UAV air signal strength prediction method and device based on neural network

Abstract

The present invention provides a method and device for predicting the air signal strength of an unmanned aerial vehicle (UAV) based on a neural network. The method comprises: obtaining an air signal test data packet at a preset altitude under current weather conditions; inputting the air signal test data packet into a signal prediction model, and outputting a signal strength prediction result for a target weather condition and a target altitude corresponding to the air signal test data packet; wherein the signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and different preset altitudes. When conducting a test, it is only necessary to fly at one altitude and one weather condition to output air coverage data sequences corresponding to other altitudes and other weather conditions, thereby reducing the workload of air signal measurement and accurately reflecting the signal strength.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method and device for predicting the air signal strength of an unmanned aerial vehicle based on a neural network. Background Art

[0002] Existing methods for acquiring aerial signal strength primarily rely on drones conducting aerial signal coverage tests to collect data. To obtain three-dimensional aerial signal strength data, drones must be assigned flight paths at different altitudes to collect data at various levels. Existing drone aerial signal coverage testing schemes often target 100-meter, 150-meter, and 200-meter horizontal planes to reflect the distribution of aerial signal strength in the area.

[0003] Signal strength in the air is complex and highly variable, with weather being a significant factor. Clear weather conditions are characterized by dry air, which facilitates wireless signal transmission. However, rainy and snowy weather, with high humidity, reduces wireless signal transmission rates and transmission range. Extreme weather, such as heavy rain, can further impact wireless network signal propagation due to changes in atmospheric waveguide. Therefore, in actual airborne signal testing, it's important to consider the impact of weather conditions on airborne signals. However, existing testing methods, to simplify the problem and ensure drone flight safety, don't account for weather effects during airborne coverage testing.

[0004] Existing air signal coverage testing solutions require three air signal tests at different altitudes within the same area. The main wireless coverage metrics measured include RSRP, SINR, UL-speed, and TA. Consequently, existing solutions suffer from a high workload and test results that fail to accurately reflect the signal strength in the area.

[0005] Therefore, how to provide an air signal strength prediction solution that can reduce the workload of air signal measurement and accurately reflect the signal strength is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0006] The present invention provides a method and device for predicting the air signal strength of an unmanned aerial vehicle (UAV) based on a neural network, which can reduce the workload of air signal measurement and accurately reflect the signal strength.

[0007] In a first aspect, the present invention provides a method for predicting the air signal strength of a UAV based on a neural network, comprising:

[0008] Obtain air signal test data packets at a preset altitude under current weather conditions;

[0009] Inputting the air signal test data packet into a signal prediction model, and outputting a signal strength prediction result of the target weather conditions and target altitude corresponding to the air signal test data packet;

[0010] The signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and at different preset altitudes.

[0011] Furthermore, the step of obtaining an air signal test data packet at a preset altitude under current weather conditions includes:

[0012] Controlling a drone to take off to a preset altitude in a target test area, wherein the drone is provided with a signal testing device;

[0013] Controlling the UAV to fly along a preset trajectory in the target test area;

[0014] When the UAV flies according to a preset trajectory in the target test area, the signal test device is controlled to obtain an air signal test data packet of a target sampling point in the target test area.

[0015] Furthermore, inputting the air signal test data packet into the signal prediction model and outputting the signal strength prediction result of the target weather condition and target altitude corresponding to the air signal test data packet includes:

[0016] Inputting the air signal test data packet into the first LSTM network of the signal prediction model to obtain the air signal feature information;

[0017] Inputting the air signal characteristic information into the second LSTM network of the signal prediction model to obtain a prediction result based on the cell information and weather information;

[0018] The cell information includes: access cell configuration information, access cell location information, nearest neighbor cell location information, and next nearest neighbor cell location information.

[0019] Furthermore, it also includes:

[0020] Receiving a first local attention input from a user to the first LSTM network;

[0021] In response to the first local attention input, a weight value of data in the signal-over-the-air test data packet is adjusted.

[0022] Furthermore, inputting the air signal characteristic information into the second LSTM network of the signal prediction model to obtain a prediction result based on the cell information and weather information includes:

[0023] Combine the air signal characteristic information with the corresponding cell information and weather information to obtain a combined data set;

[0024] The combined data set is input into the second LSTM network, and a signal strength prediction result of the target weather condition and target altitude corresponding to the combined data set is output.

[0025] Furthermore, it also includes:

[0026] receiving a second local attention input from the user to the second LSTM network;

[0027] In response to the second local attention input, weight values of the data in the combined dataset are adjusted.

[0028] Furthermore, the signal prediction model uses a loss function to optimize the network:

[0029] When the path to be predicted is a unidirectional line segment with a long distance, L1 is used as the loss function;

[0030] When the path to be predicted is a round trip within the measured range, L2 is used as the loss function;

[0031] in,

[0032]

[0033]

[0034] where y i Indicates the signal strength data actually obtained at the i-th data point, represents the signal strength of the neural network prediction value at the i-th point, and n represents the total number of data points.

[0035] In a second aspect, the present invention provides a device for predicting the air signal strength of a UAV based on a neural network, comprising:

[0036] A data acquisition module is used to obtain air signal test data packets at a preset altitude under current weather conditions;

[0037] a strength prediction module, configured to input the air signal test data packet into a signal prediction model and output a signal strength prediction result of a target weather condition and a target altitude corresponding to the air signal test data packet;

[0038] The signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and at different preset altitudes.

[0039] Furthermore, the data acquisition module includes:

[0040] A UAV control unit, configured to control the takeoff of the UAV to a preset altitude within the target test area, wherein the UAV is provided with a signal testing device;

[0041] A trajectory control unit, used to control the UAV to fly along a preset trajectory in the target test area;

[0042] The data sampling unit is used to control the signal testing device to obtain an air signal test data packet of a target sampling point in the target test area when the UAV flies according to a preset trajectory in the target test area.

[0043] Furthermore, the intensity prediction module includes:

[0044] A first LSTM submodule is configured to input the air signal test data packet into a first LSTM network of a signal prediction model to obtain the air signal feature information;

[0045] A second LSTM submodule is configured to input the air signal characteristic information into a second LSTM network of the signal prediction model, and obtain a prediction result based on the cell information and weather information;

[0046] The cell information includes: access cell configuration information, access cell location information, nearest neighbor cell location information, and next nearest neighbor cell location information.

[0047] Furthermore, it also includes:

[0048] A first receiving module is configured to receive a first local attention input of a user to the first LSTM network;

[0049] The first response module is configured to adjust a weight value of data in the air signal test data packet in response to the first local attention input.

[0050] Furthermore, the second LSTM submodule includes:

[0051] a data collection unit for combining the air signal characteristic information with the corresponding cell information and weather information to obtain a combined data set;

[0052] The prediction result unit is used to input the combined data set into the second LSTM network and output the signal strength prediction result of the target weather condition and target altitude corresponding to the combined data set.

[0053] Furthermore, it also includes:

[0054] A second receiving module is configured to receive a second local attention input of the user to the second LSTM network;

[0055] A second response module is used to adjust the weight value of the data in the combined data set in response to the second local attention input.

[0056] Furthermore, the signal prediction model uses a loss function to optimize the network:

[0057] The device further includes: a function selection module for adopting L1 as a loss function when the path to be predicted is a line segment of a long distance in one direction;

[0058] When the path to be predicted is a round trip within the measured range, L2 is used as the loss function;

[0059] in,

[0060]

[0061]

[0062] where y i Indicates the signal strength data actually obtained at the i-th data point, represents the signal strength of the neural network prediction value at the i-th point, and n represents the total number of data points.

[0063] In a third aspect, the present invention provides an electronic device comprising a memory and a memory storing a computer program, wherein when the processor executes the program, the steps of the neural network-based drone air signal strength prediction method described in the first aspect are implemented.

[0064] In a fourth aspect, the present invention provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, and the computer program is used to enable the processor to execute the steps of the neural network-based drone air signal strength prediction method described in the first aspect.

[0065] The present invention provides a neural network-based UAV air signal strength prediction method and device, which predicts signal strength through a signal prediction model and can output signal strength prediction results for target weather conditions and target altitudes corresponding to the air signal test data packet. When conducting a test, it is only necessary to fly at one altitude and one weather condition to output air coverage data sequences corresponding to other altitudes and other weather conditions, thereby reducing the workload of air signal measurement and accurately reflecting signal strength. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0067] Figure 1 One of the flow charts of a method for predicting the air signal strength of a UAV based on a neural network provided by an embodiment of the present invention;

[0068] Figure 2 A second flow chart of a method for predicting the air signal strength of a UAV based on a neural network according to an embodiment of the present invention;

[0069] Figure 3 A third flow chart of a method for predicting the air signal strength of a UAV based on a neural network provided by an embodiment of the present invention;

[0070] Figure 4 A top view of the flight path trajectory of a UAV performing an aerial coverage test according to an embodiment of the present invention;

[0071] Figure 5 A data processing flow chart for signal strength prediction provided by an embodiment of the present invention;

[0072] Figure 6 A practical flow chart of signal strength prediction provided by an embodiment of the present invention;

[0073] Figure 7 A schematic diagram of the structure of a UAV air signal strength prediction device based on a neural network provided by the present invention;

[0074] Figure 8 It is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0075] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0076] The following combination Figure 1-Figure 3 The present invention describes a method for predicting the air signal strength of a UAV based on a neural network.

[0077] Figure 1One of the flow charts of a method for predicting the air signal strength of a UAV based on a neural network provided by an embodiment of the present invention; Figure 2 A second flow chart of a method for predicting the air signal strength of a UAV based on a neural network according to an embodiment of the present invention; Figure 3 The third flow chart of a method for predicting the air signal strength of a UAV based on a neural network provided in an embodiment of the present invention.

[0078] In a specific embodiment of the present invention, an embodiment of the present invention provides a method for predicting the air signal strength of a UAV based on a neural network, comprising:

[0079] Step 110: Obtain an air signal test data packet at a preset altitude under current weather conditions;

[0080] In an embodiment of the present invention, it is first necessary to obtain an air signal test data packet at a preset altitude under current weather conditions. Specifically, this can be obtained by using a drone carrying a measurement device. Of course, other methods can also be used to obtain the data packet. For example, a long pole can be used at a preset altitude of several meters to more than ten meters, and the test device can be attached to the top of the pole to obtain the air signal test data packet. The data packet mainly includes but is not limited to: number, longitude and latitude, RSRP, SINR, DL-Speed, UL-Speed, TA, and other information.

[0081] Step 120: Inputting the air signal test data packet into a signal prediction model, and outputting a signal strength prediction result of the target weather conditions and target altitude corresponding to the air signal test data packet;

[0082] The signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and at different preset altitudes.

[0083] After acquiring airborne signal test data packets, these packets can be input into a pre-trained signal prediction model to output signal strength predictions for the target altitude and target weather conditions. The signal prediction model is a neural network trained using training samples. It considers the influence of weather factors during the testing process, and the test results accurately reflect the signal strength of the area under different weather conditions. Furthermore, there is no need to measure the airborne signal strength at each altitude; the pre-trained prediction model can be used to predict the signal strength at each altitude.

[0084] Please refer to Figure 4 , Figure 4 A top view of the flight path trajectory of a drone performing an aerial coverage test according to an embodiment of the present invention.

[0085] Specifically, to obtain an air signal test data packet at a preset altitude under current weather conditions, the following steps can be performed:

[0086] Step 210: Controlling a drone to take off to a preset altitude of a target test area, wherein the drone is provided with a signal testing device;

[0087] Step 220: Control the UAV to fly along a preset trajectory in the target test area;

[0088] Step 230: When the UAV flies along a preset trajectory in the target test area, the signal test device is controlled to obtain an air signal test data packet of a target sampling point in the target test area.

[0089] The drone collects communication network coverage data in clear weather at a certain intermediate altitude (usually 150 meters) and predicts communication network coverage data in different weather conditions at a certain altitude above and below that altitude (usually 50 meters).

[0090] The top view of the flight path of the drone during the aerial coverage test is as follows: Figure 4 As shown in the figure, during the flight, the test mobile phone carried by the aircraft will collect data packets of network coverage signals at each location.

[0091] Each point represents a data packet, and there will be a corresponding service cell. We use the following D 11 to D mn express:

[0092] D 11 、D 12 、D 13 、D 21 、D 22 、D 23 、D 31 、D 23 …D mn

[0093] The entire sequence contains m*n data packets. m represents the cell number corresponding to the data packet, and n represents the nth sampling point under the cell. For example, D 24 Represents the fourth sample data packet when occupying the second cell.

[0094] Each data packet mainly includes but is not limited to: number, longitude and latitude, RSRP, SINR, DL-Speed, UL-speed, TA and other information. Through a two-layer LSTM neural network model, the DL obtained at a certain height is 11 to D mnThe data sequence is input into the neural network, and the predicted sequence is output. The reason for using a two-layer LSTM is that the data collected by the drone during flight is a sequence, and the service cells occupied by the drone along the flight path are also a sequence.

[0095] Please refer to Figure 5 , Figure 5 This is a data processing flow chart for signal strength prediction provided by an embodiment of the present invention.

[0096] In another specific embodiment of the present invention, in order to input the air signal test data packet into the signal prediction model and output the signal strength prediction result of the target weather conditions and target altitude corresponding to the air signal test data packet, the following steps can be specifically performed:

[0097] Step 310: Inputting the air signal test data packet into the first LSTM network (LSTM1) of the signal prediction model to obtain the air signal feature information;

[0098] Step 320: Input the air signal characteristic information into the second LSTM network (LSTM2) of the signal prediction model, and obtain a prediction result based on the cell information and weather information; wherein the cell information includes: access cell configuration information, access cell location information, nearest neighbor cell location information, and next nearest neighbor cell location information.

[0099] In the above figure, C1 to C m Information about the access cell, including frequency, antenna azimuth angle, antenna downtilt angle, transmit power, antenna type, etc.

[0100] L1 to L in the above figure m In order to obtain the antenna longitude and latitude information of each service cell, the drone will access L1 to L m Community.

[0101] N1 to N in the above figure m To serve the antenna longitude and latitude information of the nearest neighboring cell of each service cell, generally speaking, the drone will access the nearest neighboring cell when flying in the air, so L m Yu N m There will be a certain relationship between the sequences.

[0102] R1 to R in the above figure m The antenna longitude and latitude information of the secondary neighboring cells of each serving cell is obtained. Actual tests have found that in areas with sparsely populated sites, the information of secondary neighboring cells can be ignored. Only when the inter-site distance is less than 500m will more handovers to secondary neighboring cells occur.

[0103] W1 to W in the above picture mThe weather conditions during the flight of the UAV, including temperature, air pressure, atmospheric waveguide and other factors, are usually the same in the data sampled in successive flights. However, considering that in some cases the long-distance flight of the UAV may involve changes in environmental factors, this variable is introduced.

[0104] Due to its predictive power and ability to alleviate the vanishing gradient problem in long-distance dependent sequence problems, LSTM is used to handle this task. In our framework, we refer to the output of this layer of LSTM as O 11 to O 11 , and then use these outputs to apply a local attention mechanism1 to select the focus of the input information, such as whether the weight is more biased towards the SINR value or RSRP value in the data packet.

[0105] Again, the LSTM layer and attention mechanism are used to embed the output of the previous LSTM network into data C1 to C m , L1 to L m , and W1 to W m These are then fed into another LSTM layer. The second LSTM layer features a characteristic: considering the continuity of drone flight, the neural network connects neurons to the previously occupied cell. Ultimately, the outputs d1, d2, …, dm are obtained, resulting in the final output. This improves the interpretability of the model's results.

[0106] The input data consists of a sequence of actual test results at 150 meters. Each input is combined with the wireless network configuration parameters at that point, as well as key information influencing the 100- or 200-meter data, such as the access station's nearest neighbor and next-nearest neighbor. This information is then fed into a neural network, where an iterative algorithm optimizes network parameters to generate predicted data. The predicted data is divided into six groups: signal strength data at 100 meters for clear, overcast, and rainy weather conditions, and signal strength data at 200 meters for clear, overcast, and rainy weather conditions.

[0107] The first part of the neural network utilizes an LSTM structure. Its purpose is to sequentially input information from a test sequence into the neural network, extracting and integrating information from each feature. Because the input sequences exhibit certain correlations, an LSTM is also designed in the second part of the network to extract relevant information. The first LSTM network primarily addresses the continuity of the input signal data, while the second LSTM network focuses on the continuity of the access cell. It combines airborne signal feature information with corresponding cell and weather information to generate a combined dataset. This combined dataset is then fed into the second LSTM network, which outputs predictions for target weather conditions and signal strength at the target altitude corresponding to the combined dataset.

[0108] LSTM is a special type of recurrent neural network that integrates current input with previous input and identifies correlations between them, thereby improving the robustness of the neural network. Both layers of the neural network use the ReLU activation function as a nonlinear mapping function, employ the Dropout technique to prevent overfitting, and use the Adam algorithm to update the parameters of each neuron.

[0109] During the training phase, drones are used to collect signal strength data at three heights within a certain area, usually with a center height of 100 meters and intervals of 50 meters above and below (100 meters, 150 meters, and 200 meters), as well as data under different weather conditions (sunny, cloudy and rainy, and rainstorms) at corresponding heights. These data are divided into training and test sets, and the supervised learning method is used to train the model.

[0110] It should be noted that the user can also adjust the specific data that the first or second LSTM network focuses on. Specifically, the user can receive a first local attention input to the first LSTM network; in response to the first local attention input, the weight value of the data in the air signal test data packet can be adjusted. The user can also receive a second local attention input to the second LSTM network; in response to the second local attention input, the weight value of the data in the combined data set can be adjusted.

[0111] Based on the above embodiment, in this embodiment, the signal prediction model uses a loss function to optimize the network: when the path to be predicted is a one-way long-distance line segment, L1 is used as the loss function; when the path to be predicted is a round trip within the range to be measured, L2 is used as the loss function;

[0112] in,

[0113]

[0114]

[0115] where y i Indicates the signal strength data actually obtained at the i-th data point, represents the signal strength of the neural network prediction value at the i-th point, and n represents the total number of data points.

[0116] During the training phase, drones are used to collect signal strength data at three altitudes within a certain area, typically 100 meters in the center and 50 meters apart (100, 150, and 200 meters). The data is then divided into training and test sets, and the model is trained using supervised learning. The initialization parameters are shown in Table 1:

[0117] Table 1

[0118]

[0119] Set the loss function L1. Use the mean square error loss function to optimize the network. In the actual use of neural networks, we found that when the UAV's flight route is densely packed with round trips, the mean absolute error function can be used as the loss function. If the path to be predicted is a straight line with a long distance in one direction, then L1 is more suitable as the loss function. If the path to be predicted is a network that goes back and forth within a certain range, such as Figure 4 As shown in the figure, it is more appropriate to use L2 as the loss function. When L2 is selected as the loss function, the Dropout value can be appropriately reduced to 45%.

[0120] During the testing phase, for a specific area, the drone only needs to complete an aerial signal coverage test mission at an intermediate altitude (e.g., 150 meters, though other altitudes are possible) under clear weather conditions to use the data for neural network training. Preliminary flight sampling is then performed at two other test altitudes to obtain initial values for the two altitude sequences to be predicted, along with other parameters. These are then fed into the neural network to predict subsequent aerial coverage data.

[0121] like Figure 6 As shown, Figure 6 A practical flow chart of signal strength prediction provided by an embodiment of the present invention.

[0122] In an embodiment of the present invention, a test mobile phone is first mounted on a drone, and aerial coverage data is collected at different altitudes and in different weather conditions according to a set route (1), and the data is cleaned (2). For example, a total of 9 sets of data can be obtained at three altitudes and three weather conditions on the same route. These nine sets of data are sequentially and separately input into multiple neural networks for training. The 9 sets of data contain information about altitude and weather, so the input order has little effect on the neural network (3). This step involves the training of multiple neural networks. For example, when training neural network A, the input is RSRP data at an altitude of 150 meters on a route, and the training sample is RSRP data on the same line at an altitude of 200 meters. Then, the trained neural network A is used to predict the RSRP value of 150 meters to 200 meters.

[0123] After completing the training of the above multiple neural networks, these neural networks can be used to predict the air coverage of similar coverage areas (4). For a new test area, we can only fly at one altitude and under one weather condition, and input this data sequence into a neural network with a specific function that has been trained previously. Then, the corresponding air coverage data sequences for other altitudes and other weather conditions will be output. (5)

[0124] An embodiment of the present invention provides a neural network-based method for predicting the aerial signal strength of an unmanned aerial vehicle. The method predicts the signal strength through a signal prediction model, and can output the signal strength prediction results of the target weather conditions and target altitude corresponding to the aerial signal test data packet. When conducting a test, it is only necessary to fly at one altitude and one weather condition to output the aerial coverage data sequences corresponding to other altitudes and other weather conditions, thereby reducing the workload of aerial signal measurement and accurately reflecting the signal strength.

[0125] The following describes the UAV air signal strength prediction device based on a neural network provided by the present invention. The UAV air signal strength prediction device based on a neural network described below and the UAV air signal strength prediction method based on a neural network described above can refer to each other.

[0126] Please refer to Figure 7 , Figure 7 This is a schematic diagram of the composition structure of a neural network-based UAV air signal strength prediction device provided by the present invention.

[0127] In another specific embodiment of the present invention, an embodiment of the present invention provides a UAV air signal strength prediction device 700 based on a neural network, comprising:

[0128] The data acquisition module 710 is used to obtain an air signal test data packet at a preset altitude under current weather conditions;

[0129] The strength prediction module 720 is configured to input the air signal test data packet into a signal prediction model and output a signal strength prediction result of the target weather conditions and target altitude corresponding to the air signal test data packet;

[0130] The signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and at different preset altitudes.

[0131] Furthermore, the data acquisition module includes:

[0132] A UAV control unit, configured to control the takeoff of the UAV to a preset altitude within the target test area, wherein the UAV is provided with a signal testing device;

[0133] A trajectory control unit, used to control the UAV to fly along a preset trajectory in the target test area;

[0134] The data sampling unit is used to control the signal testing device to obtain an air signal test data packet of a target sampling point in the target test area when the UAV flies according to a preset trajectory in the target test area.

[0135] Furthermore, the intensity prediction module includes:

[0136] A first LSTM submodule is configured to input the air signal test data packet into a first LSTM network of a signal prediction model to obtain the air signal feature information;

[0137] A second LSTM submodule is configured to input the air signal characteristic information into a second LSTM network of the signal prediction model, and obtain a prediction result based on the cell information and weather information;

[0138] The cell information includes: access cell configuration information, access cell location information, nearest neighbor cell location information, and next nearest neighbor cell location information.

[0139] Furthermore, it also includes:

[0140] A first receiving module is configured to receive a first local attention input of a user to the first LSTM network;

[0141] The first response module is configured to adjust a weight value of data in the air signal test data packet in response to the first local attention input.

[0142] Furthermore, the second LSTM submodule includes:

[0143] a data collection unit for combining the air signal characteristic information with the corresponding cell information and weather information to obtain a combined data set;

[0144] The prediction result unit is used to input the combined data set into the second LSTM network and output the signal strength prediction result of the target weather condition and target altitude corresponding to the combined data set.

[0145] Furthermore, it also includes:

[0146] A second receiving module is configured to receive a second local attention input of the user to the second LSTM network;

[0147] A second response module is used to adjust the weight value of the data in the combined data set in response to the second local attention input.

[0148] Furthermore, the signal prediction model uses a loss function to optimize the network:

[0149] The device further includes: a function selection module for adopting L1 as a loss function when the path to be predicted is a line segment of a long distance in one direction;

[0150] When the path to be predicted is a round trip within the measured range, L2 is used as the loss function;

[0151] in,

[0152]

[0153]

[0154] where y i Indicates the signal strength data actually obtained at the i-th data point, represents the signal strength of the neural network prediction value at the i-th point, and n represents the total number of data points.

[0155] An embodiment of the present invention provides a neural network-based UAV air signal strength prediction device, which predicts signal strength through a signal prediction model and can output signal strength prediction results for target weather conditions and target altitudes corresponding to the air signal test data packet. When conducting a test, it is only necessary to fly at one altitude and one weather condition to output air coverage data sequences corresponding to other altitudes and other weather conditions, thereby reducing the workload of air signal measurement and accurately reflecting signal strength.

[0156] Figure 8 An example of a physical structure diagram of an electronic device is shown below. Figure 8 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 may call a computer program in the memory 830 to execute the steps of the neural network-based drone air signal strength prediction method, for example, including:

[0157] Obtain air signal test data packets at a preset altitude under current weather conditions;

[0158] Inputting the air signal test data packet into a signal prediction model, and outputting a signal strength prediction result of the target weather conditions and target altitude corresponding to the air signal test data packet;

[0159] The signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and at different preset altitudes.

[0160] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program code.

[0161] On the other hand, the present invention further provides a computer program product, comprising a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises program instructions. When the program instructions are executed by a computer, the computer is capable of executing the neural network-based UAV air signal strength prediction method provided by the above methods, the method comprising:

[0162] (Get the air signal test data packet at the preset altitude under the current weather conditions;

[0163] Inputting the air signal test data packet into a signal prediction model, and outputting a signal strength prediction result of the target weather conditions and target altitude corresponding to the air signal test data packet;

[0164] The signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and at different preset altitudes.

[0165] On the other hand, an embodiment of the present application further provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, wherein the computer program is configured to cause the processor to execute the methods provided in the above embodiments, for example, including:

[0166] Obtain air signal test data packets at a preset altitude under current weather conditions;

[0167] Inputting the air signal test data packet into a signal prediction model, and outputting a signal strength prediction result of the target weather conditions and target altitude corresponding to the air signal test data packet;

[0168] The signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and at different preset altitudes.

[0169] The processor-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO)), optical storage (such as CDs, DVDs, BDs, HVDs, etc.), and semiconductor storage (such as ROMs, EPROMs, EEPROMs, non-volatile memories (NANDFLASH), solid-state drives (SSDs)), etc.

[0170] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0171] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0172] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for predicting the air signal strength of a UAV based on a neural network, characterized in that: include: Obtain air signal test data packets at a preset altitude under current weather conditions; Inputting the air signal test data packet into a signal prediction model, and outputting a signal strength prediction result of the target weather conditions and target altitude corresponding to the air signal test data packet; The signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and different preset altitudes; The step of inputting the air signal test data packet into a signal prediction model and outputting a signal strength prediction result of a target weather condition and a target altitude corresponding to the air signal test data packet comprises: Inputting the air signal test data packet into the first LSTM network of the signal prediction model to obtain the air signal feature information; The air signal characteristic information is input into the second LSTM network of the signal prediction model, and a prediction result is obtained based on the cell information and weather information.

2. The method for predicting the air signal strength of a UAV based on a neural network according to claim 1, characterized in that: The step of obtaining an air signal test data packet at a preset altitude under current weather conditions includes: Controlling a drone to take off to a preset altitude in a target test area, wherein the drone is provided with a signal testing device; Controlling the UAV to fly along a preset trajectory in the target test area; When the UAV flies according to a preset trajectory in the target test area, the signal test device is controlled to obtain an air signal test data packet of a target sampling point in the target test area.

3. The method for predicting the air signal strength of a UAV based on a neural network according to claim 1, characterized in that: The cell information includes: access cell configuration information, access cell location information, nearest neighbor cell location information, and next nearest neighbor cell location information.

4. The method for predicting the air signal strength of a UAV based on a neural network according to claim 3, characterized in that: Also includes: Receiving a first local attention input from a user to the first LSTM network; In response to the first local attention input, a weight value of data in the signal-over-the-air test data packet is adjusted.

5. The method for predicting the air signal strength of a UAV based on a neural network according to claim 3, characterized in that: Inputting the air signal characteristic information into the second LSTM network of the signal prediction model, and obtaining a prediction result based on the cell information and weather information includes: Combine the air signal characteristic information with the corresponding cell information and weather information to obtain a combined data set; The combined data set is input into the second LSTM network, and a signal strength prediction result of the target weather condition and target altitude corresponding to the combined data set is output.

6. The method for predicting the air signal strength of a UAV based on a neural network according to claim 5, characterized in that: Also includes: receiving a second local attention input from the user to the second LSTM network; In response to the second local attention input, weight values of the data in the combined dataset are adjusted.

7. The method for predicting the air signal strength of a UAV based on a neural network according to any one of claims 1 to 6, characterized in that: The signal prediction model uses a loss function to optimize the network: When the path to be predicted is a unidirectional line segment with a long distance, L1 is used as the loss function; When the path to be predicted is a round trip within the measured range, L2 is used as the loss function; in, where y i Indicates the signal strength data actually obtained at the i-th data point, represents the signal strength of the neural network prediction value at the i-th point, and n represents the total number of data points.

8. A UAV air signal strength prediction device based on neural network, characterized in that: include: A data acquisition module is used to obtain air signal test data packets at a preset altitude under current weather conditions; a strength prediction module, configured to input the air signal test data packet into a signal prediction model and output a signal strength prediction result of a target weather condition and a target altitude corresponding to the air signal test data packet; The signal prediction model is obtained after training based on air signal sample data packets under different weather conditions and different preset altitudes; Wherein, the intensity prediction module includes: A first LSTM submodule is configured to input the air signal test data packet into a first LSTM network of a signal prediction model to obtain the air signal feature information; The second LSTM submodule is used to input the air signal feature information into the second LSTM network of the signal prediction model to obtain a prediction result based on the cell information and weather information.

9. An electronic device comprising a processor and a memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the neural network-based drone air signal strength prediction method according to any one of claims 1 to 7 are implemented.

10. A processor-readable storage medium, characterized in that: The processor-readable storage medium stores a computer program, which is used to enable the processor to execute the steps of the neural network-based drone air signal strength prediction method according to any one of claims 1 to 7.

Images