A method, system, equipment and medium for generating collaborative decision-making criteria for unmanned aerial vehicles

Abstract

The present invention discloses a method, system, equipment and medium for generating collaborative decision criteria for unmanned aerial vehicles, and relates to the technical field of unmanned aerial vehicles. When decision criteria trigger data is received, the mission data of the unmanned aerial vehicle and the corresponding mission emergency status data are obtained. When the mission emergency status data is not urgent, the surrounding environment image data corresponding to the unmanned aerial vehicle is used to identify the opponent's equipment, and the number of opponent's equipment and the opponent's equipment feature data are obtained. When the number of opponent's equipment is less than a preset value, the unmanned aerial vehicle equipment feature data corresponding to the unmanned aerial vehicle and the ground data collected by the ground control station are extracted. The opponent's equipment feature data, the unmanned aerial vehicle equipment feature data and the ground data are rasterized to construct an environmental situation grid image. The environmental situation grid image is input into a multi-collaborative decision criterion classification model, which can scientifically and accurately generate collaborative decision criteria that meet the current environmental situation of the unmanned aerial vehicle.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle technology, and in particular to a method, system, equipment and medium for generating collaborative decision-making criteria for unmanned aerial vehicles. Background Art

[0002] With the rapid development of drone technology, drones can perform various complex tasks, but during the execution of tasks, drones are constrained by factors such as task diversity, environmental complexity, and insufficient computing power, making it difficult for them to make autonomous decisions in complex and changing task environments. When encountering emergencies during the execution of tasks, drones need to adjust their tasks according to real-time conditions, but when adjusting their own task planning, they need to follow certain decision-making criteria, such as maximizing the task completion rate, maximizing the area coverage rate, and maximizing the task planning time.

[0003] The existing methods for generating decision criteria for UAVs mainly include rule matching based on rule base and rule prediction based on deep reinforcement learning. The former manually extracts rules and builds a rule base after analyzing and processing a large amount of existing historical data for subsequent rule matching. This method requires manual extraction, is time-consuming and labor-intensive, and is difficult to adapt to dynamic situation environments. The latter can match different situations in a dynamic simulation manner and is committed to predicting various situations that may occur in the future. However, it predicts a single criterion and does not consider the correlation and priority between multiple criteria, resulting in poor applicability of the generated decision criteria. Summary of the invention

[0004] The present invention provides a method, system, equipment and medium for generating collaborative decision criteria for unmanned aerial vehicles, which solves the technical problem that the existing method for generating decision criteria for unmanned aerial vehicles can only predict a single criterion, does not consider the correlation and priority between multiple criteria, and causes the generated decision criteria to have poor applicability.

[0005] The present invention provides a method for generating a collaborative decision-making criterion for an unmanned aerial vehicle, comprising:

[0006] When decision criterion trigger data is received, the mission data of the UAV and the corresponding mission emergency status data are obtained;

[0007] When the mission emergency status data is not urgent, the surrounding environment image data corresponding to the UAV is used to identify the opponent's equipment to obtain the number of opponent's equipment and the opponent's equipment feature data;

[0008] When the number of the opponent's equipment is less than a preset value, extracting the UAV equipment feature data corresponding to the UAV and the ground data collected by the ground control station;

[0009] rasterizing the opponent equipment feature data, the UAV equipment feature data and the ground data to construct an environmental situation grid image;

[0010] The environmental situation grid image is input into a multi-cooperative decision-making criterion classification model to obtain the cooperative decision-making criterion corresponding to the UAV.

[0011] Optionally, the step of obtaining the mission data of the drone and the corresponding mission emergency status data includes:

[0012] Get the drone authorization status data corresponding to the drone;

[0013] When the authorization status data of the drone is authorized, emergency status judgment is performed on the mission data of the drone to obtain mission emergency status data.

[0014] Optionally, the step of rasterizing the opponent equipment feature data, the drone equipment feature data and the ground data to construct an environmental situation grid image includes:

[0015] Performing feature fusion on the opponent equipment feature data, the UAV equipment feature data and the ground data to obtain an initial real situation grid image set;

[0016] Rasterizing the initial true situation grid image set to obtain a target true situation grid image set;

[0017] According to the target real situation grid image set, an upsampling algorithm SMOTE is used to generate samples to obtain a first virtual situation grid image set;

[0018] According to the target real situation grid image set, using the information diffusion method MTD to generate samples, to obtain a second virtual situation grid image set;

[0019] Using the first virtual situation grid image set and the second virtual situation grid image set to construct a virtual situation grid image set;

[0020] Grid image samples are selected from the target real situation grid image set and the virtual situation grid image set to construct an environment situation grid image.

[0021] Optionally, the step of generating samples by using an upsampling algorithm SMOTE according to the target real situation grid image set to obtain a first virtual situation grid image set comprises:

[0022] Using the equipment in the target real situation grid image set as first sample generation objects respectively to obtain a first grid image sample set;

[0023] Selecting a root sample of a synthetic sample from the first grid image sample set according to a preset selection standard;

[0024] Selecting a plurality of first sample generation objects whose Euclidean distances to the root sample are within a preset distance range from the first grid image sample set to obtain a neighbor auxiliary sample point set;

[0025] Selecting auxiliary sample points from the neighboring auxiliary sample point set according to a preset upsampling rate to obtain an auxiliary sample point set;

[0026] Substituting the auxiliary sample points in the auxiliary sample point set into the preset sample formula to perform sample update and generate virtual samples;

[0027] The preset sample formula is:

[0028] x new =x i +(x′ i -x i )×rand(0,1);

[0029] Among them, x new is the first virtual situation grid image; x i is the root sample; x′ i is an auxiliary sample point; rand(0,1) is a random function between 0 and 1;

[0030] All of the virtual samples are used to construct a first virtual situation grid image set.

[0031] Optionally, the step of generating samples by using an information diffusion method MTD according to the target real situation grid image set to obtain a second virtual situation grid image set comprises:

[0032] Using the equipment in the target real situation grid image set as second sample generation objects respectively to obtain a second grid image sample set;

[0033] Select the maximum value and the minimum value corresponding to each equipment feature in the second grid image sample set respectively to obtain the maximum value and the minimum value of the feature;

[0034] Calculate the average value between the maximum value of the feature and the minimum value of the feature to obtain the feature mean;

[0035] Calculate the number of equipment with a value less than the characteristic mean in the second grid image sample set to obtain a first number of equipment;

[0036] Calculate the number of devices with a value greater than the feature mean in the second grid image sample set to obtain a second number of devices;

[0037] Substituting the first equipment number, the second equipment number and the characteristic mean into a preset boundary value calculation formula to perform boundary value calculation, and obtaining an upper boundary value and a lower boundary value corresponding to the equipment characteristic;

[0038] The preset boundary value calculation formula is:

[0039]

[0040]

[0041]

[0042]

[0043]

[0044] Among them, UB is the upper boundary value; CL is the characteristic mean; Skew U is the right skewness based on the feature mean; is the equipment variance corresponding to the second grid image sample set; N U is the number of second equipment; LB is the lower boundary value; Skew L is the left skewness based on the feature mean; N L is the number of the first equipment; x j is the observed value; is the mean; n is the number of devices corresponding to the second grid image sample set;

[0045] Substituting the upper boundary value and the lower boundary value into a preset triangular membership function value calculation formula to calculate and obtain a triangular membership function value corresponding to the equipment feature;

[0046] The preset triangular membership function value calculation formula is:

[0047]

[0048] Among them, MF is the triangular membership function value; x′ is a virtual sample; LB is the lower boundary value; CL is the feature mean; UB is the upper boundary value;

[0049] The triangular membership function value greater than the preset function value is used as a virtual sample of equipment;

[0050] A second virtual situation grid image set is constructed using all the virtual samples of the equipment.

[0051] Optionally, the step of selecting grid image samples from the target real situation grid image set and the virtual situation grid image set to construct an environment situation grid image comprises:

[0052] According to a preset equipment type selection standard, selecting equipment types from the target real situation grid image set to obtain equipment types;

[0053] According to the equipment type and the preset selection quantity, multiple supplementary samples are selected from the virtual situation grid image set to obtain a supplementary sample set corresponding to the equipment type and count the number of sample supplements in real time;

[0054] When the number of sample supplements is less than a preset number threshold, the step of selecting equipment types from the target real situation grid image set according to a preset equipment type selection standard to obtain equipment types is executed;

[0055] When the number of sample supplementation times is equal to a preset number threshold, all supplementary sample sets corresponding to the current moment are used to construct an environmental situation grid image.

[0056] Optionally, the step of inputting the environmental situation grid image into a multi-cooperative decision-making criterion classification model to obtain the collaborative decision-making criterion corresponding to the UAV includes:

[0057] Experts label the environmental situation grid image according to preset labels to obtain category labels corresponding to the environmental situation grid image;

[0058] Taking the environmental situation grid image as input and the category label as output, a preset multi-classification neural network model is trained to obtain a multi-cooperative decision-making criterion classification model;

[0059] The environmental situation grid image is input into the multi-cooperative decision-making criterion classification model to obtain the cooperative decision-making criterion corresponding to the UAV.

[0060] The present invention also provides a UAV collaborative decision-making criterion generation system, comprising:

[0061] A mission emergency data acquisition module, used to acquire the mission data of the UAV and the corresponding mission emergency data when receiving the decision criterion trigger data;

[0062] A module for obtaining the number of opponent equipment and the characteristic data of opponent equipment, which is used to identify the opponent equipment using the surrounding environment image data corresponding to the UAV when the mission emergency status data is not urgent, and obtain the number of opponent equipment and the characteristic data of opponent equipment;

[0063] The UAV equipment characteristic data and ground data extraction module is used to extract the UAV equipment characteristic data corresponding to the UAV and the ground data collected by the ground control station when the number of the opponent's equipment is less than a preset value;

[0064] An environment situation grid image acquisition module is used to rasterize the opponent equipment feature data, the UAV equipment feature data and the ground data to construct an environment situation grid image;

[0065] The collaborative decision criterion obtaining module is used to input the environmental situation grid image into a multi-collaborative decision criterion classification model to obtain the collaborative decision criterion corresponding to the UAV.

[0066] The present invention also provides an electronic equipment, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of implementing any of the above-mentioned methods for generating drone collaborative decision-making criteria.

[0067] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed, the method for generating a collaborative decision criterion for unmanned aerial vehicles as described above is implemented.

[0068] It can be seen from the above technical solutions that the present invention has the following advantages:

[0069] During the flight of the UAV, the present invention comprehensively considers factors such as the emergency situation of the UAV's mission execution and the size of the evaluated calculation amount. Under the premise that the mission execution is not urgent and the evaluated calculation amount is small, it can scientifically and accurately generate collaborative decision-making criteria that meet the UAV's current environmental situation. Secondly, in the process of generating collaborative decision-making criteria, the ground data sent by the ground control station and the data collected by the UAV are integrated, effectively combining the global ground information and the local information of the UAV. Finally, a multi-collaborative decision-making criterion classification model is used to generate decision criteria under different environmental situations, realizing a one-to-many matching of dynamic environmental situations and decision criteria, and is committed to making predictions for various situations that may occur in the future, so that the UAV can adapt to complex and changeable environments. In addition, in the process of training the multi-collaborative decision-making criterion classification model, in order to address the problem of insufficient training samples, the present invention proposes to use SMOTE that does not consider sample labels and MTD method that does not consider sample labels to generate a virtual situation grid image set, which effectively solves the problem of model overfitting caused by insufficient sample data. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0071] Figure 1A flowchart of a method for generating a collaborative decision-making criterion for unmanned aerial vehicles provided in Embodiment 1 of the present invention;

[0072] Figure 2 A flowchart of a method for generating a collaborative decision-making criterion for unmanned aerial vehicles provided in Embodiment 2 of the present invention;

[0073] Figure 3 A schematic diagram of one-to-many matching between environmental situation data and decision criteria provided in the second embodiment of the present invention;

[0074] Figure 4 A schematic diagram of an environmental situation grid image provided by Embodiment 2 of the present invention;

[0075] Figure 5 A schematic diagram of the model structure of a preset multi-classification neural network model provided in the second embodiment of the present invention;

[0076] Figure 6 A flowchart for constructing a multi-cooperative decision-making criterion classification model provided in the second embodiment of the present invention;

[0077] Figure 7 A schematic diagram of an image of a first virtual situation grid image set generated by an upsampling algorithm provided in Embodiment 2 of the present invention;

[0078] Figure 8 An image schematic diagram of a second virtual situation grid image set generated by the information diffusion method provided in the second embodiment of the present invention;

[0079] Fig. 9 A flowchart of a collaborative decision-making criterion generation method provided in Embodiment 2 of the present invention;

[0080] Fig.10 This is a structural block diagram of a UAV collaborative decision-making criterion generation system provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0081] The embodiments of the present invention provide a method, system, equipment and medium for generating collaborative decision criteria for unmanned aerial vehicles, which are used to solve the technical problem that the existing method for generating decision criteria for unmanned aerial vehicles can only predict a single criterion, does not consider the correlation and priority between multiple criteria, and causes the generated decision criteria to have poor applicability.

[0082] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0083] See also Figure 1 , Figure 1 A flowchart of the steps of a method for generating a collaborative decision criterion for unmanned aerial vehicles provided in Embodiment 1 of the present invention.

[0084] A method for generating a collaborative decision criterion for a drone provided in Example 1 of the present invention includes:

[0085] Step 101: When decision criterion trigger data is received, the mission data of the drone and the corresponding mission emergency status data are obtained.

[0086] In the embodiment of the present invention, the decision criterion trigger data is triggered when an emergency is received, and can also be set to be called at a scheduled time. Specifically, the scheduled task calling time interval is set to T or time point t, and according to the time interval T or time point t, the UAV collaborative decision criterion generation method in the embodiment of the present invention is automatically called every T time or at the specified time point t.

[0087] UAV authorization status check: Before the UAV flies, the intelligent processing chip installed on it will pre-set decision criteria so that it can make decisions according to the criteria when encountering emergencies. If it is not authorized before the flight, it will perform tasks according to ground instructions even if it encounters emergencies during the flight. Therefore, it is necessary to obtain the UAV authorization status data corresponding to the UAV. When the UAV authorization status data is authorized, the UAV's mission data is judged as an emergency situation to obtain the mission emergency status data.

[0088] Step 102: When the mission emergency status data is not urgent, the surrounding environment image data corresponding to the UAV is used to identify the opponent's equipment to obtain the number of opponent's equipment and the opponent's equipment feature data.

[0089] In an embodiment of the present invention, if the mission emergency status data of the mission to be performed is urgent, the UAV continues to complete the mission, and the decision criterion at this time is the maximum mission compliance rate. If the mission emergency status data of the mission to be performed is not urgent, the UAV uses the surrounding environment image data corresponding to the UAV to identify the opponent's equipment, and obtains the number of opponent's equipment and opponent's equipment feature data, wherein the opponent's equipment data refers to equipment, items, animals, etc. that do not appear in the mission plan, and the types of opponent equipment are different according to different technical fields. The opponent's equipment feature data generally includes the speed, position, height and type characteristics of the equipment. Based on the collected surrounding environment image data, the number of opponent's equipment and the opponent's equipment feature data are calculated, so that the calculation amount of the subsequent generation of the decision criterion can be evaluated, and based on the amount of calculation, it is decided whether to request ground help.

[0090] Step 103: When the number of opponent equipment is less than a preset value, extract the UAV equipment feature data corresponding to the UAV and the ground data collected by the ground control station.

[0091] In the embodiment of the present invention, if the amount of computation for evaluation is too large, resulting in the limitation of the computing power of the intelligent processing chip, the UAV requests the ground server to perform the calculation, and the decision criteria returned by the ground server to the UAV are used as the final decision criteria, i.e., the collaborative decision criteria. If the amount of computation for evaluation is small, the UAV autonomously generates the collaborative decision criteria, i.e., based on the ground data collected by the ground control station, obtains the speed, position, height and type feature information of the equipment of the two opposing parties, and extracts the UAV equipment feature data corresponding to the UAV.

[0092] Step 104: rasterize the opponent's equipment feature data, the UAV equipment feature data and the ground data to construct an environmental situation grid image.

[0093] In an embodiment of the present invention, the opponent equipment feature data calculated by the drone, the drone equipment feature data, and the ground data collected by the ground control station are feature fused to obtain an initial real situation grid image set. The initial real situation grid image set is rasterized to obtain a target real situation grid image set. According to the target real situation grid image set, the upsampling algorithm SMOTE is used to generate samples to obtain a first virtual situation grid image set. According to the target real situation grid image set, the information diffusion method MTD is used to generate samples to obtain a second virtual situation grid image set. The first virtual situation grid image set and the second virtual situation grid image set are used to construct a virtual situation grid image set. Grid image samples are selected from the target real situation grid image set and the virtual situation grid image set to construct an environmental situation grid image.

[0094] Step 105: Input the environmental situation grid image into the multi-cooperative decision-making criterion classification model to obtain the cooperative decision-making criterion corresponding to the UAV.

[0095] In an embodiment of the present invention, the environmental situation grid image is annotated by experts according to preset labels to construct category labels corresponding to the environmental situation grid image. The environmental situation grid image is used as input, and the category label is used as output to train the preset multi-classification neural network model to obtain a multi-cooperative decision-making criterion classification model. The environmental situation grid image is input into the multi-cooperative decision-making criterion classification model to obtain the collaborative decision-making criterion corresponding to the drone.

[0096] In an embodiment of the present invention, when decision criterion trigger data is received, the mission data of the drone and the corresponding mission emergency data are obtained. When the mission emergency data is not urgent, the surrounding environment image data corresponding to the drone is used to identify the opponent's equipment, and the number of opponent's equipment and the opponent's equipment feature data are obtained. When the number of opponent's equipment is less than the preset value, the drone equipment feature data corresponding to the drone and the ground data collected by the ground control station are extracted. The opponent's equipment feature data, the drone equipment feature data and the ground data are rasterized to construct an environmental situation grid image. The environmental situation grid image is input into a multi-cooperative decision criterion classification model to generate a collaborative decision criterion that conforms to the current situation. By integrating the ground transmission data, the drone collection data and the drone's own tasks, the collaborative decision criterion corresponding to the drone-ground control station is intelligently generated in a complex situation environment. The technical problem that the existing drone decision criterion generation method can only predict a single criterion without considering the correlation and priority between multiple criteria, resulting in poor applicability of the generated decision criterion is solved. Taking into account the emergency situation of the drone's mission execution, the size of the number of opponent's equipment evaluated, and other factors, the collaborative decision criterion that conforms to the drone in the current environmental situation can be scientifically and accurately generated on the premise that the mission execution is not urgent and the number of opponent's equipment evaluated is small.

[0097] See also Figure 2 , Figure 2 A flowchart of the steps of a method for generating a collaborative decision criterion for unmanned aerial vehicles provided in Embodiment 2 of the present invention.

[0098] Another method for generating a collaborative decision criterion for unmanned aerial vehicles provided in Example 2 of the present invention includes:

[0099] Step 201: When decision criterion trigger data is received, the mission data of the drone and the corresponding mission emergency status data are obtained.

[0100] Further, step 201 may include the following sub-steps S11-S12:

[0101] S11. Obtain the drone authorization status data corresponding to the drone.

[0102] S12. When the authorization status data of the drone is authorized, the mission data of the drone is subjected to emergency status judgment to obtain mission emergency status data.

[0103] In an embodiment of the present invention, the authorization status of the drone is checked. Before the drone flies, the intelligent processing chip installed on it will pre-set decision criteria so that it can make decisions according to the criteria when encountering emergencies. If it has been authorized before the flight, the drone needs to communicate with the ground control station in real time and make collaborative decisions during the flight. Before the drone flies, its task management module stores the task sequence (such as penetration, reconnaissance, circling and strike, etc.) that it needs to perform during the flight and the emergency situation of the corresponding tasks (such as emergency, non-emergency, etc.). Therefore, when the decision criterion trigger data is received, first, it is determined whether the drone authorization status data corresponding to the drone is authorized. If it is authorized, it is further determined whether the task emergency status data corresponding to the drone is non-urgent.

[0104] Step 202: When the mission emergency status data is not urgent, the surrounding environment image data corresponding to the UAV is used to identify the opponent's equipment to obtain the number of opponent's equipment and the opponent's equipment feature data.

[0105] In an embodiment of the present invention, during the flight, when the task management module is called up, the emergency situation of the current task performed by the drone will be evaluated. If it is a non-urgent task, the surrounding environment image data collected by the drone will be used to identify the opponent's equipment, and the number of opponent's equipment and the characteristic data of opponent's equipment will be obtained. The drone will generate a decision criterion after a comprehensive evaluation based on the environmental situation and the amount of calculation. That is, based on the data collected by the drone, the speed, position, and height characteristic information of the equipment of both parties are calculated, and then the ground data and the equipment characteristic information are integrated to calculate the environmental situation grid image that represents the environment. Finally, the grid image is input into the multi-cooperative decision criterion prediction model to generate a multi-cooperative decision criterion that conforms to the current environmental situation. If it is an urgent task, it will continue to execute the previously scheduled task.

[0106] Step 203: When the number of opponent equipment is less than a preset value, extract the UAV equipment feature data corresponding to the UAV and the ground data collected by the ground control station.

[0107] In an embodiment of the present invention, the drone calculates the types and quantities of equipment collected from the surrounding environment image data. If it is calculated that the number of surrounding opponent equipment is not large, that is, the number of opponent equipment is less than a preset value, the drone can autonomously generate a decision criterion by extracting the drone equipment features collected by the volumetric drone and the ground data collected by the ground control station to autonomously generate multiple collaborative decision criteria. If it is calculated that the number of surrounding opponent equipment is too large, resulting in the computing power of the intelligent processing chip being limited, the drone requests the ground server of the ground control station to perform calculations, and the decision criterion returned to the drone by the ground server is used as the final collaborative decision criterion.

[0108] Step 204: feature fusion of the opponent's equipment feature data, the UAV equipment feature data and the ground data to obtain an initial real situation grid image set.

[0109] In an embodiment of the present invention, characteristic information on the speed, position, height and type of equipment of the opposing parties is obtained, i.e., the opponent's equipment characteristic data and the UAV's equipment characteristic data are obtained, and characteristic information on the speed, position, height and type of equipment is obtained from data sent from the ground, i.e., the ground data. The equipment characteristic information calculated by the UAV is fused with the equipment characteristic information obtained from the ground data to obtain an initial real situation grid image set.

[0110] like Figure 3 As shown in the figure, the environmental situation corresponding to the initial real situation grid image set is composed of the characteristic information of each equipment in the environment around the drone. The equipment information comes from the data collected by the drone and the data sent from the ground. The characteristic information is generally a characteristic vector of the equipment attribute set, such as the attribute set of the equipment = {equipment of both sides of the confrontation, equipment type, speed, x coordinate, y coordinate, height}. The decision criteria can be the highest cost-effectiveness, the maximum task completion rate, the maximum area coverage rate, the fastest task planning time, the maximum damage ratio of both sides of the confrontation, the maximum target damage rate, the maximum own survival rate, etc. The environmental situation at a certain moment can correspond to multiple decision criteria.

[0111] Step 205: rasterize the initial true situation grid image set to obtain a target true situation grid image set.

[0112] In the embodiment of the present invention, Figure 4 As shown in the figure, according to the calculated equipment feature information around the drone, the map near the drone is gridded so that there is only one equipment in each cell. In order to vividly display the targets of both sides and the drone itself, the equipment of the opponent is marked as black in the cell, the equipment of the opponent is marked as gray in the cell, the drone itself is represented by a black triangle, and the target is white if there is no target in the cell. The scattered black and gray marks are some random targets. Different from the RGB values ​​stored in the image, the grid data stores the various attribute states of the target, such as speed, x coordinate, y coordinate, etc. The final environmental situation information is saved as a target real situation grid image set in the form of rasterized images. The drone detection data reflects local information, and the ground data reflects global information. By fusing the data of the two and expressing them with rasterized images, that is, the target real situation grid image set combines the ground global information and the drone local information, it can more accurately reflect the environmental situation.

[0113] Step 206: According to the target real situation grid image set, an upsampling algorithm SMOTE is used to generate samples to obtain a first virtual situation grid image set.

[0114] Further, step 206 may include the following sub-steps S21-S26:

[0115] S21. Use the equipment in the target real situation grid image set as the first sample generation object respectively to obtain a first grid image sample set.

[0116] S22: Select a root sample of the synthesized sample from the first grid image sample set according to a preset selection standard.

[0117] S23. Selecting a plurality of first sample generation objects whose Euclidean distances to the root sample are within a preset distance range from the first grid image sample set to obtain a nearest auxiliary sample point set.

[0118] S24. Select auxiliary sample points from the neighboring auxiliary sample point set according to a preset upsampling rate to obtain an auxiliary sample point set.

[0119] S25, substituting the auxiliary sample points in the auxiliary sample point set into the preset sample formula to perform sample update and generate virtual samples.

[0120] The preset sample formula is:

[0121] x new =x i +(x′ i -x i )×rand(0,1);

[0122] Among them, x new is the first virtual situation grid image; x i is the root sample; x′ i is an auxiliary sample point; rand(0,1) is a random function between 0 and 1.

[0123] S26. Use all virtual samples to construct a first virtual situation grid image set.

[0124] The preset selection criteria refers to a selection method for root samples that is set in advance based on actual needs.

[0125] The preset distance range refers to the selection range constructed based on the preset number k of neighbor auxiliary sample points after calculating the Euclidean distances between all first sample generation objects and the root sample, and then sorting them in ascending order. That is, the first sample generation objects are sorted in ascending order according to their Euclidean distances, and the first k first sample generation objects are selected to construct a neighbor auxiliary sample point set. The number k of neighbor auxiliary sample points is generally an odd number, such as k=5.

[0126] The preset up-sampling rate N is N=100%, 200%, 300%, etc.

[0127] In the embodiment of the present invention, the equipment in the target real situation grid image set is respectively used as the first sample generation object, and the first sample generation object includes equipment feature data. The selected equipment E = {x1, x2, ..., x n}, that is, the first grid image sample set is obtained. From the first grid image sample set E = {x1, x2, …, x n}Select sample x i As the root sample of the synthetic sample, calculate the sample x i In the first grid image sample set E = {x1, x2, ..., x i}, k (generally an odd number, such as k = 5) nearest neighbor auxiliary sample points based on Euclidean distance are used to obtain a nearest neighbor auxiliary sample point set.

[0128] Set the preset upsampling rate N, and randomly select from the k nearest neighbor auxiliary sample points according to the upsampling rate N. Auxiliary sample points x′ i , and obtain the auxiliary sample point set. For each randomly selected auxiliary sample point, a new sample point is generated using a preset sample formula, and all real equipment data are traversed to obtain the first virtual situation grid image set.

[0129] Step 207: According to the target real situation grid image set, the information diffusion method MTD is used to generate samples to obtain a second virtual situation grid image set.

[0130] Further, step 207 may include the following sub-steps S31-S39:

[0131] S31. Use the equipment in the target real situation grid image set as second sample generation objects respectively to obtain a second grid image sample set.

[0132] S32, respectively selecting the maximum value and the minimum value corresponding to each equipment feature in the second grid image sample set to obtain the maximum value and the minimum value of the feature.

[0133] S33. Calculate the average value between the maximum value and the minimum value of the feature to obtain the feature mean.

[0134] S34. Calculate the number of equipment that is smaller than the feature mean in the second grid image sample set to obtain a first number of equipment.

[0135] S35. Calculate the number of equipment with a value greater than the feature mean in the second grid image sample set to obtain a second number of equipment.

[0136] S36. Substitute the first equipment number, the second equipment number and the characteristic mean into a preset boundary value calculation formula to perform boundary value calculation to obtain an upper boundary value and a lower boundary value corresponding to the equipment characteristic.

[0137] The preset boundary value calculation formula is:

[0138]

[0139]

[0140]

[0141]

[0142]

[0143] Among them, UB is the upper boundary value; CL is the characteristic mean; Skew U is the right skewness based on the feature mean; is the equipment variance corresponding to the second grid image sample set; N U is the number of second equipment; LB is the lower boundary value; Skew L is the left skewness based on the feature mean; N L is the number of the first equipment; x j is the observed value; x is the mean; n is the number of devices corresponding to the second grid image sample set.

[0144] S37. Substitute the upper boundary value and the lower boundary value into a preset triangular membership function value calculation formula to perform calculation to obtain a triangular membership function value corresponding to the equipment feature.

[0145] The preset triangular membership function value calculation formula is:

[0146]

[0147] Among them, MF is the triangular membership function value; x′ is a virtual sample; LB is the lower boundary value; CL is the feature mean; UB is the upper boundary value.

[0148] S38. Use the triangular membership function value that is greater than the preset function value as the equipment virtual sample.

[0149] S39. Use all equipment virtual samples to construct a second virtual situation grid image set.

[0150] In an embodiment of the present invention, the equipment in the target real situation grid image set is respectively used as the second sample generation object to obtain the second grid image sample set. The maximum value and minimum value corresponding to each equipment feature in the second grid image sample set are respectively selected to obtain the feature maximum value max and the feature minimum value min. The average value between the feature maximum value and the feature minimum value is calculated to obtain the feature mean value CL = (min + max) / 2. Based on the equipment features corresponding to the feature mean value, the number of equipment in the second grid image sample set that is less than the feature mean value CL is calculated to obtain the first equipment number N LBased on the equipment features corresponding to the feature mean, the number of equipment in the second grid image sample set that is greater than the feature mean CL is calculated to obtain the second equipment number N U . Substitute the first number of equipment, the second number of equipment and the characteristic mean into the preset boundary value calculation formula to perform boundary value calculation, and obtain the upper boundary value and the lower boundary value corresponding to the equipment feature. According to the upper boundary value and the lower boundary value, generate a virtual sample, that is, substitute the upper boundary value and the lower boundary value into the preset triangular membership function value calculation formula for calculation, and obtain the triangular membership function value corresponding to the equipment feature, and use the triangular membership function value MF as the generated virtual sample. Generate a randomly uniformly distributed value r between [0, 1], that is, the preset function value. If MF(x′)>r, the generated sample is qualified and the equipment virtual sample is obtained. Otherwise, the sample will be discarded. After the equipment virtual samples corresponding to all equipment features are calculated, all equipment virtual samples corresponding to the current moment are used to construct the second virtual situation grid image set.

[0151] Step 208: Use the first virtual situation grid image set and the second virtual situation grid image set to construct a virtual situation grid image set.

[0152] In an embodiment of the present invention, the SMOTE upsampling algorithm that does not consider sample labels and the MTD information diffusion method that does not consider sample labels are respectively used to obtain virtual situation grid images corresponding to the target real situation grid image set, and all virtual situation grid images, i.e., the first virtual situation grid image set and the second virtual situation grid image set, are used to construct a virtual situation grid image set.

[0153] Step 209: Select grid image samples from the target real situation grid image set and the virtual situation grid image set to construct an environment situation grid image.

[0154] Further, step 209 may include the following sub-steps S41-S44:

[0155] S41. According to a preset equipment type selection standard, select equipment types from a target real situation grid image set to obtain equipment types.

[0156] S42. Select multiple supplementary samples from the virtual situation grid image set according to the equipment type and the preset selection quantity, obtain a supplementary sample set corresponding to the equipment type and count the number of sample supplements in real time.

[0157] S43. When the number of sample supplements is less than a preset number threshold, jump to execute the step of selecting equipment types from the target real situation grid image set according to the preset equipment type selection standard to obtain the equipment type.

[0158] S44. When the number of sample supplementation times is equal to the preset number threshold, all supplementary sample sets corresponding to the current moment are used to construct an environmental situation grid image.

[0159] The preset equipment type selection standard refers to randomly selecting equipment types from the target real situation grid image set.

[0160] The preset selection quantity n refers to the critical value of the quantity corresponding to each selection of supplementary samples.

[0161] The preset number threshold m refers to the critical value corresponding to the number of selections.

[0162] In the embodiment of the present invention, the equipment E corresponding to the target real situation grid image set is E={E1, E2, ..., E j} Randomly select equipment type E i , according to equipment type E i , randomly select n virtual samples from the virtual samples generated by SMOTE and MTD as equipment E i Repeat the above steps m times, selecting different equipment each time, to obtain a supplementary sample set containing n×m equipment feature information and count the number of sample supplements in real time. When the number of sample supplements is less than the preset number threshold, jump to the step of selecting equipment types from the target real situation grid image set according to the preset equipment type selection standard to obtain the equipment type. When the number of sample supplements is equal to the preset number threshold, all the supplementary sample sets corresponding to the current moment are used to construct the environmental situation grid image.

[0163] Step 210: Input the environmental situation grid image into the multi-cooperative decision-making criterion classification model to obtain the cooperative decision-making criterion corresponding to the UAV.

[0164] Further, step 210 may include the following sub-steps S51-S53:

[0165] S51, annotating the environment situation grid image by experts according to preset labels, and constructing category labels corresponding to the environment situation grid image.

[0166] S52. Taking the environmental situation grid image as input and the category label as output, the preset multi-classification neural network model is trained to obtain a multi-cooperative decision-making criterion classification model.

[0167] S53, inputting the environmental situation grid image into the multi-cooperative decision-making criterion classification model to obtain the cooperative decision-making criterion corresponding to the UAV.

[0168] The preset tags may be one or more of the following: highest cost-effectiveness, highest mission completion rate, highest area coverage rate, fastest mission planning time, highest damage ratio between the opposing parties, highest target damage rate, and highest own survival rate.

[0169] In an embodiment of the present invention, an expert annotation is performed on a sample data set of real and virtual pipe situation grid images to obtain a category label of an environmental situation grid image, that is, an expert annotation is performed on the environmental situation grid image according to a preset label to construct a category label corresponding to the environmental situation grid image. The environmental situation grid image is used as input and the category label is used as output to train a preset multi-classification neural network model to obtain a multi-cooperative decision-making criterion classification model.

[0170] Specifically, if Figure 6 As shown in the figure, firstly, the speed, position and height feature information of the equipment of both sides in the UAV data and the ground data sent by the ground control station are obtained and fused and calculated to obtain the grid image representing the environmental situation, and the real situation grid image set of the target is obtained. Since the collected environmental situation data is very limited, the virtual environmental situation grid image, that is, the virtual situation grid image set, is generated based on a small amount of real environmental situation data and the SMOTE method without considering the sample label and the MTD method without considering the sample label. Then, the environmental situation grid image is annotated by experts to obtain the category label of the environmental situation grid image. After obtaining the environmental situation grid image and the category label, a preset multi-classification neural network model is established, and the real and virtual environmental situation grid images are used as the input of the neural network model, and the category label is used as the output. The neural network model is trained to obtain a multi-cooperative decision-making criterion classification model. By inputting the environmental situation grid image into the multi-cooperative decision-making criterion classification model, the collaborative decision criterion corresponding to the UAV can be obtained.

[0171] In the process of model training and testing, the real and virtual environment situation grid images, i.e., environment situation grid images, are divided into training sets, validation sets, and test sets. The training sets and validation sets are used to train and evaluate the neural network model. The validation set data is used to verify the classification accuracy of the multi-cooperative decision-making criteria. When the classification accuracy reaches the set classification accuracy, the trained model is the multi-cooperative decision-making criteria classification model. Otherwise, the model architecture and parameters are adjusted, and training is continued until the set accuracy is reached. After obtaining the final multi-cooperative decision-making criteria classification model, the generalization ability and accuracy of the model are evaluated using the test set.

[0172] During the model training process, the cross entropy loss function is selected, and the optimizer uses Adam for back-propagation to update the network model parameters. By setting different mini-batch-size, epoch, learning rate and other parameters, a model with higher accuracy is obtained in the process of continuous tuning.

[0173] Preferably, the model is evaluated using a sample-specific evaluation method, that is, the F1 value is used for evaluation, as shown in the following formula.

[0174]

[0175]

[0176]

[0177] Among them, F1 is the index value; precision is the precision rate; recall is the recall rate; p is the number of samples; Y i is the set of labels associated with x; h(x i ) returns the correct set of labels for x for a multi-label classifier.

[0178] like Figure 7 As shown, Figure 7 This is a schematic diagram of the first virtual situation grid image set generated by the upsampling algorithm provided in the second embodiment of the present invention. The black dots are real situation data samples, and the gray triangles are generated virtual situation data samples. It can be seen from the figure that the samples generated by SMOTE are all on the line connecting the two known samples.

[0179] like Figure 8 As shown, Figure 8 This is a schematic diagram of the second virtual situation grid image set generated by the information diffusion method provided in the second embodiment of the present invention. The black dots are real situation data samples, and the gray triangles are generated virtual situation data samples. It can be seen from the figure that the distribution of samples generated by MTD is not necessarily between the two samples.

[0180] like Figure 5 As shown in the figure, the preset multi-classification neural network model includes an input layer, a convolution layer, a pooling layer and a fully connected layer. The input layer inputs the environmental situation grid image. The convolution layer is used to read the local information of the image-like environmental situation data. Each convolution layer contains multiple convolution kernels of the same size. A maximum pooling layer is added after the convolution layer to reduce parameters and prevent overfitting. A linear layer is added before inputting to the fully connected layer to reduce the data dimension. Finally, the fully connected layer is used to map the extracted features to the physical features. The last fully connected layer uses the activation function softmax to output multiple collaborative decision criteria, such as the maximum task completion rate, the maximum area coverage rate, the highest cost-effectiveness, etc. The number of decision criteria output is determined by comparing the predicted probability of each decision criterion with the set threshold.

[0181] In the embodiment of the present invention, Fig. 9As shown, during the flight, the drone sets the parameters according to the time in the scheduled task calling module, and calls the decision criteria to generate related programs. First, check whether the drone is authorized. Before the drone flies, the intelligent processing chip installed on it will pre-set the decision criteria so that it can make decisions according to the criteria when encountering emergencies. If it is not authorized before the flight, even if it encounters emergencies during the flight, it will not communicate with the ground and continue to perform the task according to the pre-set decision criteria. If it has been authorized before the flight, the drone needs to communicate with the ground in real time and make collaborative decisions during the flight. Then, the emergency situation of the drone's mission is checked. If the emergency situation of its mission is urgent, the drone continues to complete the mission. The decision criterion at this time is the maximum task compliance rate. It should be noted that when the drone performs the assigned task, it can still use the collaborative decision criteria to generate related methods. If the emergency situation of its mission is not urgent, the drone generates a decision criterion after a comprehensive assessment of the environmental situation.

[0182] When the above conditions are met, the UAV can generate collaborative decision-making criteria with the ground. First, the UAV calculates the types and quantities of surrounding equipment based on the collected surrounding image information. If the calculated number of surrounding opponent equipment is too large, resulting in the limitation of the computing power of the intelligent processing chip, the UAV requests the ground server to calculate and returns the decision criteria returned to the UAV by the ground server as the final decision criteria; if the calculated number of surrounding opponent equipment is not large, the UAV will generate the decision criteria autonomously. At this time, the UAV calculates the speed, position, height and type feature information of the equipment of both sides based on the collected environmental image data, and obtains the speed, position, height and type feature information of the equipment from the data sent from the ground. The equipment feature information calculated by the UAV is integrated with the equipment feature information obtained from the ground data to obtain the final environmental situation information, and the final environmental situation information is saved as a grid image in the form of a rasterized image. Finally, the calculated grid image is input into the constructed multi-collaborative decision-making criterion classification model to generate multi-collaborative decision-making criteria that conform to the current environmental situation.

[0183] See also Fig.10 , Fig.10 This is a structural block diagram of a UAV collaborative decision-making criterion generation system provided in Example 3 of the present invention.

[0184] A third example of the present invention provides a UAV collaborative decision-making criterion generation system, comprising:

[0185] The mission emergency data acquisition module 1001 is used to acquire the mission data of the drone and the corresponding mission emergency data when receiving the decision criterion trigger data.

[0186] The module 1002 for obtaining the number of opponent equipment and the characteristic data of the opponent equipment is used to identify the opponent equipment with the surrounding environment image data corresponding to the UAV when the mission emergency status data is not urgent, so as to obtain the number of opponent equipment and the characteristic data of the opponent equipment.

[0187] The UAV equipment characteristic data and ground data extraction module 1003 is used to extract the UAV equipment characteristic data corresponding to the UAV and the ground data collected by the ground control station when the number of opponent equipment is less than a preset value.

[0188] The environment situation grid image obtaining module 1004 is used to rasterize the opponent equipment feature data, the UAV equipment feature data and the ground data to construct an environment situation grid image.

[0189] The collaborative decision criterion obtaining module 1005 is used to input the environmental situation grid image into the multi-collaborative decision criterion classification model to obtain the collaborative decision criterion corresponding to the UAV.

[0190] Optionally, the mission emergency status data acquisition module 1001 includes:

[0191] The drone authorization status data acquisition module is used to obtain the drone authorization status data corresponding to the drone.

[0192] The mission emergency status data acquisition submodule is used to perform emergency status judgment on the mission data of the drone when the drone authorization status data is authorized, so as to obtain mission emergency status data.

[0193] Optionally, the environment situation grid image obtaining module 1004 includes:

[0194] The module for obtaining the initial real situation grid image set is used to fuse the opponent's equipment feature data, the UAV equipment feature data and the ground data to obtain the initial real situation grid image set.

[0195] The target real situation grid image set obtaining module is used to rasterize the initial real situation grid image set to obtain the target real situation grid image set.

[0196] The first virtual situation grid image set obtaining module is used to generate samples using the upsampling algorithm SMOTE according to the target real situation grid image set to obtain the first virtual situation grid image set.

[0197] The second virtual situation grid image set obtaining module is used to generate samples according to the target real situation grid image set by using the information diffusion method MTD to obtain the second virtual situation grid image set.

[0198] The virtual situation grid image set construction module is used to construct a virtual situation grid image set by using the first virtual situation grid image set and the second virtual situation grid image set.

[0199] The environment situation grid image acquisition submodule is used to select grid image samples from the target real situation grid image set and the virtual situation grid image set to construct the environment situation grid image.

[0200] Optionally, the first virtual situation grid image set obtaining module may perform the following steps:

[0201] The equipment in the target real situation grid image set is respectively used as the first sample generation object to obtain the first grid image sample set;

[0202] Selecting a root sample of the synthetic sample from the first grid image sample set according to a preset selection standard;

[0203] Selecting a plurality of first sample generation objects whose Euclidean distances to the root sample are within a preset distance range from the first grid image sample set to obtain a nearest neighbor auxiliary sample point set;

[0204] Select auxiliary sample points from the neighboring auxiliary sample point set according to a preset upsampling rate to obtain an auxiliary sample point set;

[0205] Substituting the auxiliary sample points in the auxiliary sample point set into the preset sample formula to update the sample, a virtual sample is generated;

[0206] The preset sample formula is:

[0207] x new =x i +(x′ i -x i )×rand(0,1);

[0208] Among them, x ne is the first virtual situation grid image; x i is the root sample; x′ i is an auxiliary sample point; rand(0,1) is a random function between 0 and 1;

[0209] All virtual samples are used to construct a first virtual situation grid image set.

[0210] Optionally, the second virtual situation grid image set obtaining module may perform the following steps:

[0211] The equipment in the target real situation grid image set is respectively used as the second sample generation object to obtain a second grid image sample set;

[0212] Select the maximum value and the minimum value corresponding to each equipment feature in the second grid image sample set respectively to obtain the maximum value and the minimum value of the feature;

[0213] Calculate the average value between the maximum value and the minimum value of the feature to get the feature mean;

[0214] Calculate the number of equipments with a value less than the feature mean in the second grid image sample set to obtain the first number of equipments;

[0215] Calculate the number of equipment with a value greater than the feature mean in the second grid image sample set to obtain a second number of equipment;

[0216] Substituting the first equipment number, the second equipment number and the characteristic mean into a preset boundary value calculation formula to perform boundary value calculation, and obtaining an upper boundary value and a lower boundary value corresponding to the equipment characteristic;

[0217] The preset boundary value calculation formula is:

[0218]

[0219]

[0220]

[0221]

[0222]

[0223] Among them, UB is the upper boundary value; CL is the characteristic mean; Skew U is the right skewness based on the feature mean; is the equipment variance corresponding to the second grid image sample set; N U is the number of second equipment; LB is the lower boundary value; Skew L is the left skewness based on the feature mean; N L is the number of the first equipment; x j is the observed value; x is the mean; n is the number of devices corresponding to the second grid image sample set;

[0224] Substitute the upper boundary value and the lower boundary value into the preset triangular membership function value calculation formula to calculate and obtain the triangular membership function value corresponding to the equipment feature; the preset triangular membership function value calculation formula is:

[0225]

[0226] Among them, MF is the triangular membership function value; x′ is a virtual sample; LB is the lower boundary value; CL is the feature mean; UB is the upper boundary value;

[0227] The triangular membership function value greater than the preset function value is used as a virtual sample of equipment;

[0228] The second virtual situation grid image set is constructed using all equipment virtual samples.

[0229] Optionally, the environment situation grid image obtaining submodule may perform the following steps:

[0230] According to a preset equipment type selection standard, equipment types are selected from a target real situation grid image set to obtain equipment types;

[0231] According to the equipment type and the preset selection quantity, multiple supplementary samples are selected from the virtual situation grid image set to obtain a supplementary sample set corresponding to the equipment type and count the number of sample supplements in real time;

[0232] When the number of sample supplements is less than a preset number threshold, the execution jumps to the step of selecting equipment types from the target real situation grid image set according to the preset equipment type selection standard to obtain the equipment type;

[0233] When the number of sample supplements is equal to the preset number threshold, all supplementary sample sets corresponding to the current moment are used to construct the environmental situation grid image.

[0234] Optionally, the collaborative decision criterion obtaining module 1005 includes:

[0235] The category label obtaining module is used to perform expert annotation on the environmental situation grid image according to preset labels and construct the category label corresponding to the environmental situation grid image.

[0236] The module for obtaining a multi-cooperative decision-making criterion classification model is used to take an environmental situation grid image as input and a category label as output, perform model training on a preset multi-classification neural network model, and obtain a multi-cooperative decision-making criterion classification model.

[0237] The collaborative decision criterion obtaining submodule is used to input the environmental situation grid image into the multi-collaborative decision criterion classification model to obtain the collaborative decision criterion corresponding to the UAV.

[0238] An embodiment of the present invention further provides an electronic device, which includes: a memory and a processor, wherein a computer program is stored in the memory; when the computer program is executed by the processor, the processor executes a method for generating collaborative decision criteria for unmanned aerial vehicles as described in any of the above embodiments.

[0239] The memory may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk or a ROM. The memory has a storage space for program codes for executing any method step in the above method. For example, the storage space for program codes may include individual program codes for implementing the various steps in the above method, respectively. These program codes may be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards or floppy disks. The program code may be compressed, for example, in an appropriate form. When these codes are run by a computing and processing device, the computing and processing device performs the various steps in the method for generating collaborative decision criteria for unmanned aerial vehicles described above.

[0240] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for generating collaborative decision-making criteria for unmanned aerial vehicles according to any of the above embodiments is implemented.

[0241] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0242] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0243] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0244] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0245] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or 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, including several instructions for a computer device (which can be a personal computer, server, or network equipment, etc.) to perform all or part of the steps of the methods of 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 codes.

[0246] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. 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 embodiments of the present invention.

Claims

1. A method for generating collaborative decision-making criteria for unmanned aerial vehicles, characterized in that: include: When decision criterion trigger data is received, the mission data of the UAV and the corresponding mission emergency status data are obtained; When the mission emergency status data is not urgent, the surrounding environment image data corresponding to the UAV is used to identify the opponent's equipment to obtain the number of opponent's equipment and the opponent's equipment feature data; When the number of the opponent's equipment is less than a preset value, extracting the UAV equipment feature data corresponding to the UAV and the ground data collected by the ground control station; rasterizing the opponent equipment feature data, the UAV equipment feature data and the ground data to construct an environmental situation grid image; Inputting the environmental situation grid image into a multi-cooperative decision-making criterion classification model to obtain a cooperative decision-making criterion corresponding to the UAV; The step of rasterizing the opponent equipment feature data, the UAV equipment feature data and the ground data to construct an environmental situation grid image comprises: Performing feature fusion on the opponent equipment feature data, the UAV equipment feature data and the ground data to obtain an initial real situation grid image set; Rasterizing the initial true situation grid image set to obtain a target true situation grid image set; According to the target real situation grid image set, an upsampling algorithm SMOTE is used to generate samples to obtain a first virtual situation grid image set; According to the target real situation grid image set, using the information diffusion method MTD to generate samples, to obtain a second virtual situation grid image set; Using the first virtual situation grid image set and the second virtual situation grid image set to construct a virtual situation grid image set; Grid image samples are selected from the target real situation grid image set and the virtual situation grid image set to construct an environment situation grid image.

2. The method for generating UAV collaborative decision-making criteria according to claim 1, characterized in that: The step of obtaining the mission data of the drone and the corresponding mission emergency status data includes: Get the drone authorization status data corresponding to the drone; When the authorization status data of the drone is authorized, emergency status judgment is performed on the mission data of the drone to obtain mission emergency status data.

3. The method for generating UAV collaborative decision-making criteria according to claim 1, characterized in that: The step of generating samples by using an upsampling algorithm SMOTE according to the target real situation grid image set to obtain a first virtual situation grid image set comprises: Using the equipment in the target real situation grid image set as first sample generation objects respectively to obtain a first grid image sample set; Selecting a root sample of a synthetic sample from the first grid image sample set according to a preset selection standard; Selecting a plurality of first sample generation objects whose Euclidean distances to the root sample are within a preset distance range from the first grid image sample set to obtain a neighbor auxiliary sample point set; Selecting auxiliary sample points from the neighboring auxiliary sample point set according to a preset upsampling rate to obtain an auxiliary sample point set; Substituting the auxiliary sample points in the auxiliary sample point set into the preset sample formula to perform sample update and generate virtual samples; The preset sample formula is: x new =x i +(x′ i -x i )×rand(0,1); Among them, x new is the first virtual situation grid image; x i is the root sample; x′ i is an auxiliary sample point; rand(0,1) is a random function between 0 and 1; All of the virtual samples are used to construct a first virtual situation grid image set.

4. The method for generating UAV collaborative decision-making criteria according to claim 1, characterized in that: The step of generating samples by using the information diffusion method MTD according to the target real situation grid image set to obtain the second virtual situation grid image set comprises: Using the equipment in the target real situation grid image set as second sample generation objects respectively to obtain a second grid image sample set; Select the maximum value and the minimum value corresponding to each equipment feature in the second grid image sample set respectively to obtain the maximum value and the minimum value of the feature; Calculate the average value between the maximum value of the feature and the minimum value of the feature to obtain the feature mean; Calculate the number of equipment with a value less than the characteristic mean in the second grid image sample set to obtain a first number of equipment; Calculate the number of devices with a value greater than the feature mean in the second grid image sample set to obtain a second number of devices; Substituting the first equipment number, the second equipment number and the characteristic mean into a preset boundary value calculation formula to perform boundary value calculation, and obtaining an upper boundary value and a lower boundary value corresponding to the equipment characteristic; The preset boundary value calculation formula is: Among them, UB is the upper boundary value; CL is the characteristic mean; Skew U is the right skewness based on the feature mean; is the equipment variance corresponding to the second grid image sample set; N U is the number of second equipment; LB is the lower boundary value; Skew L is the left skewness based on the feature mean; N L is the number of the first equipment; x j is the observed value; is the mean; n is the number of devices corresponding to the second grid image sample set; Substituting the upper boundary value and the lower boundary value into a preset triangular membership function value calculation formula to calculate and obtain a triangular membership function value corresponding to the equipment feature; The preset triangular membership function value calculation formula is: Among them, MF is the triangular membership function value; x′ is a virtual sample; LB is the lower boundary value; CL is the feature mean; UB is the upper boundary value; The triangular membership function value greater than the preset function value is used as a virtual sample of equipment; A second virtual situation grid image set is constructed using all the virtual samples of the equipment.

5. The method for generating UAV collaborative decision-making criteria according to claim 1, characterized in that: The step of selecting grid image samples from the target real situation grid image set and the virtual situation grid image set to construct an environment situation grid image comprises: According to a preset equipment type selection standard, selecting equipment types from the target real situation grid image set to obtain equipment types; According to the equipment type and the preset selection quantity, multiple supplementary samples are selected from the virtual situation grid image set to obtain a supplementary sample set corresponding to the equipment type and count the number of sample supplements in real time; When the number of sample supplements is less than a preset number threshold, the step of selecting equipment types from the target real situation grid image set according to a preset equipment type selection standard to obtain equipment types is executed; When the number of sample supplementation times is equal to a preset number threshold, all supplementary sample sets corresponding to the current moment are used to construct an environmental situation grid image.

6. The method for generating UAV collaborative decision-making criteria according to claim 1, characterized in that: The step of inputting the environmental situation grid image into a multi-cooperative decision-making criterion classification model to obtain a collaborative decision-making criterion corresponding to the UAV includes: Experts label the environmental situation grid image according to preset labels to obtain category labels corresponding to the environmental situation grid image; Taking the environmental situation grid image as input and the category label as output, a preset multi-classification neural network model is trained to obtain a multi-cooperative decision-making criterion classification model; The environmental situation grid image is input into the multi-cooperative decision-making criterion classification model to obtain the cooperative decision-making criterion corresponding to the UAV.

7. A UAV collaborative decision-making criterion generation system, characterized in that: include: A mission emergency data acquisition module, used to acquire the mission data of the UAV and the corresponding mission emergency data when receiving the decision criterion trigger data; A module for obtaining the number of opponent equipment and the characteristic data of opponent equipment, which is used to identify the opponent equipment using the surrounding environment image data corresponding to the UAV when the mission emergency status data is not urgent, and obtain the number of opponent equipment and the characteristic data of opponent equipment; The UAV equipment characteristic data and ground data extraction module is used to extract the UAV equipment characteristic data corresponding to the UAV and the ground data collected by the ground control station when the number of the opponent's equipment is less than a preset value; An environment situation grid image acquisition module is used to rasterize the opponent equipment feature data, the UAV equipment feature data and the ground data to construct an environment situation grid image; A collaborative decision criterion obtaining module is used to input the environmental situation grid image into a multi-collaborative decision criterion classification model to obtain a collaborative decision criterion corresponding to the UAV; The environment situation grid image obtaining module includes: An initial real situation grid image set acquisition module is used to perform feature fusion on the opponent's equipment feature data, the UAV equipment feature data and the ground data to obtain an initial real situation grid image set; A target real situation grid image set obtaining module is used to rasterize the initial real situation grid image set to obtain a target real situation grid image set; A first virtual situation grid image set obtaining module is used to generate samples using an upsampling algorithm SMOTE according to the target real situation grid image set to obtain a first virtual situation grid image set; A second virtual situation grid image set obtaining module is used to generate samples using the information diffusion method MTD according to the target real situation grid image set to obtain the second virtual situation grid image set; A virtual situation grid image set construction module, used to construct a virtual situation grid image set using the first virtual situation grid image set and the second virtual situation grid image set; The environment situation grid image acquisition submodule is used to select grid image samples from the target real situation grid image set and the virtual situation grid image set to construct the environment situation grid image.

8. An electronic device, characterized in that: It includes a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the method for generating drone collaborative decision criteria as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method for generating collaborative decision-making criteria for unmanned aerial vehicles as described in any one of claims 1 to 6 is implemented.

Images