Multi-machine task dynamic planning method and system, electronic equipment and storage medium

Abstract

The invention discloses a multi-machine task dynamic planning method and system, electronic equipment and a storage medium, and the method comprises the steps: a ground station receives the state information of each reconnaissance front end and battlefield environment information, and transmits the obtained data to each reconnaissance front end; the ground station decomposes an area to be reconnaissance into a plurality of sub-areas, obtains the positions of all current reconnaissance front ends, extracts the entrance points of the reconnaissance paths of the sub-areas, finds the reconnaissance front end closest to the entrance point for each entrance point, and distributes the reconnaissance front end to the sub-area corresponding to the entrance point; repeating the process until the reconnaissance front ends are distributed to all the sub-regions; the method has the advantages that distributed task allocation of the unmanned aerial vehicle is realized, and the number of reconnaissance task areas is as large as possible.

Description

Technical Field

[0001] This invention relates to the field of command and control systems, and specifically to a multi-machine task dynamic planning method, system, electronic device, and storage medium. Background Technology

[0002] Intelligent multi-drone mission dynamic planning technology refers to an unmanned technology that uses intelligent algorithms to plan and coordinate the missions of multiple drones. It features a highly autonomous and intelligent drone system capable of autonomously adjusting its flight trajectory, mission plan, and collaborative strategies according to mission requirements and environmental changes, achieving efficient and precise mission execution.

[0003] Intelligent multi-drone mission dynamic planning technology typically involves multiple unmanned aerial vehicles (UAVs), each with its own independent flight control and sensor systems, capable of independently completing specific tasks. These UAVs also possess robust data transmission and communication capabilities. Intelligent algorithms are key to achieving multi-drone collaboration in this technology. Based on advanced technologies such as machine learning and deep learning, these algorithms automatically generate mission plans, path plans, and collaborative strategies by analyzing mission requirements, environmental information, and sensor data, enabling collaborative control and mission execution among the multiple UAVs. While significant progress has been made in multi-UAV collaborative control systems, including task allocation and route planning, some challenges remain to be addressed. 1. Research on centralized systems for task allocation is more extensive than on distributed systems. For example, the centralized UAV selection method for centralized UAV task planning disclosed in Chinese Patent Publication No. CN111949047A is a type of centralized task allocation. Although centralized allocation systems can formulate detailed combat plans for UAVs before combat, unexpected situations often arise during mission execution, requiring UAVs to update or correct the task allocation results in real time to better adapt to the battlefield environment. Therefore, research on distributed systems is urgently needed.

[0004] 2. Insufficient analysis of mission characteristics in multi-UAV collaborative systems; Due to the special nature of missions, the focus of research should differ for different types of missions, but currently there is still a lack of specific analysis of particular missions. For example, when studying the problem of multi-UAV collaborative reconnaissance in multiple mission areas, it is necessary to start from the actual battlefield and not only consider common issues such as mission allocation and route planning, but also analyze the characteristics of reconnaissance missions. For example, what is the purpose of the UAVs performing reconnaissance missions, and how should the resources carried by the UAVs be allocated to maximize the reconnaissance mission area? These are all issues that cannot be ignored when UAVs perform reconnaissance missions. Summary of the Invention

[0005] The technical problem to be solved by this invention is how to achieve distributed task allocation for UAVs and maximize the reconnaissance task area.

[0006] The present invention solves the above-mentioned technical problems through the following technical means: a multi-aircraft mission dynamic planning method, comprising: a ground station receiving status information and battlefield environment information of each reconnaissance front-end, and transmitting the obtained data to each reconnaissance front-end; the ground station decomposes the area to be reconnoitered into multiple sub-regions, obtains the current positions of all reconnaissance front-ends, extracts the entry point of the reconnaissance path of each sub-region, and for each entry point, finds the reconnaissance front-end closest to the entry point and assigns it to the sub-region corresponding to the entry point, and repeats the above process until reconnaissance front-ends are assigned to all sub-regions.

[0007] Furthermore, the process of decomposing the area to be reconnoitered into multiple areas includes: The area to be investigated is a polygon, with the vertices of the polygon as the starting point. vi Starting from a point, find a point on the edge of the polygon. vd , center of mass c.p. , vi , vd as well as vi and vd The vertices of the polygon between them are used as the vertices of the sub-region, and the area of ​​the sub-region is... Sd and Sub The absolute error is less than d .

[0008] Furthermore, vd The method for determining the location of an edge is as follows: the region to be investigated is described as a polygon with p vertices, and the set of vertices of the polygon is { v 1, v 2, ..., vp};Will vi As the starting point of the sub-region, it is regarded as the left boundary. vleft The next vertex of the polygon after the starting point is considered the right boundary. right , and then right As vd calculate Sd ,like Sd Greater than Ssub+d This means vd In vleft and right On the edge of the vertex; otherwise, vleft and right Update each vertex to its next polygon vertex and continue the above process until a vertex is found. vd The edge where it is located vd For a vertex of a subregion, Sub This represents the average area of ​​each sub-region. dThis indicates the allowable error for the area of ​​the sub-region.

[0009] Furthermore, vd The method for determining the location is: calculation. vleft and right midpoint coordinates ( xm, ymm ), taking the midpoint as vd, Calculate the area of ​​the sub-region Sd ,like Sd in and Ssub-d and Ssub+d Between, the midpoint is vd ;like Sd Greater than Sub + d This means vd exist vleft Between and the midpoint, right After updating to the midpoint, continue the above operations; if Sd Less than Ssub-d ,illustrate vd At the midpoint and right Between, vleft After updating to the midpoint, continue the above operations.

[0010] Furthermore, the reconnaissance front-ends corresponding to each sub-region use a round-trip reconnaissance path planning algorithm to plan reconnaissance paths. The specific process is as follows: The sub-region is a convex polygon; the construction direction of the reconnaissance path is set to the longest edge of the sub-region. it In parallel directions, according to the given order of the vertices of the sub-regions, determine the longest edge. it The starting and ending vertices are respectively vs. , vnext ( s );distance it The farthest vertex is vf , vf arrive it The vertical distance is dfs ; use size Lx×Ly The scanning rectangle covers the sub-region along a straight line; the covered area is called the coverage rectangle, representing the area covered by the sensor's field of view on this path segment; the spacing between the coverage rectangles... dc equal Lx — ovx , ovx Indicates horizontal coverage; number of covered rectangles , Indicates rounding up; Through formula renew ovx Through formula renew dc ; Length of the covering rectangle Parallel to it straight line Sure, and it distance , This is the number of covering rectangles that have been constructed so far; let... The intersection points with the left and right boundaries are respectively and The angles between the left and right boundary line segments and the scanning direction are respectively and ,but ; After the current covering rectangle is constructed, Towards vf Move in the direction of dc, It continues to construct new coverage rectangles until the sub-region is fully covered. The scanning rectangle moves within the coverage rectangle along the construction direction, and the trajectory of the center point of the scanning rectangle is the track of the reconnaissance front end.

[0011] Furthermore, each reconnaissance front-end identifies targets within its respective reconnaissance sub-region. The specific process is as follows: The Darknet-53 in the YOLOv3 model is replaced with the backbone of MobileNet, while other structures remain unchanged, thus forming a target recognition network. This network is useful for identifying target categories in images collected by the reconnaissance front-end. During training, the target recognition network is used as a generator. Training samples are input into the generator, and the discriminator distinguishes between the generator's generated results and the real results. Training stops when the generator's loss function is minimized or when the preset number of iterations is reached, resulting in a trained target recognition network. This trained network is then used to identify the properties of targets.

[0012] Furthermore, after each reconnaissance front-end identifies the target, it also plans the target engagement sequence. The specific process is as follows: The K-means clustering algorithm is used to cluster the targets, resulting in multiple target groups. The Hungarian algorithm is then used to assign these target groups to different fire units. After target group assignment, reconnaissance and strike operations are conducted on the discretely distributed targets within each target group. The strike order planning is then solved using the Traveling Salesman Problem. The shortest paths through each target obtained from this solution form the initial strike order scheme. Finally, the target closest to a fire unit within each target group is identified and designated as the first strike target. Tfirst ; Calculate according to the initial order of attack. Tfirst Distance to the next target dtn and distance to the previous target dtp ,like dtn Less than dtp If the initial attack order is correct, then the direction of the initial attack order will be used as the final attack order; otherwise, the direction of the initial attack order will be reversed and used as the final attack order.

[0013] This invention also provides a multi-aircraft mission dynamic planning system, including multiple reconnaissance front-ends, a ground station, a communication relay, and a command post terminal server. The multiple reconnaissance front-ends are wirelessly connected to the ground station, the ground station is wirelessly connected to the command post communication radio via the communication relay, the ground station is wiredly connected to the unit communication radio, the unit communication radio is wirelessly connected to the command post communication radio, and the command post communication radio is wiredly connected to the command post terminal server. The ground station executes the aforementioned multi-aircraft mission dynamic planning method.

[0014] The present invention also provides an electronic device, including a processor and a memory, wherein the memory stores computer program instructions that can be executed by the processor, and when the processor executes the computer program instructions, it implements the steps of the method described above.

[0015] The present invention also provides a computer-readable storage medium storing computer program instructions, which, when called and executed by a processor, implement the steps of the method described above.

[0016] The advantages of this invention are: (1) This invention decomposes the area to be reconnoitered into multiple sub-regions and assigns reconnaissance front-ends to all sub-regions, realizing distributed task allocation for UAVs. This allows UAVs to update or correct task allocation results in real time according to the situation in order to better adapt to the battlefield environment. At the same time, for each sub-region, the entry point of the reconnaissance path of each sub-region is extracted. For each entry point, the reconnaissance front-end closest to the entry point is found and assigned to the sub-region corresponding to the entry point. This allows all reconnaissance front-ends to be reasonably allocated according to the principle of proximity, thereby maximizing the number of reconnaissance areas.

[0017] (2) This invention designs and applies an "M"-shaped intelligent multi-drone mission dynamic planning technology for multi-drone scheduling and collaborative reconnaissance. Before the UAVs take off to carry out battlefield reconnaissance and monitoring missions, the missions of each UAV are analyzed and planned according to the battlefield environment and actual conditions. The reconnaissance capability of a single UAV is limited, especially when the area to be reconnoitered is large, which will take a long time. In order to reduce the reconnaissance time, the area to be reconnoitered is first decomposed into multiple areas, and then the reconnaissance route is planned for each sub-area using a round-trip reconnaissance path planning algorithm. Finally, a greedy algorithm is used to arrange multiple UAVs to simultaneously reconnoiter each sub-target area. Effectively coordinating multiple UAVs to reconnoiter multiple mission areas maximizes the benefits of UAVs for multiple mission areas to be reconnoitered.

[0018] (3) After completing the reconnaissance mission across the entire battlefield, the reconnaissance front-end of this invention uses the YOLOv3 model to identify battlefield targets and marks the nature, location, and status of targets across the entire battlefield on an electronic map. This enhances the autonomy and stability of target detection, identification, and tracking technologies, strengthens the UAV's maneuverability and attack capabilities, improves weapon delivery accuracy, increases launch opportunities, and reduces weapon delivery time and operator workload, thereby contributing to improving the survivability and effective combat performance of UAVs. The algorithm replaces Darknet-53 in YOLOv3 with the backbone of MobileNet, while keeping other details unchanged, thus reducing computational costs and achieving sufficiently fast response speeds and accurate detection results.

[0019] (4) This invention uses the K-means clustering algorithm to cluster the targets to obtain multiple target groups. The Hungarian algorithm is used to assign the multiple target groups to different fire units. After the target groups are assigned, a target attack order scheme is formulated within the target groups. Then, in order to shorten the total attack distance, the target closest to the fire unit is found within the target group and is used as the first target to be attacked. Tfirst ; Calculate according to the initial order of attack. Tfirst Distance to the next target dtn and distance to the previous target dtp ,like dtn Less than dtp If the initial attack order is correct, then the direction of the initial attack order will be used as the final attack order; otherwise, the direction of the initial attack order will be reversed and used as the final attack order. Attached Figure Description

[0020] Figure 1 This is a schematic diagram illustrating the intelligent multi-UAV collaborative reconnaissance mission decision-making problem in a multi-UAV dynamic programming method disclosed in Embodiment 1 of the present invention; Figure 2This is a schematic diagram of the centralized control structure of a multi-UAV collaborative mission planning system in a multi-UAV mission dynamic planning method disclosed in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the round-trip reconnaissance path planning algorithm in a multi-machine task dynamic planning method disclosed in Embodiment 1 of the present invention; Figure 4 This is a system interface diagram of UAV aerial photography footage in a multi-aircraft task dynamic planning method disclosed in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the interface for flight altitude, distance, position, and other information in a multi-aircraft mission dynamic planning method disclosed in Embodiment 1 of the present invention. Figure 6 This is a schematic diagram of the architecture of a multi-machine task dynamic planning system disclosed in Embodiment 2 of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1 like Figure 2 As shown, Embodiment 1 of the present invention provides a multi-machine task dynamic planning method, the specific process of which is as follows: The intelligent multi-UAV collaborative reconnaissance mission decision-making problem refers to: given the performance and reconnaissance resources of each UAV, how to effectively coordinate multiple UAVs to conduct reconnaissance of multiple mission areas, ultimately obtaining the mission set, sequence, and corresponding time for each mission area for each UAV, so as to maximize the benefits of the UAVs for multiple reconnaissance mission areas. A specific problem description is as follows: Figure 1 As shown.

[0023] In the dynamically changing battlefield environment, the control structure of a multi-UAV collaborative mission planning system is one of the important factors affecting the planning effect. Its level of intelligence determines the reaction speed, responsiveness, and flexibility of the entire decision-making system. The selection of the control structure should meet the following requirements: high reliability and dynamic reconfiguration capability. This study allows for flexible assembly and modification of the system's control structure based on mission requirements and environmental changes. The control structures of the multi-UAV collaborative mission planning system can be divided into the following three types: 1) Distributed control structure: one main ground station controls multiple sub-ground stations, and one sub-ground station controls one UAV. 2) Airborne node control structure: one ground station controls one UAV, and one UAV controls multiple other UAVs. 3) Centralized control structure: one ground station controls multiple UAVs. Actual testing shows that the communication resource load rates of the three structures are basically similar with no fundamental difference, but the centralized control structure has significantly better reliability. The other two control structures are unreliable because excessive communication bridging and cascading amplify the probability of signal failure. Considering the reliability requirements of UAVs, this invention adopts a centralized control structure. Figure 2 As shown, in a centralized control structure, all decisions are made by the central control node, which is the ground station. The central control node then receives the status information of each UAV and the battlefield environment information, and through the analysis and calculation of the node, it transmits the obtained data to each UAV.

[0024] Once the hardware architecture is determined, this invention begins dynamic route planning for multi-aircraft reconnaissance missions. Specifically, the ground station receives status information and battlefield environment information from each reconnaissance front-end (in this embodiment, the reconnaissance front-end is an unmanned aerial vehicle (UAV); in actual applications, it could be other equipment such as a deep-space exploration balloon carrying relevant sensors. For convenience, the reconnaissance front-end will be referred to as an UAV). The ground station downloads the obtained data to each UAV; the ground station decomposes the area to be reconnaissance into multiple sub-regions, obtains the current positions of all UAVs, extracts the entry point of the reconnaissance path for each sub-region, finds the UAV closest to each entry point, and assigns it to the sub-region corresponding to that entry point. This process is repeated until UAVs are assigned to all sub-regions; the UAVs corresponding to each sub-region use a round-trip reconnaissance path planning algorithm to plan their reconnaissance paths.

[0025] Among them, such as Figure 3 As shown, the UAVs corresponding to each sub-region use a round-trip reconnaissance path planning algorithm to plan their reconnaissance paths, including: The sub-region is a convex polygon; the construction direction of the reconnaissance path is set to the longest edge of the sub-region. it In parallel directions, according to the given order of the vertices of the sub-regions, determine the longest edge. it The starting and ending vertices are respectively vs. , vnext ( s );distance it The farthest vertex is vf , vf arrive it The vertical distance is dfs ; use size Lx×Ly The scanning rectangle covers the sub-region along a straight line; the covered area is called the coverage rectangle, representing the area covered by the sensor's field of view on this path segment; the spacing between the coverage rectangles... dc equal Lx — ovx , ovx Indicates horizontal coverage; number of covered rectangles , This indicates rounding up; during UAV reconnaissance, sensors detect in real time, so only horizontal coverage needs to be considered when constructing the coverage path. ovx And without having to consider vertical coverage ovy .

[0026] because n.c. The data will be rounded up to ensure that the area covered by the last coverage rectangle outside the area to be detected is minimized. Each calculation needs to be performed separately. dc and ovx Through the formula renew ovx Through formula renew dc ; Length of the covering rectangle Parallel to it straight line Sure, and it distance , This is the number of covering rectangles that have been constructed so far; let... The intersection points with the left and right boundaries are respectively and The angles between the left and right boundary line segments and the scanning direction are respectively and ,but ; After the current covering rectangle is constructed, Towards vf Move in the direction of dc, It continues to construct new coverage rectangles until the sub-region is fully covered. The scanning rectangle moves within the coverage rectangle along the construction direction, and the trajectory of the center point of the scanning rectangle is the drone's flight path.

[0027] The reconnaissance capability of a single UAV is limited, especially when the area to be reconnoitered is large, which will result in a long reconnaissance time. To reduce reconnaissance time, the area to be reconnoitered is first decomposed into multiple regions. Then, the reconnaissance route planning algorithm described above is used to plan reconnaissance routes for each sub-region. Finally, a greedy algorithm is used to arrange multiple UAVs to simultaneously reconnoiter each sub-target area. The process of decomposing the area to be reconnoitered into multiple regions is as follows: The area to be investigated is a polygon, with the vertices of the polygon as the starting point. vi Starting from a point, find a point on the edge of the polygon. vd , center of mass c.p. , vi , vd as well as vi and vd The vertices of the polygon between them are used as the vertices of the sub-region, and the area of ​​the sub-region is... Sd and Sub The absolute error is less than d The key to decomposition lies in determining vd The location is determined in two steps: the first step is to determine vd The second step is to determine the edge where it is located. vd The precise location.

[0028] Sure vd The method of defining the region to be investigated is as follows: the region to be investigated is described as a polygon with p vertices, and the set of vertices of the polygon is { v 1, v 2, ..., vp};Will vi As the starting point of the sub-region, it is regarded as the left boundary. vleft The next vertex of the polygon after the starting point is considered the right boundary. right , and then right As vd calculate Sd ,like Sd Greater than Ssub+d This means vd In vleft and right On the edge of the vertex; otherwise, vleft and right Update each vertex to its next polygon vertex and continue the above process until a vertex is found. vd The edge where it is located vd For a vertex of a subregion, Sub This represents the average area of ​​each sub-region. d Indicates the allowable error for the area of ​​the sub-region; Sure vd The precise location method is: calculation vleft and rightmidpoint coordinates ( xm,ym ), taking the midpoint as vd, Calculate the area of ​​the sub-region Sd ,like Sd in and Ssub-d and Ssub+d Between, the midpoint is vd ;like Sd Greater than Sub + d This means vd exist vleft Between and the midpoint, right After updating to the midpoint, continue the above operations; if Sd Less than Ssub-d ,illustrate vd At the midpoint and right Between, vleft After updating to the midpoint, continue the above operations.

[0029] By optimizing multi-aircraft flight paths for full coverage, comprehensive and efficient coverage of the operational target area can be achieved. This provides more accurate and comprehensive battlefield intelligence and situational awareness for multi-aircraft collaborative operations, helping to improve decision-making efficiency and effectiveness. Secondly, real-time push of planned cruise paths to UAVs connected to the system enables multi-aircraft collaborative flight, improving the collaborative combat capabilities and efficiency of UAVs. By pushing cruise paths in real time, UAVs can quickly respond to the actions of other UAVs, better coordinate actions, avoid collisions and conflicts, and improve the safety of collaborative flight. Finally, acquiring flight data provides more accurate and real-time data support for multi-aircraft collaborative battlefield situational awareness technology. By analyzing flight data, real-time information such as the position, speed, and heading of each UAV, as well as data such as the relative positions and distances between UAVs, can be obtained, thereby better grasping the battlefield situation and changing trends, providing decision-makers with more accurate and timely battlefield intelligence and situational awareness.

[0030] This invention monitors the parameters and mission completion status of each UAV during multi-UAV collaborative missions and displays them on a ground station. The monitored objects include: UAV aerial footage, flight altitude, flight distance, flight position information, satellite navigation signals, aircraft damage status, mission completion rate, battery status, target coordinate positioning accuracy, etc. Figure 4 The image shown is a system interface diagram of drone aerial footage, as follows: Figure 5The diagram shows an interface displaying flight altitude, distance, and location information. The update frequency of the monitoring data significantly impacts the load on the communication link. Testing revealed that a monitoring data update frequency of 20Hz meets practical application requirements in terms of both data refresh rate and link load. By adjusting the Kalman filter algorithm parameters, the target positioning accuracy error at different data refresh frequencies was verified. Testing showed that a coordinate calculation refresh frequency of 10Hz resulted in a coordinate accuracy error within 5m, indicating an optimal solution for coordinate data update delay and coordinate error.

[0031] Target detection and identification play a crucial role in battlefield surveillance and reconnaissance. Deciphering the identity and location information of targets from remote sensing images is of great value for command and decision support and precision strikes. In the future intelligent combat mode, the system can identify abnormal targets through artificial intelligence technology, thereby assisting commanders in decision-making.

[0032] Deep learning technology enables the automatic segmentation and recognition of targets such as people, equipment, and geographical environments in video footage. In real-world environments, image quality is often affected by natural geographical factors such as lighting, fog, and occlusion. To address these issues, this system employs an improved YOLO regression model for target detection. Trained on millions of samples of people, equipment, and geographical environment targets, a neural network classifier is obtained, reliably and accurately detecting targets in environmental images and acquiring target-related information. The improved YOLO regression model replaces the Darknet-53 in the YOLOv3 model with the MobileNet backbone, while keeping other structures unchanged, thus forming a target recognition network. This network can then classify targets with labeled coordinates. During training, the target recognition network acts as a generator. Training samples are input into the generator, and a discriminator distinguishes the generator's output from the actual output. Training stops when the generator's loss function is minimized or when a preset number of iterations is reached, resulting in a trained target recognition network. This network is then used to identify the properties of targets. The generator's loss function is existing technology and will not be elaborated upon here. In this embodiment, the target tracking algorithm employs a normalized cross-correlation matching algorithm and an affine transformation image matching algorithm. After inputting the target coordinates, it can continuously and stably track the target quickly and output the target position information in real time according to the video rate, without any frame dropping. Both the normalized cross-correlation matching algorithm and the affine transformation image matching algorithm are existing technologies and will not be described in detail here.

[0033] After target identification, the targets are clustered using the K-means clustering algorithm to obtain multiple target groups. The K-means clustering algorithm is an existing technology, and its main steps are as follows: (1) Randomly select from all targets. KOne target is set as the initial cluster center for each cluster; (2) Calculate the distance to each target separately. K The Euclidean distance between the cluster centers is used to determine the category of each target and then the category of the target is set to the category corresponding to its nearest cluster center. (3) Calculate the clustering objective function value, and calculate the values ​​of the clustering objective function values ​​respectively. K The cluster center of each class is used as the new cluster center for that class; (4) If the cluster center results do not change or the objective function value does not decrease, output the final result; otherwise, repeat steps (2) and (3).

[0034] After clustering the targets, all targets are divided into K= Nau The purpose of target group allocation is to... Nau Target groups are assigned to Nau The assignment of fire units is such that the total distance from each UAV to its corresponding cluster center is minimized. The Hungarian algorithm, a method for calculating maximum matching in a bipartite graph using augmenting paths, is used to solve the assignment problem and can address the issue of assigning target groups to fire units in this invention. Distance cost is used to construct... Nau × Nau Given the cost matrix, the steps for target group allocation using the Hungarian algorithm are as follows: (1) Find the minimum value of each column in the cost matrix and subtract the minimum value of the current column from each column element; (2) Find the minimum value of each row in the cost matrix, and subtract the minimum value of the current row from each element of the current row; (3) Cover the zeros in the cost matrix with the fewest straight lines in the rows or columns; (4) Determine whether the number of covered lines is equal to Nau If so, find the allocation scheme with the minimum distance cost and output the final result. Otherwise, proceed to step (5). (5) Find the minimum value among the uncovered elements, subtract this minimum value from each row, and then add it to each column. Skip to step (3) to continue.

[0035] After target group allocation, reconnaissance and strike operations are required for multiple discretely distributed targets within the target group. At this stage, the strike order planning is solved using the Traveling Salesman Problem (TSP). The shortest access path through each target obtained from the solution forms the initial strike order scheme. Due to the characteristics of TSP solutions, starting from any target, visiting all targets in any direction, and returning to the starting point, the distance traveled is the same. To shorten the total strike distance, appropriate initial strike targets and strike directions can be selected based on these characteristics. The target closest to the fire unit within the target group is found and designated as the first strike target. Tfirst ; Calculate according to the initial order of attack. Tfirst Distance to the next target dtn and distance to the previous target dtp ,like dtn Less than dtp If the initial attack order is correct, the direction of the initial attack order will be used as the final attack order; otherwise, the direction of the initial attack order will be reversed and used as the final attack order. There are various methods to solve the TSP problem, including analytical methods, enumeration methods, dynamic programming methods, and genetic algorithms. Analytical methods often only yield local optima, enumeration methods are computationally inefficient, and dynamic programming suffers from exponential explosion. Genetic algorithms (GA) perform efficient heuristic searches simultaneously from multiple positions in the solution space, achieving high efficiency while preventing the search from converging to local optima. Therefore, this invention uses a genetic algorithm to solve the attack order planning problem. Genetic algorithms are existing technology and will not be elaborated upon here.

[0036] Through the above technical solutions, this invention designs and applies an "M"-shaped intelligent multi-aircraft mission dynamic planning technology for multi-aircraft scheduling and collaborative reconnaissance. Before the UAVs take off to carry out battlefield reconnaissance and monitoring missions, the missions of each UAV are analyzed and planned according to the battlefield environment and actual conditions.

[0037] Example 2 like Figure 6 As shown, the present invention also provides a multi-aircraft mission dynamic planning system, including multiple UAVs, a ground station, a communication relay, and a command post terminal server. The multiple UAVs are wirelessly connected to the ground station, the ground station is wirelessly connected to the command post communication radio via the communication relay, the ground station is wiredly connected to the squad communication radio, the squad communication radio is wirelessly connected to the command post communication radio, and the command post communication radio is wiredly connected to the command post terminal server. The ground station executes a multi-aircraft mission dynamic planning method as described in Embodiment 1.

[0038] Example 3 Embodiment 3 of the present invention provides an electronic device, including a processor and a memory. The memory stores computer program instructions that can be executed by the processor. When the processor executes the computer program instructions, it implements the steps of the method described in Embodiment 1.

[0039] Example 4 Embodiment 4 of the present invention provides a computer-readable storage medium storing computer program instructions, which, when called and executed by a processor, implement the steps of the method described in Embodiment 1.

[0040] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dynamic programming method for multi-machine tasks, characterized in that, include: The ground station receives status information and battlefield environment information from each reconnaissance front-end, and then transmits the obtained data to each reconnaissance front-end. The ground station decomposes the area to be reconnaissance into multiple sub-regions, obtains the current positions of all reconnaissance front-ends, extracts the entry points of the reconnaissance paths in each sub-region, finds the reconnaissance front-end closest to each entry point, and assigns it to the sub-region corresponding to that entry point. The above process is repeated until reconnaissance front-ends have been assigned to all sub-regions.

2. The multi-machine task dynamic programming method according to claim 1, characterized in that, The process of dividing the area to be investigated into multiple areas includes: The area to be investigated is a polygon, with the vertices of the polygon as the starting point. vi Starting from a point, find a point on the edge of the polygon. vd , center of mass cp , vi , vd as well as vi and vd The vertices of the polygon between them are used as the vertices of the sub-region, and the area of ​​the sub-region is... Sd and Ssub The absolute error is less than δ .

3. The multi-machine task dynamic programming method according to claim 2, characterized in that, vd The method for determining the location of an edge is as follows: the region to be investigated is described as a polygon with p vertices, and the set of vertices of the polygon is { v 1, v 2, ..., vp };Will vi As the starting point of the sub-region, it is regarded as the left boundary. vleft The next vertex of the polygon after the starting point is considered the right boundary. vright , and then vright As vd calculate Sd ,like Sd Greater than Ssub+δ This means vd In vleft and vright On the edge of the vertex; otherwise, vleft and vright Update each to its next polygon vertex and continue the above operation until a vertex is found. vd The edge where it is located vd For a vertex of a subregion, Ssub This represents the average area of ​​each sub-region. δ This indicates the allowable error for the area of ​​the sub-region.

4. The multi-machine task dynamic programming method according to claim 3, characterized in that, vd The method for determining the location is: calculation. vleft and vright midpoint coordinates ( xm, ym ), taking the midpoint as vd, Calculate the area of ​​the sub-region Sd ,like Sd in and Ssub-δ and Ssub+δ Between, the midpoint is vd ;like Sd Greater than Ssub + δ This means vd exist vleft Between and the midpoint, vright After updating to the midpoint, continue the above operations; if Sd Less than Ssub-δ ,illustrate vd At the midpoint and vright Between, vleft After updating to the midpoint, continue the above operations.

5. The multi-machine task dynamic programming method according to claim 4, characterized in that, The reconnaissance front-end corresponding to each sub-region uses a round-trip reconnaissance path planning algorithm to plan the reconnaissance path. The specific process is as follows: The sub-region is a convex polygon; the construction direction of the reconnaissance path is set to the longest edge of the sub-region. es In parallel directions, according to the given order of the vertices of the sub-regions, determine the longest edge. es The starting and ending vertices are respectively vs , vnext ( s );distance es The farthest vertex is vf , vf arrive es The vertical distance is dfs ; use size as Lx×Ly The scanning rectangle covers the sub-region along a straight line; the covered area is called the coverage rectangle, representing the area covered by the sensor's field of view on this path segment; the spacing between the coverage rectangles... DC equal Lx — ovx , ovx Indicates horizontal coverage; number of covered rectangles. , Indicates rounding up; Through formula renew ovx Through formula renew DC ; Length of the covering rectangle Parallel to es straight line Sure, and es distance , This is the number of covering rectangles that have been constructed so far; let... The intersection points with the left and right boundaries are respectively and The angles between the left and right boundary line segments and the scanning direction are respectively and ,but ; After the current covering rectangle is constructed, Towards vf Move in the direction of DC It continues to construct new coverage rectangles until the sub-region is fully covered. The scanning rectangle moves within the coverage rectangle along the construction direction, and the trajectory of the center point of the scanning rectangle is the track of the reconnaissance front end.

6. The multi-machine task dynamic programming method according to claim 1, characterized in that, Each reconnaissance front-end identifies targets within its respective reconnaissance sub-region. The specific process is as follows: The Darknet-53 in the YOLOv3 model is replaced with the backbone of MobileNet, while other structures remain unchanged, thus forming a target recognition network. This network is useful for identifying target categories in target images collected by the reconnaissance front-end. During training, the target recognition network is used as a generator. Training samples are input into the generator, and the discriminator distinguishes between the generator's generated results and the real results. Training stops when the generator's loss function is minimized or when the preset number of iterations is reached, resulting in a trained target recognition network. The trained target recognition network is then used to identify the properties of the target.

7. The multi-machine task dynamic programming method according to claim 6, characterized in that, After each reconnaissance front-end identifies the target, it also plans the target engagement sequence. The specific process is as follows: The K-means clustering algorithm is used to cluster the targets to obtain multiple target groups. The Hungarian algorithm is then used to assign the multiple target groups to different fire units. After target group allocation, reconnaissance and strike operations are required for multiple discretely distributed targets within the target group. At this point, the strike order planning is solved using the Traveling Salesman Problem. The shortest access paths obtained through each target are the preliminary strike order scheme. Then, the target closest to the fire unit within the target group is identified and designated as the first strike target. Tfirst ; Calculate according to the initial order of attack. Tfirst Distance to the next target dtn and distance to the previous target dtp ,like dtn Less than dtp If the initial attack order is correct, then the direction of the initial attack order will be used as the final attack order; otherwise, the direction of the initial attack order will be reversed and used as the final attack order.

8. A multi-machine task dynamic programming system, characterized in that, It includes multiple reconnaissance front-ends, ground stations, communication relays, and command post terminal servers. The multiple reconnaissance front-ends are wirelessly connected to the ground stations. The ground stations are wirelessly connected to the command post communication radios via the communication relays. The ground stations are wired to the unit communication radios. The unit communication radios are wirelessly connected to the command post communication radios. The command post communication radios are wired to the command post terminal servers. The ground stations execute a multi-aircraft mission dynamic planning method as described in any one of claims 1-7.

9. An electronic device, characterized in that, The method includes a processor and a memory, the memory storing computer program instructions executable by the processor, wherein when the processor executes the computer program instructions, it implements the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The device stores computer program instructions that, when invoked and executed by a processor, implement the steps of the method described in any one of claims 1-7.

Images