient size corresponding to each position in the first radio frequency map through a gradient extraction algorithm (such as the Sobel algorithm), and then the positions with a gradient greater than a certain threshold can be used as target positions. In some possible embodiments, the first device can also select M target positions from the N positions corresponding to the first radio frequency map in combination with the topological structure of the first radio frequency map and the topological structure of the environment information (such as point cloud information, high-definition map, etc.) on the map. It can be understood that for the case that the first radio frequency map is a 3D radio frequency map, the contour feature, edge feature, gradient feature, etc. included in the topological structure of the first radio frequency map can be three-dimensional.

[0135] In a possible implementation, for a case where the first radio frequency map is a 3D radio frequency map, the first device can further select a plurality of positions (sub-regions) according to a 2D radio frequency map of the average value / median value, and a plurality of frequencies corresponding to each position in the 3D radio frequency map can be selected. Specifically, for a case where the first radio frequency map is a 3D radio frequency map, the first device can calculate the average value, median value, or maximum value of each position (sub-region) at all frequencies, and then obtain a corresponding 2D radio frequency map. Then, the first device can select a plurality of positions based on the 2D radio frequency map (for example, random selection, grouping selection, topological selection, etc.), and obtain a plurality of positions. Then, the first device can select positions corresponding to the plurality of positions at a plurality of frequencies as M target positions. For example, for a case where the size of the first radio frequency map is m*n*k (m*n*k=N), the first device can calculate the average value, median value, or maximum value of m*n positions (sub-regions) at k frequencies, and then obtain a 2D radio frequency map of m*n. Then, the first device can select a plurality of positions from the 2D radio frequency map of m*n, and then select positions corresponding to the plurality of positions at a plurality of frequencies (for example, k frequencies) as M target positions. Assuming that m is 5, n is 6, and k is 6, the positions selected by the first device from the 2D radio frequency map of m*n include position (1, 2) and position (4, 2). In this case, the M target positions selected by the first device in the first radio frequency map can include (1, 2, 1), (1, 2, 2), (1, 2, 3), (1, 2, 4), (1, 2, 5), (1, 2, 6), position (4, 2, 1), position (4, 2, 2), position (4, 2, 3), position (4, 2, 4), position (4, 2, 5), and position (4, 2, 6). In some embodiments, the first device can also select positions corresponding to part of the frequencies, for example, k1, k3, and k5. In this case, the M target positions selected by the first device in the first radio frequency map can include (1, 2, 1), (1, 2, 3), (1, 2, 5), position (4, 2, 1), position (4, 2, 3), and position (4, 2, 5).

[0136] Grouping selection, that is, N positions corresponding to N first data in the first radio frequency map are divided into L groups of positions, and then at least one position is selected from each group of positions in the L groups of positions to obtain M target positions. Grouping selection for a case where the first radio frequency map is a 3D radio frequency map is similar to grouping selection for a case where the first radio frequency map is a 2D radio frequency map, and the corresponding description of the grouping selection for the case where the first radio frequency map is a 2D radio frequency map can be referred to.

[0137] It should be understood that the above-mentioned several position selection manners are only illustrative and do not constitute a limitation. In addition to the above-mentioned several position selection manners, the first device can also select a target position based on other rules.

[0138] 403. The first device obtains K second data based on the M target positions, K being an integer less than N and greater than 0.

[0139] For example, after the first device selects M target positions from the N positions corresponding to the first radio map, the first device can perform data extraction based on the first data extraction manner and the M target positions to obtain K second data.

[0140] In the embodiments of the present application, the first data extraction manner can be original data extraction, weighted data extraction, or other data extraction manners such as maximum data extraction, which will be introduced below.

[0141] The original data extraction is to directly extract the original data (first data) corresponding to the selected M target positions to obtain K second data. In this case, K can be equal to M, and the second data is the first data. As can be seen, the original data extraction manner does not require additional processing and has high data extraction efficiency.

[0142] The weighted data extraction is to perform weighted calculation on the first data corresponding to the selected M target positions and the positions having a distance less than a first threshold (such as 2) from the M target positions to obtain K second data. In this case, K can be equal to M, and the second data is the extracted feature data or weighted data. For example, in the case of a 2D radio map, the distance between different positions can be considered in two dimensions. For example, assuming that position 1 is (m1, n1) and position 2 is (m2, n2), the distance between position 1 and position 2 can be max(|m1-m2|, |n1-n2|), or can be The following will be described by taking the first manner as an example. Max(a, b) represents the maximum value of a and b, and || is the absolute value symbol. The value of the first threshold can be determined according to the actual situation or can be obtained according to the training result of the reconstruction model. For example, different distance thresholds can be used for training during the training process, and finally the distance threshold corresponding to the best performance of the reconstruction model can be taken as the first threshold. Please refer to Figure 6 , Figure 6 is a schematic diagram of a weighted calculation disclosed in the embodiments of the present application. As shown in Figure 6As shown, assuming the first threshold value is 1, and one of the target positions is (3, 3), i.e. the position corresponding to S15, in this case, the positions having a distance less than the first threshold value from the target position (3, 3) can include the position (2, 2), the position (2, 3), the position (2, 4), the position (3, 2), the position (3, 4), the position (4, 2), the position (4, 3), and the position (4, 4). The weighting parameter (coefficient) matrix is a 3*3 matrix, which is W1~W9 respectively. The corresponding weighted value based on the target position (3, 3) can be (W1*S8+W2*S9+W3*S 10 +W4*S 14 +W5*S 15 +W6*S 16 +W7*S 20 +W8*S 21 +W9*S 22 ). If the target position also includes (4, 4), the weighted calculation can also be based on W1~W9, and different target positions can share the weighting parameter matrix. It can be understood that for the positions of the edge of the first radio frequency map (such as the positions corresponding to S1, S2, etc.), there can be no data in some directions near the positions, in this case, the corresponding positions can be supplemented with 0 for weighted calculation.

[0143] Further, in the case of the first radio frequency map being a 3D radio frequency map, the distance between different positions can be considered in two dimensions or three dimensions. In the case of considering two dimensions, the first radio frequency map can be regarded as k 2D radio frequency maps, and then the target positions in the k 2D radio frequency maps are weighted and calculated respectively, which can be referred to the above description. In the case of considering three dimensions, assuming that the position 1 is (m1, n1, k1), and the position 2 is (m2, n2, k2), the distance between the position 1 and the position 2 can be max(|m1-m2|, |n1-n2|, |k1-k2|), or can be The following is described by taking the first way as an example. For example, assuming that the first threshold value is 1, and one of the target positions is (3, 3, 2), in this case, the positions that are less than the first threshold value away from the target position (3, 3, 2) can include the positions (2, 2, 1), (2, 3, 1), (2, 4, 1), (3, 2, 1), (3, 3, 1), (3, 4, 1), (4, 2, 1), (4, 3, 1), (4, 4, 1), (2, 2, 2), (2, 3, 2), (2, 4, 2), (3, 2, 2), (3, 4, 2), (4, 2, 2), (4, 3, 2), (4, 4, 2), (2, 2, 3), (2, 3, 3), (2, 4, 3), (3, 2, 3), (3, 3, 3), (3, 4, 3), (4, 2, 3), (4, 3, 3), and (4, 4, 3). In this case, the weighted parameter matrix can be a 3*3*3 matrix, which is W1~W9 respectively. The weighted parameter matrix can be directly given or obtained based on the training of the reconstruction model. 27 The weighted parameter matrix can be shared by different target positions based on the target position (3, 3, 2) and the data corresponding to the positions within a distance of 1. It should be noted that the different parameters (such as W1~W9) in the above weighted parameter matrix can be the same or different. It should be further noted that the above weighted parameter matrix can be directly given or obtained based on the training of the reconstruction model.

[0144] It should be noted that in the above weighted data extraction manner, the first data corresponding to the selected M target positions and all positions within a distance of less than the first threshold value (such as 2) from the M target positions are weighted and calculated to obtain K second data. However, in some possible implementation manners, the first data corresponding to the selected M target positions and part of the positions within a distance of less than the first threshold value (such as 2) from the M target positions can be weighted and calculated to obtain K second data. The part of the positions within a distance of less than the first threshold value (such as 2) from the M target positions can be randomly selected or selected according to a preconfigured rule, such as selecting odd-numbered columns, selecting even-numbered columns, selecting positions with odd distances, selecting positions with even distances, and the like. For example, if the first threshold value is 2, the positions within a distance of less than 2 from the M target positions can include the positions (2, 2, 1), (2, 3, 1), (2, 4, 1), (3, 2, 1), (3, 3, 1), (3, 4, 1), (4, 2, 1), (4, 3, 1), (4, 4, 1), (2, 2, 2), (2, 3, 2), (2, 4, 2), (3, 2, 2), (3, 4, 2), (4, 2, 2), (4, 3, 2), (4, 4, 2), (2, 2, 3), (2, 3, 3), (2, 4, 3), (3, 2, 3), (3, 3, 3), (3, 4, 3), (4, 2, 3), (4, 3, 3), and (4, 4, 3). Figure 6As shown, assuming the first threshold value is 1, and one of the target positions is (3, 3), i.e., the position corresponding to S15, in this case, the positions having a distance less than the first threshold value from the target position (3, 3) can include positions (2, 2), (2, 3), (2, 4), (3, 2), (3, 4), (4, 2), (4, 3), and (4, 4). The first device can select a part of the positions (e.g., 3 positions) from the 8 positions and then perform a weighted calculation with the target position (3, 3). It can be understood that in some possible implementation, the first device can not include the selected target position when performing the weighted calculation, and can perform a weighted calculation based on the first data corresponding to all or part of the positions (not including the target position) having a distance less than the first threshold value (e.g., 2) from the selected M target positions to obtain K second data.

[0145] It can be understood that the weighted data extraction manner can make the extracted second data include data information of the target position and data information of the region near the target position, and can make the generalization better and the reconstruction effect better.

[0146] In some possible implementation, after the weighted calculation, a nonlinear transformation can also be performed on the weighted value, and the value obtained by the nonlinear transformation can be taken as the second data.

[0147] In the above two data extraction manners, one corresponding second data can be extracted for one target position. However, it should be understood that in some possible implementation, multiple corresponding second data can also be extracted for one target position, in which case K can be greater than M, or one corresponding second data can be extracted for multiple target positions, in which case K can be less than M. For example, the first device can extract the maximum value and the second maximum value near the target position, such as extracting the maximum value and the second maximum value within a range having a distance less than the first threshold value (e.g., 2) from the target position. For another example, the first device can extract the maximum value and the average value within a range having a distance less than the first threshold value (e.g., 2) from the target position. In some possible implementation, in the case of extracting multiple corresponding second data for one target position, one dimension can be expanded to splice the multiple second data corresponding to the target position, or the second data corresponding to the target position can be stored in different positions in the first radio map, in which case the relative positions of these positions from the center position of the radio map or the target position need to be known. In the case of extracting one corresponding second data for multiple target positions, the second data can be taken as the second data corresponding to the multiple target positions, and for example, the second data can be stored in the first radio map corresponding to the multiple target positions.

[0148] It can be understood that the electromagnetic information corresponding to the first geographic area can change over time, and therefore, in a possible implementation, the K second data extracted by the first device based on the target position can be time feature data. For example, for a target position, the first device can perform weighted calculation on the first data corresponding to the target position over a period of time to obtain a corresponding weighted value, which can be used as the second data, and the corresponding weighted matrix can be given or trained. It can also be understood that in the embodiments of the present application, the corresponding second data can also be obtained by combining time weighting and area weighting, which is not limited herein. Since the time feature data takes into account the time sequence information of the target position, it can play a role in data smoothing.

[0149] It can be understood that in some cases (such as random selection), the M target positions selected by the above-mentioned first position selection method have a certain randomness, that is, the M target positions selected by the first device through the first position selection method are not fixed. Therefore, in order to ensure that the second device can obtain a better reconstruction effect, in some possible embodiments, the first device can perform position selection and data extraction multiple times to obtain multiple sets of target positions and corresponding second data, and then the first device can input the multiple sets of target positions and corresponding second data into the first reconstruction model to obtain multiple reconstructed radio frequency maps, and then the error between the reconstructed radio frequency maps and the first radio frequency map can be calculated, and the position information of the M target positions and the K second data corresponding to the reconstructed radio frequency map with the smallest error can be sent to the second device. In the embodiments of the present application, the above-mentioned method can also be referred to as iterative selection. For example, the first device can perform multiple position selections on the first radio frequency map by random selection, and can obtain multiple sets of M target positions, and then can perform data extraction based on the multiple sets of M target positions to obtain multiple sets of K second data. Then, the first device can process the multiple sets of M target positions and corresponding K second data based on the first reconstruction model to obtain multiple reconstructed radio frequency maps. Then, the first device can calculate the error between the multiple reconstructed radio frequency maps and the first radio frequency map, and can send the M target positions and K second data corresponding to the reconstructed radio frequency map with the smallest error to the second device, and the M target positions and K second data corresponding to the reconstructed radio frequency map with relatively larger error can be discarded. It can be understood that in the iterative selection process, the way in which the first device performs position selection multiple times can be different or the same, and the embodiments of the present application are not limited in this regard.

[0150] 404. The first device sends the K second data and the position information of the M target positions to the second device.

[0151] After the first device extracts data based on the M selected target locations and obtains K second data, it can send the K second data and the location information of the M target locations to the second device. Correspondingly, the second device can receive the K second data and the location information of the M target locations from the first device. The location information of the M target locations can be used to indicate the corresponding locations of the M target locations selected by the first device in the first radio frequency map.

[0152] The location information of the M target locations may be numbers or indexes of the M target locations, or may be a location map (bitmap). The number of locations included in the location map may be the same as the number of locations included in the first radio frequency map. In the location map, the value of the location corresponding to the M target locations may be 1, and the value of the remaining locations may be 0. For example, see Figure 7 , Figure 7 This is a schematic diagram of a 2D position map disclosed in the embodiment of this application. Figure 7 As shown, assuming that the first radio frequency map is a 2D radio frequency map, the M target positions selected by the first device are position (2, 5), position (3, 3) and position (5, 5). In this case, in the corresponding 2D position mapping diagram, the values ​​corresponding to position (2, 5), position (3, 3) and position (5, 5) can be 1, and the values ​​corresponding to the remaining non-target positions can be 0, and the position compression rate can be 6*5 / 3 times.

[0153] It is understandable that there may be an association relationship between the K second data and the M target locations. In this case, the first device may also indicate the association relationship between the K second data and the M target locations to the second device. Exemplarily, the order of the location information of the K second data and the M target locations may implicitly indicate the corresponding association relationship. For example, assuming that Q second data are associated with the location information of one target location, then each Q second data in the K second data may be associated with one of the location information of the M target locations in turn. For example, assuming that K is equal to M and both are 5, the K second data may be {data 1, data 2, data 3, data 4, data 5}, and the location information of the M target locations may be {location information 1, location information 2, location information 3, location information 4, location information 5}. In this case, data 1 may be associated with location information 1, data 2 may be associated with location information 2, data 3 may be associated with location information 3, data 4 may be associated with location information 4, and data 5 may be associated with location information 5.

[0154] It should be noted that the embodiment of the present application does not limit the form of the position information. The position information can be a two-dimensional index, or the two-dimensional index can be converted into a one-dimensional vector (one-dimensional index). It can be converted in rows or columns. The embodiment of the present application does not limit this.

[0155] Optionally, in some possible implementation manners, the first device can further compress the K second data and / or the position information of the M target positions, and then send corresponding compressed data to the second device. The compression manner of the first device can be quantization, entropy coding, etc., which is not limited in the embodiments of the present application.

[0156] It should be noted that, in some embodiments, the size of the first radio map can be fixed, and the first device and the second device can know the size of the first radio map. Alternatively, in some embodiments, the first device can indicate the size of the first radio map to the second device, for example, by means of a position mapping diagram, or the first device can directly send the size information of the first radio map to the second device.

[0157] It can be understood that, in some cases, the first device can receive a radio map request message from the second device, which can be used to request the first radio map. Then, the first device can perform position selection and data extraction on the first radio map based on the radio map request message, to obtain the K second data and the position information of the M target positions, and then the first device can send the K second data and the position information of the M target positions to the second device.

[0158] 405. The second device reconstructs the radio map corresponding to the K second data and the position information of the M target positions based on the first reconstruction model to obtain a second radio map, the second radio map including N first data, and M and K are integers less than N and greater than 0.

[0159] After the second device receives the K second data and the position information of the M target positions from the first device, the second device can reconstruct the radio map corresponding to the K second data and the position information of the M target positions based on the first reconstruction model to obtain a second radio map. The first reconstruction model can be a pre-trained reconstruction model. The second radio map can be understood as a radio map inferred by the first reconstruction model based on the K second data and the position information of the M target positions. The second radio map can be associated with the first geographic region, and the second radio map can include N first data corresponding to N positions. The first reconstruction model can be a convolutional neural network (CNN), or other machine learning models or neural network models, which are not limited in the embodiments of the present application.

[0160] In some possible implementations, the second device may process the K second data and the position information of the M target locations to obtain third data, and the third data may be used as input data for the first reconstruction model. For example, the second device may map the K second data to the corresponding M target locations to obtain corresponding third data. Figure 8 As shown, assuming that M and K are both 3, the data extraction method is original data extraction, and the three second data received by the second device are {S 11 、S 15 、S 29}, the values ​​of position (2, 5), position (3, 3) and position (5, 5) in the received corresponding position map can be 1. Based on this, {S 11 、S 15 、S 29} is mapped to the corresponding position to obtain the third data. It is understandable that, in some possible implementations, the second device can input the K second data and the position information of the M target positions into the first reconstruction model.

[0161] In some possible implementations, different location selection methods and / or data extraction methods may correspond to different reconstruction models. Accordingly, in order to ensure the reconstruction effect, the second device may store multiple reconstruction models, and the second device may use the first reconstruction model corresponding to the K second data and the location information of the M target locations for reconstruction. Exemplarily, after the first device selects the target location using the first location selection method and extracts the corresponding second data using the first data extraction method, it may send first information to the second device. The first information may include identification information of the first location selection method and / or identification information of the first data extraction method. Accordingly, the second device may receive the first information from the first device, and then may determine the first reconstruction model corresponding to the first location selection method and / or the first data extraction method. Thereafter, the first device may reconstruct the radio frequency map corresponding to the location information of the K second data and the M target locations based on the first reconstruction model.

[0162] It can be understood that, in some possible implementation, in order to ensure that the second device can employ a correct reconstruction model to perform reconstruction, the first device and the second device can also negotiate the corresponding position selection manner and / or data extraction manner in advance, for example, the second device sends configuration information to the first device, the configuration information can include the position selection manner and / or data extraction manner that the second device hopes to employ, for another example, the first device can send configuration information to the second device, the configuration information can include the position selection manner and / or data extraction manner that the first device employs. Alternatively, in some other possible implementation, the position selection manner and / or data extraction manner can be configured for the first device and the second device by other devices (such as a core network device).

[0163] Optionally, in some cases, after the first device selects the partial target positions from the N positions corresponding to the first radio map, the first device can continue to select partial target positions, and then can send the position information of the newly selected partial target positions and the second data extracted based on the newly selected partial target positions to the second device, so that the second device can perform reconstruction based on more data when performing reconstruction, and the accuracy of reconstruction can be improved and the reconstruction error can be reduced. For example, the first device can select J target positions from the N positions corresponding to the first radio map, the J target positions are different from the target positions that have been sent, that is, different from the target positions (such as the M target positions) selected from the N positions corresponding to the first radio map before, then the first device can obtain H second data based on the J target positions, and then the first device can send the H second data and the position information of the J target positions to the second device. J and H are integers less than N and greater than 0. Correspondingly, the second device can receive the H second data and the position information of the J target positions from the first device, the J target positions are different from the target positions that have been received, that is, different from the target positions corresponding to the position information that has been received (such as the position information of the M target positions that has been received) before, then the second device can reconstruct a third radio map based on the received second data and the position information of the target positions (such as the K second data and the position information of the M target positions described above), and the H second data and the position information of the J target positions. The third radio map can be associated with the first geographic region, and the third radio map can include N first data corresponding to N positions.

[0164] It can be understood that the first reconstruction model used by the second device can be a pre-trained reconstruction model. The first reconstruction model can be trained by the first device, the second device, or other devices (such as an AI device in the core network), and the embodiments of the present application do not limit this. In the case where the first reconstruction model is trained by the first device, the first device can obtain a plurality of radio frequency maps, which are also training data. Then, the first device can process the plurality of radio frequency maps based on the first position selection manner and the first data extraction manner respectively to obtain a plurality of sets of model input data. Then, the first device can train the initial first reconstruction model based on the plurality of sets of model input data and the plurality of radio frequency maps to obtain the trained first reconstruction model. Training the initial first reconstruction model based on the plurality of sets of model input data and the plurality of radio frequency maps means that the plurality of sets of model input data are first input into the first reconstruction model to obtain a plurality of reconstructed radio frequency maps, and then the parameters of the first reconstruction model are optimized based on the error between the reconstructed radio frequency maps and the corresponding original radio frequency maps (training data). One position selection and one data extraction for an original radio frequency map can obtain one set of reconstruction model input data. Correspondingly, inputting one set of reconstruction model input data into the first reconstruction model can obtain one reconstructed radio frequency map. After the first device trains the first reconstruction model, the first device can send the trained first reconstruction model to the second device, such as the parameters of the model, the structure of the model, etc. Correspondingly, the second device can receive the trained first reconstruction model from the first device, and can deploy the trained first reconstruction model. In the case where the first reconstruction model is trained by the second device, the second device can obtain a plurality of radio frequency maps, and then the second device can process the plurality of radio frequency maps based on the first position selection manner and the first data extraction manner respectively to obtain a plurality of sets of model input data. Then, the second device can train the initial first reconstruction model based on the plurality of sets of model input data and the plurality of radio frequency maps to obtain the trained first reconstruction model. After the second device trains the first reconstruction model, the second device can deploy the trained first reconstruction model. In some possible embodiments, the second device can also send the trained first reconstruction model to the first device. In the case where the first reconstruction model is trained by a third device other than the first device and the second device, the third device can obtain a plurality of radio frequency maps, and then the third device can process the plurality of radio frequency maps based on the first position selection manner and the first data extraction manner respectively to obtain a plurality of sets of model input data. Then, the third device can train the initial first reconstruction model based on the plurality of sets of model input data and the plurality of radio frequency maps to obtain the trained first reconstruction model. After the third device trains the first reconstruction model, the third device can send the trained first reconstruction model to the first device and the second device.

[0165] It can be understood that in actual cases, the resources of the first device for transmitting the compressed data corresponding to the radio map (such as the K second data and the position information of the M target positions described above) are variable. When the available transmission resources are more, the first device can select more target positions from the radio map, and when the available transmission resources are less, the first device can select less target positions from the radio map. For this case, it is necessary to make the first reconstruction model support different position selection rates, so in the process of processing a plurality of radio maps based on the first position selection mode and the first data extraction mode respectively to obtain a plurality of sets of model input data, different position selection rates can be used, that is, the plurality of sets of model input data can be obtained based on different position selection rates.

[0166] It should be noted that the first device or the second device or other devices can also use the second position selection mode and the second data extraction mode to process a plurality of radio maps respectively to obtain a plurality of sets of model input data, and then train the corresponding reconstruction model based on the plurality of sets of model input data, so as to obtain the reconstruction model corresponding to the second position selection mode and the second data extraction mode. The second position selection mode is different from the first position selection mode, and / or the second data extraction mode is different from the first data extraction mode. Even in some embodiments, when training the first reconstruction model, different data position selection modes and / or data extraction modes can also be used to obtain model input data, and the embodiments of the present application are not limited thereto.

[0167] In the above processing flow, the first device can select part of the positions in the first radio map, and then extract data based on the selected part of the positions. Then, the first device can send the extracted data and the position information of the selected part of the positions to the second device. The second device can reconstruct the radio map corresponding to the received data and the position information of the part of the positions based on the reconstruction model. In this way, the compression benefit can be greatly improved. Moreover, the embodiments of the present application include a plurality of position selection modes and a plurality of data extraction modes, which can be flexibly adapted to different scenarios. In addition, by obtaining model data input based on different position selection rates in the model training stage, the trained reconstruction model can be applicable to different position selection rates (compression rate conditions).

[0168] The training of the first reconstruction model will be described below. Please refer to Figure 9 , Figure 9 is a flowchart of a reconstruction model training method disclosed by an embodiment of the present application. As shown in Figure 9 , the method can include but is not limited to the following steps:

[0169] 901. The second device sends training data, a configuration model and configuration information to the first device.

[0170] The training data can be a plurality of radio frequency maps, or the training data can be a plurality of radio frequency maps and data obtained based on position selection and data extraction of the plurality of radio frequency maps. It can be understood that the embodiments of the present application do not limit the manner in which the second device obtains the training data, which can be sent by the second device or other devices, or stored locally in the first device.

[0171] The configuration model, that is, the model to be trained, can be an initial model (such as an initial first reconstruction model) or a model to be updated (such as a first reconstruction model pre-trained or trained based on a small amount of data). It can be understood that the embodiments of the present application do not limit the manner in which the second device obtains the configuration model, which can be sent by the second device or other devices, or stored locally in the first device.

[0172] The configuration information can include position selection methods, data extraction methods, selection rates k / selection rate sets, compression configurations (such as compression methods, compression parameters, etc.), and can include parameters corresponding to the position selection methods and parameters (such as weighting parameters) corresponding to the data extraction methods. In the case of random selection of position selection, the corresponding parameters can include random number seeds, etc., and in the case of grouping selection of the position selection method, the corresponding parameters can include a plurality of grouping intervals, selection rates corresponding to each grouping, etc. It can be understood that the embodiments of the present application do not limit the manner in which the second device obtains the configuration information, which can be sent by the second device or other devices, or stored locally in the first device.

[0173] It can be understood that step 901 is optional, for example, the training data, the configuration model, and the configuration information come from other devices.

[0174] 902. The first device trains the first reconstruction model based on the training data.

[0175] After the first device obtains the training data, the configuration model, and the configuration information, the first device can train the configuration model based on the training data and the configuration information. Please refer to Figure 10 , Figure 10 is a schematic diagram of model training disclosed by the embodiments of the present application. As shown in Figure 10As shown, the model can include two parts, an encoding model and a decoding model (such as a CNN model), the encoding model is mainly used for position selection and data extraction. The decoding model is mainly used for reconstructing the corresponding radio frequency map based on the selected target position information and the extracted second data. In some embodiments, the encoding model can also perform data compression, and accordingly, the decoding model can also be used for data decompression. During model training, the first device can input the training data (original radio frequency map) into the encoding model for position selection and data extraction, to obtain the corresponding position information and second data, which can be used as input data of the decoding model. The first device can input the position information and second data output by the encoding model into the decoding model, and the decoding model can output the reconstructed radio frequency map. Then, the first device can calculate the error (such as mean-square error (MSE)) between the reconstructed radio frequency map and the original radio frequency map. Then, the first device can optimize (update) the model parameters of the decoding model based on the error between the reconstructed radio frequency map and the original radio frequency map. In some embodiments, the encoding model can also include model parameters that can be trained, such as weighting parameters when the data extraction method is weighted data extraction. In this case, the related parameters in the encoding model can also be updated based on the model error. It should be noted that in the embodiments of the present application, the data extraction can adopt an AI method (such as training weighting parameters, CNN network, etc.), or a non-AI method (such as directly giving weighting parameters, original data extraction, etc.). For example, the data extraction can be a position area feature encoding module (such as a CNN module), or a position area weighting coefficient (such as the above 3*3 weighting parameter matrix).

[0176] In some possible implementation manners, the first device can use different model parameters for the encoding model in the model training process, such as a position selection corresponding position selection rate, a position selection manner, a data extraction manner, and the like. Taking the use of different position selection rates as an example, for the input data of each round of parameter update or each decoding model, a position selection rate can be selected from a configured set of position selection rates for processing to obtain the output data of the decoding model. In some embodiments, for each position selection rate in the set of selection rates, a corresponding selection probability can be set, and the sum of the selection probabilities corresponding to all the position selection rates in the set of selection rates can be 1. In this way, the preference of the decoding model for the position selection rate can be adjusted. For example, assuming that in actual situations, the position selection rate 1 is used more frequently, the position selection rate 2 and the position selection rate 3 are used less frequently, and the position selection rate 4 is used the least frequently, the selection probability corresponding to the position selection rate 1 can be 0.5, the selection probability corresponding to the position selection rate 2 and the position selection rate 3 can be 0.2, and the selection probability corresponding to the position selection rate 4 can be 0.1. In this way, the finally trained decoding model can perform better in the case of the position selection rate being the position selection rate 1.

[0177] In the embodiments of this application, the same original radio frequency map can be processed by the encoding model for multiple times to obtain multiple different outputs, so as to increase the training data corresponding to the decoding model. For example, the encoding model processes the same original radio frequency map for different position selection rates, and the outputs obtained can be different. For example, the encoding model processes the same original radio frequency map for random selection, group selection, and the like, and the outputs obtained can also be different because the target positions selected can have randomness.

[0178] It can be understood that, in the iterative training process of the decoding model, when the total loss value (error value) is less than the preset model error, the training can be stopped to obtain the trained decoding model. Optionally, the first device can set a preset iteration number of the initial decoding model and record the training iteration number of the initial decoding model. When the training iteration number is equal to the preset iteration number, the training of the initial decoding model can be stopped, and the model with the training iteration number equal to the preset iteration number is determined as the finally trained decoding model.

[0179] It should be noted that the manner of updating the model parameters is not limited in the embodiments of this application, and various parameter updating manners such as stochastic gradient descent and batch gradient descent can be used. It can be understood that when the model parameters are updated by the gradient descent method, the gradient can be propagated in reverse, such as from the decoding model to the encoding model.

[0180] In the embodiments of this application, the first reconstruction model can be the decoding model.

[0181] 903. The first device sends the trained first reconstruction model to the second device.

[0182] After the first device trains the first reconstruction model to obtain the trained first reconstruction model, the first device can send the trained first reconstruction model to the second device. In addition, the first device itself can also deploy the encoding model, such as the feature extraction model (such as the weighted feature extraction module) corresponding to the selected target position.

[0183] 904. The second device deploys the trained first reconstruction model.

[0184] After the second device receives the trained first reconstruction model from the first device, the second device can deploy the trained first reconstruction model.

[0185] It can be understood that in some cases, the training of the model can also be performed by the second device. Please refer to Figure 11 , Figure 11 is another flowchart of a reconstruction model training method disclosed in the embodiments of the present application. As Figure 11 shown, the method can include but is not limited to the following steps:

[0186] 1101. The first device sends training data, a configuration model, and configuration information to the second device.

[0187] Step 1101 is similar to step 901 described above, and reference can be made to the related description in step 901 described above.

[0188] It can be understood that step 1101 is optional, for example, the training data, the configuration model, and the configuration information come from other devices.

[0189] 1102. The second device trains the first reconstruction model based on the training data.

[0190] Step 1102 is similar to step 902, and reference can be made to the related description in step 902 described above.

[0191] 1103. The second device deploys the trained first reconstruction model.

[0192] After the second device trains the first reconstruction model, the second device can deploy the trained first reconstruction model.

[0193] 1104. The second device sends the trained encoding model to the first device.

[0194] In some embodiments, during the model training process, the related parameters of the encoding model can be trained, and in this case, the second device can send the trained encoding model to the first device, that is, send the related parameters of the encoding model, such as the weighting parameter, to the first device.

[0195] It can be understood that in some embodiments, the second device can also send the trained first reconstruction model (decoding model) to the second device, so that the first device can then perform iterative selection, layer-by-layer transmission, etc.

[0196] 1105. The first device deploys the trained encoding model.

[0197] After the first device receives the encoding model from the second device, the trained encoding model can be deployed.

[0198] It can be understood that steps 1104 and 1105 are optional, such as in the case of random selection and original data extraction, the decoding model parameters that need to be trained can not be included.

[0199] It should be noted that in addition to the first device or the second device training the encoding model and the decoding model, other devices can also train the encoding model and the decoding model, which is not limited here.

[0200] It can be understood that in order to ensure that the first device and the second device use corresponding encoding models and decoding models, in some possible embodiments, the first device and the second device can first negotiate configuration information, and then train the model, compress the radio map (position selection, data extraction, and data compression), etc. Please refer to Figure 12 , Figure 12 is another flow diagram of the data processing method disclosed in the embodiments of the present application. As Figure 12 shown, the method can include but is not limited to the following steps:

[0201] 1201. The second device sends configuration information to the first device.

[0202] The configuration information can include position selection method configuration information and data extraction method configuration information.

[0203] The position selection mode configuration information can comprise identification information of the first position selection mode. For example, the identification information of the first position selection mode can be a number or an index of the first position selection mode, or other information that can identify the first position selection mode. The first position selection mode can be any one of random selection, topology selection, grouping selection, and iteration selection. In some embodiments, the position selection mode configuration information can further comprise parameters corresponding to the first position selection mode. For example, when the first position selection mode is random selection, the position selection mode configuration information can further comprise a random seed corresponding to the random selection. For another example, when the first position selection mode is grouping selection, the position selection mode configuration information can further comprise a grouping interval. For another example, when the first position selection mode is iteration selection, the position selection mode configuration information can further comprise a maximum number of iterations. For another example, the position selection mode configuration information can further comprise one or more position selection rates (a set of selection rates).

[0204] The data extraction mode configuration information can comprise identification information of the first data extraction mode. For example, the identification information of the first data extraction mode can be a number or an index of the first data extraction mode, or other information that can identify the first data extraction mode. The data extraction mode can be any one of raw data extraction, weighted data extraction, and time feature data extraction. In some embodiments, the data extraction mode configuration information can further comprise parameters corresponding to the first data extraction mode. For example, when the first data extraction mode is weighted data extraction, the data extraction mode configuration information can further comprise weighting parameters (including a weighting area size, a weighting coefficient, etc.). For another example, when the first position selection mode is time feature data extraction, the data extraction mode configuration information can further comprise a time weighting interval, i.e., an indication of a time range for which data is weighted.

[0205] It can be understood that, in some possible implementations, the configuration information can also be configured by the first device, which can send the configuration information to the second device. Alternatively, in some possible implementations, the configuration information can be configured by a third device, which can send the configuration information to the first device and the second device.

[0206] 1202. The first device compresses the first radio map based on the configuration information to obtain compressed data.

[0207] 1203. The first device sends the compressed data corresponding to the first radio map to the second device.

[0208] 1204. The second device processes the compressed data corresponding to the first radio map based on the first reconstruction model to obtain a reconstructed radio map.

[0209] It can be understood that in some cases, if the target position selected by the first device is less (such as the corresponding position selection rate is 0.01), the reconstruction effect of the second device based on the compressed data sent by the first device may be poor. For example, due to the limited transmission resources available to the first device, the compressed data that the first device can send will also be limited accordingly, and accordingly, the number of target positions selected by the first device from the first radio map will also be limited, so that the second device in the radio map reconstruction may cause poor reconstruction effect, for this case, the first device can perform layering transmission, that is, on the basis of the first time sending compressed data, continue to select target positions and data extraction, and then send the corresponding new compressed data to the second device. Please refer to Figure 13 , Figure 13 is a flow diagram of a layering transmission disclosed in an embodiment of the present application. As shown in Figure 13 , the method can include but is not limited to the following steps:

[0210] 1301. The second device sends configuration information to the first device.

[0211] Correspondingly, the first device can receive the configuration information from the second device, which can include the first position selection method and the parameters corresponding to the first position selection method, the first data extraction method and the parameters corresponding to the first data extraction method, and the base layer position selection rate. The base layer position selection rate can also be referred to as layer 1 position selection rate.

[0212] It can be understood that step 1301 is optional. For example, in some possible embodiments, the third device can send the configuration information to the first device.

[0213] 1302. The first device obtains base layer compressed data based on the base layer position selection rate.

[0214] The first device can obtain the corresponding compressed data based on the base layer position selection rate, that is, the base layer compressed data. For example, the first device can first select a plurality of target positions from the radio map to be transmitted based on the first position selection method and the base layer position selection rate, and then can extract data based on the plurality of target positions selected by the first data extraction method. After that, the first device can send the extracted data and the position information of the plurality of target positions selected to the second device. That is, the base layer compressed data can include the position information of the plurality of target positions selected based on the first position selection and the base layer position selection rate, and the data extracted based on the plurality of target positions. Optionally, the first device can also compress (such as entropy encoding, etc.) the position information and the extracted data.

[0215] 1303. The first device sends the base layer compressed data to the second device.

[0216] Accordingly, the second device may receive the base layer compressed data from the first device.

[0217] 1304. The first device sends a termination message or a performance message to the second device.

[0218] Correspondingly, the second device can receive a termination message or a performance message from the first device, and the termination message or the performance message can be associated with layer 1, that is, the termination message or the performance message is a termination message or the performance message corresponding to layer 1. The termination message can be used to indicate whether the first device continues to send the added layer compressed data. For example, the termination message can be 0 or 1, 0 represents the termination of sending the added layer data, and 1 represents the continuation of sending the added layer data. The performance message (error data) can be used to indicate the reconstruction performance or reconstruction error of the corresponding compressed data. For example, the performance message can be the mean square error between the reconstructed radio frequency map obtained by the first device based on all the compressed data currently sent or all the compressed data currently generated and the corresponding original radio frequency map.

[0219] In some possible implementations, the first device may obtain a termination message or a performance message based on the mean square error between the reconstructed radio frequency map obtained by reconstructing all the compressed data currently sent or all the compressed data currently generated and the corresponding original radio frequency map. Specifically, the first device may locally store a corresponding mean square error threshold (second threshold). When the mean square error between the reconstructed radio frequency map obtained by reconstructing all the compressed data currently sent or all the compressed data currently generated and the corresponding original radio frequency map is greater than the second threshold, the termination message sent by the first device may be 0. When the mean square error between the reconstructed radio frequency map obtained by reconstructing all the compressed data currently sent or all the compressed data currently generated and the corresponding original radio frequency map is less than or equal to the second threshold, the termination message sent by the first device may be 1.

[0220] Optionally, the second device may send the second threshold to the first device, or the third device may send the second threshold to the first device.

[0221] It is understandable that step 1301 is optional.

[0222] 1305. The second device sends an increase layer sending instruction to the first device.

[0223] For example, the second device can send a layer-increase sending instruction to the first device, where the layer-increase sending instruction can be used to instruct the first device to send layer-increase data (e.g., layer 2 compressed data). The layer-increase sending instruction can include a layer 2 position selection rate. In some embodiments, the layer-increase sending instruction can also not include the layer 2 position selection rate. For example, the layer 2 position selection rate can be pre-configured for the first device. For another example, the first device can determine the layer 2 position selection rate according to an error corresponding to the layer 1 position selection rate.

[0224] In step 1304, after the second device receives the performance message, the second device can determine whether the performance message satisfies the performance requirement of the second device. If yes, the second device can not send the layer-increase sending instruction. If no, the second device can send the layer-increase sending instruction. For example, after the second device receives the performance message, the second device can determine whether the performance message is greater than a second threshold. If yes, the second device can determine that the performance requirement of the second device cannot be satisfied, and the second device can send the layer-increase sending instruction to the first device. If no, the second device can determine that the performance requirement of the second device can be satisfied, and the second device can not send the layer-increase sending instruction to the first device.

[0225] It can be understood that step 1305 is optional. For example, in the case where the first device sends the termination message in step 1304, the first device can determine whether the first device will continue to send layer-increase data based on the termination message. In this case, the second device can not send the layer-increase sending instruction to the first device.

[0226] 1306. The first device obtains layer 2 compressed data based on the layer 2 position selection rate.

[0227] Step 1306 is similar to step 1302, and reference can be made to the related description in step 1302 above, which will not be repeated here.

[0228] 1307. The first device sends the layer 2 compressed data to the second device.

[0229] Correspondingly, the second device can receive the layer 2 compressed data from the first device.

[0230] 1308. The first device sends a termination message or a performance message to the second device.

[0231] Correspondingly, the second device can receive the termination message or the performance message from the first device, which can be associated with layer 2, i.e., the termination message or the performance message is a layer 2 corresponding termination message or performance message. The layer 2 corresponding performance message can be obtained based on the base layer compressed data and the layer 2 compressed data to reconstruct the reconstructed radio frequency map and the corresponding original radio frequency map, such as the mean square error between the reconstructed radio frequency map obtained based on the base layer compressed data and the layer 2 compressed data and the corresponding original radio frequency map.

[0232] For example, in the case that the mean square error between the reconstructed radio frequency map obtained based on the base layer compressed data and the layer 2 compressed data and the corresponding original radio frequency map is greater than the second threshold, the termination message sent by the first device can be 0, and in the case that the mean square error between the reconstructed radio frequency map obtained based on the base layer compressed data and the layer 2 compressed data and the corresponding original radio frequency map is less than or equal to the second threshold, the termination message sent by the first device can be 1.

[0233] It can be understood that step 1308 is optional.

[0234] 1309. The second device obtains a reconstructed radio frequency map based on the base layer compressed data and the layer 2 compressed data.

[0235] After the second device receives the layer 2 compressed data from the first device, the second device can integrate the base layer compressed data and the layer 2 compressed data to obtain integrated data, and then can reconstruct based on the integrated data to obtain a reconstructed radio frequency map.

[0236] It should be noted that the target positions selected by the first device at each layer can be different, i.e., the position information corresponding to different layers can be different. For example, for the base layer compressed data, the target positions selected by the first device from the to-be-transmitted radio frequency map can be position 1 and position 2, and for the layer 2 compressed data, the target positions selected by the first device from the to-be-transmitted radio frequency map can be position 5 and position 6.

[0237] It can be understood that the above processing flow only illustrates the case of two layers of increments, but the number of layers of increments is not limited in the embodiments of the present application, and can also be three layers, four layers or more layers. For example, in the case that the layer 2 corresponding termination message is 1, the first device can also send layer 3 corresponding compressed data to the second device. Or, in the case that the second device determines that the performance requirement is not met based on the layer 2 corresponding performance message, the second device can continue to send an increment layer sending instruction to the second device to trigger the first device to continue to send layer 3 corresponding compressed data.

[0238] It can be understood that, in combination with the position selection rate adaptation (the reconstruction model is applicable to different position selection rates), the layer-by-layer transmission process can be designed, so that the reconstruction performance under the condition of limited transmission resources can be ensured.

[0239] Some simulation diagrams of the present scheme are provided below.

[0240] Please refer to Figure 14 , Figure 14 is a reconstruction error diagram under different conditions disclosed by the embodiments of the present application. Figure 14 The model used can be a convolutional neural network including 26 convolutional layers, and the position selection rate or compression rate can be 0.02-0.1. Figure 14 The combinations of the three position selection methods and data extraction methods include random selection+original data extraction, grouping selection+original data extraction, and random selection+weighted data extraction. Among them, the deepest color, that is, the uppermost broken line, corresponds to random selection+original data extraction. The lightest color, that is, the lowermost broken line, corresponds to random selection+weighted data extraction. The broken line in the middle corresponds to grouping selection+original data extraction.

[0241] From Figure 14 it can be seen that, in the case of original data extraction in the data extraction method, the error corresponding to grouping selection is smaller than that of random selection. In the case of random selection in the position selection method, the error corresponding to weighted data extraction is smaller than that of original data extraction. Among them, when the position selection rate is 0.08, the model performs well and the error is small. Based on this, in some cases, the position selection rate used can be determined according to the actual performance of the model (such as the performance on the test set).

[0242] Please refer to Figure 15 , Figure 15 is a comparison diagram of radio map reconstruction disclosed by the embodiments of the present application. As Figure 15 indicated, for the original radio map, random selection and grouping selection can be used for processing, a plurality of target positions (the position selection rate is 0.04) can be selected, the same data extraction method (such as original data extraction) can be used for data extraction, and finally the reconstruction can be performed based on the position information of the selected target positions and the extracted data to obtain the reconstructed radio map. Among them, the mean square error between the reconstructed radio map and the original radio map under the random selection method can be 8.33 dB, and the mean square error between the reconstructed radio map and the original radio map under the grouping selection method can be 3.94 dB. It can be seen that the mean square error corresponding to the grouping selection method is smaller than that of the random selection method. In Figure 15In this way, the part of the circle mark in the reconstructed radio map is randomly selected to have a larger deviation from the corresponding position in the original radio map, which is caused by the randomness of the random selection.

[0243] Please refer to Figure 16 , Figure 16 is another comparison diagram of the radio map reconstruction disclosed by the embodiment of the present application. As shown in Figure 16 , for the original radio map, the original data extraction and the weighted data extraction can be used for data extraction, and the reconstructed radio map can be obtained based on the position information of the selected target position and the extracted data. The mean square error between the reconstructed radio map and the original radio map in the original data extraction mode can be 8.33 dB, and the mean square error between the reconstructed radio map and the original radio map in the grouping selection mode can be 3.7534 dB. It can be seen that the mean square error corresponding to the weighted data extraction mode is smaller than that of the original data extraction mode.

[0244] It can be understood that the scheme provided by the present application is to improve the compression efficiency of the radio map at the cost of a smaller error, thereby saving transmission resources. In the case where the error between the original radio map and the reconstructed radio map is small (e.g., less than a second threshold), the reconstructed radio map is almost the same as the original radio map, and the reconstructed radio map can be used normally without affecting.

[0245] It should be noted that the related information (i.e., the same information or similar information) and the related description in the above different embodiments can be mutually referred.

[0246] It should be understood that the above various flowcharts (such as Figure 4 , Figure 9 , Figure 11 , Figure 12 , Figure 13 , etc.) mainly take the first device and the second device as the interactive execution subject to illustrate the above processing flow, but the present application does not limit the interactive execution subject. For example, the first device in Figure 4 may also be a chip, a chip system, or a processor supporting the first device to implement the method, and can also be a logic module or software capable of implementing all or part of the function of the first device. For another example, Figure 4 the second device in may also be a chip, a chip system, or a processor supporting the second device to implement the method, and can also be a logic module or software capable of implementing all or part of the function of the second device.

[0247] The above mainly introduces the data processing method provided by the embodiments of the present application. It can be understood that the first device and the second device can include the hardware structure and / or software module corresponding to each function in order to realize the above corresponding functions. The units and steps of each example described in combination with the embodiments disclosed herein can be implemented in hardware or a combination of hardware and computer software. Whether a certain function is implemented in hardware or computer software driven hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of the present application.

[0248] The embodiments of the present application can divide the first device, the second device, etc. into functional modules according to the above method examples. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one module. The above integrated module can be realized in the form of hardware or software functional module. It should be noted that the division of the module by the embodiments of the present application is illustrative, and is only a logical functional division. Actual implementation can have another division method.

[0249] In the case of dividing each functional module according to each function, Figure 17 A possible structural diagram of a data processing apparatus 1700 is shown. The data processing apparatus 1700 includes a processing unit 1701 and a sending unit 1702. In a possible design, the data processing apparatus 1700 can be the first device described above, or can be a chip in the first device, or can be a processing system in the first device, etc. Wherein:

[0250] The processing unit 1701 is configured to obtain first data information, the first data information including N first data, the N first data corresponding to N positions, N being an integer greater than 1;

[0251] The processing unit 1701 is further configured to select M target positions from the N positions, M being an integer less than N and greater than 0;

[0252] The processing unit 1701 is further configured to obtain K second data based on the M target positions, K being an integer less than N and greater than 0;

[0253] The sending unit 1702 is configured to send the K second data and the position information of the M target positions to the second device.

[0254] In a possible implementation, the first data information is two-dimensional data information or three-dimensional data information.

[0255] In a possible implementation, the first data information is a radio map.

[0256] In a possible implementation, the processing unit 1701 selects M target positions from the N positions based on the first position selection manner, where the first position selection manner is one of random selection, topology selection, grouping selection, and iteration selection.

[0257] In a possible implementation, when the first position selection manner is grouping selection, the processing unit 1701 selects M target positions from the N positions based on the first position selection manner includes: dividing the N positions into L groups of positions based on the N first data, where L is an integer greater than 1; and selecting at least one position from each of the L groups of positions to obtain the M target positions.

[0258] In a possible implementation, when the first position selection manner is topology selection, the processing unit 1701 selects M target positions from the N positions based on the first position selection manner includes: selecting M target positions from the N positions based on a topology structure of the N first data.

[0259] In a possible implementation, the processing unit 1701 obtains K second data based on the M target positions includes: obtaining K second data based on a first data extraction manner and the M target positions, where the first data extraction manner is one of original data extraction and weighted data extraction.

[0260] In a possible implementation, when the first data extraction manner is weighted data extraction, the processing unit 1701 obtains K second data based on the first data extraction manner and the M target positions includes: performing weighted calculation on the first data corresponding to the M target positions and all or part of positions that are less than a first threshold from the M target positions, respectively, to obtain K second data.

[0261] In a possible implementation, the sending unit 1702 is further configured to send, to the second device, first information, where the first information includes identification information of the first position selection manner and / or identification information of the first data extraction manner.

[0262] In a possible implementation, the processing unit 1701 is further configured to select J target positions from the N positions, where the J target positions are different from the target positions that have been sent, and J is an integer less than N and greater than 0; the processing unit 1701 is further configured to obtain H second data based on the J target positions, where H is an integer less than N and greater than 0; and the sending unit 1702 is further configured to send, to the second device, the H second data and position information of the J target positions.

[0263] In a possible implementation, the processing unit 1701 is further configured to obtain a plurality of data information; the processing unit 1701 is further configured to process the plurality of data information based on the first position selection manner and the first data extraction manner respectively, to obtain a plurality of groups of model input data; the processing unit 1701 is further configured to train an initial first reconstruction model based on the plurality of groups of model input data and the plurality of data information, to obtain a trained first reconstruction model; and the sending unit 1702 is further configured to send the trained first reconstruction model to the second device.

[0264] In a possible implementation, the plurality of groups of model input data are obtained based on different position selection rates, where the position selection rate is used to indicate a proportion of a number of selected positions in the data information to a total number of positions in the data information.

[0265] The specific operations of each unit in the data processing apparatus 1700 described above can be referred to the descriptions of the data processing apparatus 1700 and the possible implementations of the first device, which will not be repeated here. Figure 4 、 Figure 9 、 Figure 11 、 Figure 12 、 Figure 13 The specific operations of each unit in the data processing apparatus 1700 described above can be referred to the descriptions of the data processing apparatus 1700 and the possible implementations of the first device, which will not be repeated here.

[0266] Figure 18 FIG. 18 shows a possible structural schematic diagram of a data processing apparatus 1800. The data processing apparatus 1800 includes a receiving unit 1801 and a processing unit 1802. In a possible design, the data processing apparatus 1800 can be the second device described above, or can be a chip in the second device, or can be a processing system in the second device, and the like. Wherein:

[0267] The receiving unit 1801 is configured to receive K second data and position information of M target positions from a first device.

[0268] The processing unit 1802 is configured to reconstruct data information corresponding to the K second data and the position information of the M target positions based on a first reconstruction model, to obtain second data information, where the second data information includes N first data; N is an integer greater than 1, and M and K are integers less than N and greater than 0.

[0269] In a possible implementation, the first data information is two-dimensional data information or three-dimensional data information.

[0270] In a possible implementation, the first data information is a radio frequency map.

[0271] In a possible implementation, before the processing unit 1802 reconstructs the second data information based on the first reconstruction model and the data information corresponding to the K second data and the position information of the M target positions, the receiving unit 1801 further receives first information from the first device, where the first information includes identification information of the first position selection manner and / or identification information of the first data extraction manner; and the processing unit 1802 further determines the first reconstruction model corresponding to the first position selection manner and / or the first data extraction manner; different position selection manners and / or data extraction manners correspond to different reconstruction models.

[0272] In a possible implementation, the first position selection manner is one of random selection, topology selection, grouping selection, and iteration selection, and the first data extraction manner is original data extraction or weighted data extraction.

[0273] In a possible implementation, the receiving unit 1801 further receives H second data and position information of J target positions from the first device, where the J target positions are different from the target positions that have been received, and H and J are integers less than N and greater than 0; and the processing unit 1802 further reconstructs third data information based on the second data and the position information of the target positions that have been received, and the H second data and the position information of the J target positions, where the third data information includes N first data.

[0274] In a possible implementation, before the processing unit 1802 reconstructs the second data information based on the first reconstruction model and the data information corresponding to the K second data and the position information of the M target positions, the receiving unit 1801 further receives the trained first reconstruction model from the first device.

[0275] In a possible implementation, before the processing unit 1802 reconstructs the second data information based on the first reconstruction model and the data information corresponding to the K second data and the position information of the M target positions, the processing unit 1802 further obtains a plurality of data information; the processing unit 1802 further processes the plurality of data information based on the first position selection manner and the first data extraction manner, to obtain a plurality of groups of model input data; and the processing unit 1802 further trains the initial first reconstruction model based on the plurality of groups of model input data and the plurality of data information, to obtain the trained first reconstruction model.

[0276] In a possible implementation, the plurality of groups of model input data are obtained based on different position selection rates, where the position selection rate is used to indicate a proportion of a number of selected positions to a total number of positions in data information.

[0277] The specific operations of each unit in the data processing apparatus 1800 described above can be seen from the description of the second device in the above data processing apparatus 1800 and possible embodiments thereof, and will not be repeated here. Figure 4 、 Figure 9 、 Figure 11 、 Figure 12 、 Figure 13 The specific operations of each unit in the data processing apparatus 1800 described above can be seen from the description of the second device in the above data processing apparatus 1800 and possible embodiments thereof, and will not be repeated here.

[0278] In one possible implementation, Figure 17 and Figure 18 In the data processing apparatus shown in FIG. 17 and FIG. 18, the sending unit can be a transmitter, and the receiving unit can be a receiver. The sending unit and the receiving unit can be integrated into one device, such as a transceiver. In the embodiments of the present application, the processor and the transceiver can be coupled, and the manner in which the processor and the transceiver are connected is not limited in the embodiments of the present application. For example, the data processing apparatus can further include a processing unit, which can be one or more processors / logic circuits. In the process of executing the above method, the process of sending information (such as sending K second data and position information of M target positions) in the above method can be understood as the process of outputting the above information by the processor. When the above information is output, the processor can output the above information to the transceiver so that the transceiver transmits. After the above information is output by the processor, it can also need to be processed further before reaching the transceiver. Similarly, the process of receiving information (such as receiving K second data and position information of M target positions) in the above method can be understood as the process of receiving input by the processor. When the processor receives the input information, the transceiver receives the above information and inputs it to the processor. Furthermore, after the transceiver receives the above information, the above information can need to be processed further before being input to the processor.

[0279] In another possible implementation, Figure 17 and Figure 18 In the data processing apparatus shown in FIG. 17 and FIG. 18, the sending unit can be a transmitter, and the receiving unit can be a receiver. The sending unit and the receiving unit can be integrated into one device, such as a transceiver. In the embodiments of the present application, the processor and the transceiver can be coupled, and the manner in which the processor and the transceiver are connected is not limited in the embodiments of the present application. For example, the data processing apparatus can further include a processing unit, which can be one or more processors / logic circuits. In the process of executing the above method, the process of sending information (such as sending K second data and position information of M target positions) in the above method can be understood as the process of outputting the above information by the processor. When the above information is output, the processor can output the above information to the transceiver so that the transceiver transmits. After the above information is output by the processor, it can also need to be processed further before reaching the transceiver. Similarly, the process of receiving information (such as receiving K second data and position information of M target positions) in the above method can be understood as the process of receiving input by the processor. When the processor receives the input information, the transceiver receives the above information and inputs it to the processor. Furthermore, after the transceiver receives the above information, the above information can need to be processed further before being input to the processor.

[0280] Figure 19 FIG. 19 shows a possible hardware structure of the data processing apparatus 1900 provided by the embodiments of the present application. The data processing apparatus 1900 can include a processor 1901 and a transceiver 1902. It should be noted that Figure 19 Only the main components of the data processing apparatus 1900 are shown, and the data processing apparatus 1900 can further include a memory 1903, an input / output device (not shown in the figure), and the like.

[0281] The processor 1901 is mainly used for processing communication protocol and communication data, and controlling the whole data processing device, executing software program, and processing data of the software program. The memory 1903 is mainly used for storing software program and data. The transceiver 1902 can include control circuit and antenna, and the control circuit is mainly used for converting baseband signal and radio frequency signal, and processing radio frequency signal. The antenna is mainly used for receiving and sending radio frequency signal in the form of electromagnetic wave. The input and output device, such as touch screen, display screen, keyboard, etc. is mainly used for receiving user input data and outputting data to user.

[0282] When the data processing device is powered on, the processor 1901 can read the software program in the memory 1903, interpret and execute the instruction of the software program, and process the data of the software program. When data needs to be sent wirelessly, the processor 1901 can output baseband signal to the control circuit after baseband processing the data to be sent, and the control circuit can convert the baseband signal to radio frequency signal and send the radio frequency signal to the outside in the form of electromagnetic wave through the antenna. When data is sent to the data processing device, the control circuit receives the radio frequency signal through the antenna, converts the radio frequency signal to baseband signal, and outputs the baseband signal to the processor 1901, and the processor 1901 converts the baseband signal to data and processes the data.

[0283] In a possible implementation, the control circuit and the antenna can be arranged independently of the processor for baseband processing, for example, in a distributed scenario, the control circuit and the antenna can be arranged remotely from the data processing device.

[0284] The processor 1901, the transceiver 1902, and the memory 1903 can be connected through a communication bus.

[0285] For example, the memory 1903 can include, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read only memory (EPROM) or compact disc read-only memory (CD-ROM), etc.

[0286] The processor 1901 can be, for example, a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit, a field programmable gate array (FPGA) or other programmable logic device, transistor logic device, hardware component, or any combination thereof. The processor can also be a combination of computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0287] In one design, the data processing apparatus 1900 can be configured to perform the functions of the first device in the above-described embodiments. For example, the data processing apparatus 1900 can be configured to perform the operations of the first device in any of the above-described method embodiments. Figure 4 Figure 9 Figure 11 Figure 12 Figure 13 and possible implementations thereof, which will not be repeated here.

[0288] In another design, the data processing apparatus 1900 can be configured to perform the functions of the second device in the above-described embodiments. For example, the data processing apparatus 1900 can be configured to perform the operations of the second device in any of the above-described method embodiments. Figure 4 Figure 9 Figure 11 Figure 12 Figure 13 and possible implementations thereof, which will not be repeated here.

[0289] In one possible design, the processor 1901 can store instructions, which can be a computer program, running on the processor 1901, to cause the data processing apparatus 1900 to perform the operations of the first device or the second device in any of the above-described method embodiments. For example, the data processing apparatus 1900 can be configured to perform the operations of the first device or the second device in any of the above-described method embodiments. Figure 4 Figure 9 Figure 11 Figure 12 Figure 13 and possible implementations thereof, which will not be repeated here.

[0290] It should be noted that the data processing apparatus 1900 shown in the figure is only one implementation of the embodiments of the present application, and in actual applications, the data processing apparatus 1900 can also include more or fewer components, which are not limited here. Figure 19

[0291] The embodiments of the present application also disclose a communication system, which includes a first device configured to perform the operations of the first device in any of the above-described method embodiments, and a second device configured to perform the operations of the second device in any of the above-described method embodiments.

[0292] ​​​​​​​​​​​​​The chip includes a processor, where the processor is configured to execute computer programs or computer instructions stored in a memory, so that the chip executes operations performed by the first device in the method embodiments, or so that the chip executes operations performed by the second device in the method embodiments.

[0293] As one possible implementation, the memory is located outside the chip.

[0294] The embodiments of the present application also disclose a computer readable storage medium, which has instructions stored thereon, the instructions being executed to perform operations performed by the first device in the method embodiments or operations performed by the second device in the method embodiments.

[0295] The embodiments of the present application also disclose a computer program product including instructions, the instructions being executed to perform operations performed by the first device in the method embodiments or operations performed by the second device in the method embodiments.

[0296] Obviously, the above-described embodiments are only some of the embodiments of the present application, but not all the embodiments. In this document, the term "embodiments" means that the specific features, structures or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The occurrence of the phrase "in an embodiment" in various places in the specification does not necessarily mean that the same embodiment is being referred to, nor does it mean that the same embodiment is being referred to in combination with other embodiments. It will be apparent to those skilled in the art from this description that the embodiments described herein can be combined with other embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application. The terms "include" and "have" and any variations thereof in the present application are intended to cover non-exclusive inclusion. For example, a series of steps or units are included, or optionally, other steps or units not listed are also included, or optionally, other steps or units inherent to the processes, methods, products or equipment are also included. It can be understood that in some embodiments, the equal sign in the above condition judgment can be greater than one end or less than one end, for example, the condition judgment of greater than, less than or equal to a threshold value can be changed to greater than or equal to, less than the condition judgment of the threshold value, which is not limited here. It can also be understood that for an architecture with multiple devices or modules, if one device or module generates an information and another device or module utilizes the information, there can be multiple ways for the other device to obtain the information, for example, the device or module generating the information can directly send the information to the device or module utilizing the information (equivalent to direct sending), or the device or module generating the information can send the information to the device or module utilizing the information through other devices or modules (equivalent to indirect sending).

[0297] It is to be understood that the figures only show parts of the application and not everything. It is to be understood that some example embodiments are described as processes or methods depicted as flow diagrams. Although the flow diagrams describe the operations (or steps) as a sequential process, many of the operations can be performed in parallel, concurrently or simultaneously. In addition, the order of the operations can be re-arranged. A process can be terminated when its operations are completed, but could also have additional steps not included in the figure. A process can correspond to a method, a function, a procedure, a subroutine, a subprogram, etc.

[0298] The terms "component," "module," "system," "unit," and the like are used in this specification to refer to computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a unit can be, but is not limited to, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and / or a computer. Further, these units can execute from various computer-readable media having various data structures stored thereon. The units can communicate over a network via signals, which can include one or more data packets (e.g., from a second unit and transmitted between other units in a distributed system). For example, the units can communicate via signals transmitted between or among other systems via the Internet.

[0299] The above detailed description has further explained the purpose, technical solutions and advantages of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made on the basis of the technical solutions of the present application shall be included in the protection scope of the present application.

Claims

1. A data processing method, characterized by, The method applied to a first device comprises: obtaining first data information, the first data information comprising N first data, the N first data corresponding to N positions, N being an integer greater than 1; selecting M target positions from the N positions, M being an integer less than N and greater than 0; obtaining K second data based on the M target positions, K being an integer less than N and greater than 0; sending the K second data and position information of the M target positions to a second device.

2. The method of claim 1, wherein, The first data information is two-dimensional data information or three-dimensional data information.

3. The method according to claim 1 or 2, characterized in that, The first data information is a radio frequency map.

4. The method according to any one of claims 1 to 3, characterized in that, The selecting of the M target positions from the N positions comprises: selecting the M target positions from the N positions based on a first position selection manner, the first position selection manner being one of random selection, topology selection, grouping selection, and iteration selection.

5. The method of claim 4, wherein, In a case where the first position selection manner is the grouping selection, the selecting of the M target positions from the N positions based on the first position selection manner comprises: dividing the N positions into L groups of positions based on the N first data, L being an integer greater than 1; selecting at least one position from each of the L groups of positions to obtain the M target positions.

6. The method according to claim 4 or 5, characterized in that, In a case where the first position selection manner is the topology selection, the selecting of the M target positions from the N positions based on the first position selection manner comprises: selecting the M target positions from the N positions based on a topology structure of the N first data.

7. The method according to any one of claims 1 to 6, characterized in that, The obtaining of the K second data based on the M target positions comprises: obtaining the K second data based on a first data extraction manner and the M target positions, the first data extraction manner being original data extraction or weighted data extraction.

8. The method of claim 7, wherein, In a case where the first data extraction manner is the weighted data extraction, the obtaining of the K second data based on the first data extraction manner and the M target positions comprises: performing weighted calculation on first data corresponding to the M target positions and all or part of positions having a distance less than a first threshold from the M target positions to obtain the K second data.

9. The method according to claim 7 or 8, characterized in that, The method further comprises: sending first information to the second device, the first information comprising identification information of the first position selection manner and / or identification information of the first data extraction manner.

10. The method according to any one of claims 1 to 9, characterized in that, The method further comprises: selecting J target positions from the N positions, the J target positions being different from the target positions that have been sent, J being an integer less than N and greater than 0; obtaining H second data based on the J target positions, H being an integer less than N and greater than 0; sending the H second data and position information of the J target positions to a second device.

11. The method according to any one of claims 1 to 10, characterized in that, The method further comprises: obtaining a plurality of data information; processing the plurality of data information based on the first position selection manner and the first data extraction manner to obtain a plurality of groups of model input data; training an initial first reconstruction model based on the plurality of groups of model input data and the plurality of data information to obtain a trained first reconstruction model; sending the trained first reconstruction model to the second device.

12. The method of claim 11, wherein, The multiple sets of model input data are obtained based on different position selection rates, and the position selection rate is used to indicate a proportion of a number of selected positions to a total number of positions in data information.

13. A data processing method, characterized by, The method applied to the second device comprises: receiving K second data and position information of M target positions from the first device; reconstructing data information corresponding to the K second data and the position information of the M target positions based on a first reconstruction model to obtain second data information, wherein the second data information comprises N first data; N is an integer greater than 1, and M and K are integers less than N and greater than 0.

14. The method of claim 13, wherein, The first data information is two-dimensional data information or three-dimensional data information.

15. The method according to claim 13 or 14, characterized in that, The first data information is a radio frequency map.

16. The method according to claim 14 or 15, characterized in that Before the reconstruction based on the first reconstruction model, the method further comprises: receiving first information from the first device, wherein the first information comprises identification information of a first position selection mode and / or identification information of a first data extraction mode; determining the first reconstruction model corresponding to the first position selection mode and / or the first data extraction mode; wherein different position selection modes and / or data extraction modes correspond to different reconstruction models.

17. The method of claim 16, wherein, The first position selection mode is one of random selection, topological selection, grouping selection, and iteration selection, and the first data extraction mode is original data extraction or weighted data extraction.

18. The method according to any one of claims 14-17, characterized by, The method further comprises: receiving H second data and position information of J target positions from the first device, wherein the J target positions are different from the target positions that have been received, and H and J are integers less than N and greater than 0; reconstructing third data information based on the second data and the position information of the target positions that have been received and the H second data and the position information of the J target positions, wherein the third data information comprises N first data.

19. The method according to any one of claims 14-18, characterized by, Before the reconstruction based on the first reconstruction model, the method further comprises: receiving the trained first reconstruction model from the first device.

20. The method according to any one of claims 14-19, characterized by, Before the reconstruction based on the first reconstruction model, the method further comprises: obtaining multiple data information; processing the multiple data information based on the first position selection mode and the first data extraction mode respectively to obtain multiple sets of model input data; training an initial first reconstruction model based on the multiple sets of model input data and the multiple data information to obtain the trained first reconstruction model.

21. The method of claim 20, wherein, The multiple sets of model input data are obtained based on different position selection rates, and the position selection rate is used to indicate a proportion of a number of selected positions to a total number of positions in data information.

22. A data processing apparatus, characterized in that, A computer program product comprising computer program code or computer instructions, which when run by a processor, implements the method of any one of claims 1-12, or the method of any one of claims 13-21.

23. A communication system, characterized by A computer program product comprising computer program code or computer instructions, which when run by a processor, implements the method of any one of claims 1-12, or the method of any one of claims 13-21.

24. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer program or computer instructions, which are executed by the processor to implement the method of any one of claims 1-12, or the method of any one of claims 13-21.

25. A computer program product, characterised in that, The computer program product comprises computer program code or computer instructions, which when run by a processor, implements the method of any one of claims 1-12, or the method of any one of claims 13-21.

Images