A method, system and device for unmanned aerial vehicle path planning

Abstract

The application discloses a kind of unmanned plane path planning method, system and equipment, it is related to unmanned plane path planning field, the method includes constructing to be planned scene model;The to-be-planned scene model is multiple unmanned plane to multiple target planning;The to-be-planned scene model includes: unmanned plane target distribution result and map environment information;The map environment information includes: target position, coordinate and threat area range;Maximum rotation angle constraint is determined according to the curvature radius constraint of unmanned plane path turning;According to the curvature radius constraint of unmanned plane path turning, maximum rotation angle constraint and threat area range, using RRT* algorithm, path planning is carried out to unmanned plane in to-be-planned scene model.The application can improve the efficiency and accuracy of unmanned plane path planning, and can make path to be able to adaptively cross threat area, more conform to actual situation.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of unmanned aerial vehicle path planning, and in particular to an unmanned aerial vehicle path planning method, system and device. BACKGROUND

[0002] In recent years, unmanned aerial vehicles have been widely used in disaster relief, photography and mobile edge computing. With the advent of the information age, the application of unmanned aerial vehicles has expanded from civilian use to modern warfare. In complex scenarios, unmanned aerial vehicle path planning is one of the most important steps in completing the entire task, as it determines the space-time information of the unmanned aerial vehicle. The purpose of path planning is to find a series of points from the starting point to the end point to ensure that the unmanned aerial vehicle does not collide with obstacles. Therefore, efficient and reliable path planning algorithms are worth studying.

[0003] Path planning algorithms can be divided into search-based, sampling-based, potential field-based and learning-based directions. Dijkstra's algorithm and A* algorithm are classic search-based algorithms, both of which can guarantee the optimization of the path. A* algorithm increases heuristic information on the basis of the former, greatly reducing the search range. The basic sampling-based algorithm includes probabilistic roadmap (PRM) and rapidly-exploring random tree (RRT), of which the biggest feature of RRT is rapidity. Artificial potential field method regards the target as gravity and the obstacle as repulsive force to establish a gravity field, thereby completing path planning. With the rise of artificial intelligence, reinforcement learning is applied to path planning. Samaneh uses reinforcement learning to achieve multi-agent path planning in dense environments.

[0004] In the above methods, the RRT algorithm has attracted widespread attention due to its rapidity and the need for discrete state space. People continue to propose various schemes from different angles to improve the RRT algorithm. For optimality, Sertac Karaman proposes RRT*, which is the first asymptotically optimal RRT. In the algorithm, a node that finds the optimal cost and a reconnection mechanism are proposed. Informed RRT* optimizes the sampling range on the basis of RRT* and improves the convergence speed of RRT*. In dynamic path planning, Olzhas proposes RRT*D, which preserves the useful part of the tree after the dynamic obstacle destroys the solution path. Considering the dynamics feasibility, QS-RRT algorithm ensures the smoothness of the path by adding a detection node in the RRT algorithm. However, QS-RRT is not asymptotically optimal like RRT. Kinodynamic RRT* proposed by Dustin restricts the feasible state space through the dynamics model of the object and the control angle. However, Kinodynamic RRT* needs to calculate the new feasible state space of the path, which increases the computational complexity.

[0005] Another problem is obstacle avoidance. In the past, most path planning algorithms regard obstacles as hard constraints. However, in actual application scenarios, there may be some areas that are obstacles, but not absolutely impassable, and need to be adjusted according to actual conditions.

[0006] To solve the above two problems, it is urgent to provide a new unmanned aerial vehicle path planning method or system to improve the efficiency and accuracy of unmanned aerial vehicle path planning, and enable the path to adaptively pass through the threat area, and be more in line with actual conditions. SUMMARY

[0007] The purpose of the present application is to provide an unmanned aerial vehicle path planning method, system and device, which can improve the efficiency and accuracy of unmanned aerial vehicle path planning, and enable the path to adaptively pass through the threat area, and be more in line with actual conditions.

[0008] To achieve the above purpose, the present application provides the following solutions:

[0009] An unmanned aerial vehicle path planning method, comprising:

[0010] Constructing a to-be-planned scene model; the to-be-planned scene model is multi-unmanned aerial vehicle path planning for multiple targets; the to-be-planned scene model comprises unmanned aerial vehicle target distribution results and map environment information; the map environment information comprises target positions, coordinates and threat area ranges;

[0011] Determining a maximum rotation angle constraint according to a curvature radius constraint of unmanned aerial vehicle path turning;

[0012] According to the curvature radius constraint of unmanned aerial vehicle path turning, the maximum rotation angle constraint and the threat area range, an RRT* algorithm is used to plan a path for the unmanned aerial vehicle in the to-be-planned scene model.

[0013] Optionally, the RRT* algorithm is used to plan a path for the unmanned aerial vehicle in the to-be-planned scene model according to the curvature radius constraint of unmanned aerial vehicle path turning, the maximum rotation angle constraint and the threat area range, specifically comprising:

[0014] Obtaining a starting position and a target position of a to-be-planned unmanned aerial vehicle in the to-be-planned scene model;

[0015] Using the curvature radius constraint of unmanned aerial vehicle path turning and the maximum rotation angle constraint to perform secondary screening on a neighbor set generated in the RRT* algorithm;

[0016] Determining a risk of passing through a threat area according to the threat area range; and adding the risk of passing through the threat area to a node cost function to obtain an optimized cost function;

[0017] According to the optimized cost function, the neighbor set after the secondary screening, the search tree and the map environment information, a new search tree after reconnection is obtained;

[0018] According to the new search tree after reconnection, a planning path is determined.

[0019] Optionally, the risk of crossing the threat area is determined according to the range of the threat area, and the risk of crossing the threat area is added to the cost function of the node to obtain the optimized cost function, and the previous also includes:

[0020] According to the start point, the end point and the map environment information of the path, a numerical sampling method is used to determine the total distance of crossing the threat area.

[0021] Optionally, the risk of crossing the threat area is determined according to the range of the threat area, and the risk of crossing the threat area is added to the cost function of the node to obtain the optimized cost function, and the previous also includes:

[0022] The formula J i = Tree(i).Prev.dist+cs_weight*dist_in_cs is used to determine the optimized cost function;

[0023] Wherein, J i represents the optimized cost function of node i, Tree(i).Prev.dist represents the cost of the parent node of node i in RRT*, cs_weight represents a parameter for controlling the aggressiveness and conservatism of the path, and dist_in_cs represents the distance between the connection path of the extended node and the corresponding parent node and the obstacle region.

[0024] An unmanned aerial vehicle path planning system, comprising:

[0025] A to-be-planned scene model construction module is configured to construct a to-be-planned scene model; the to-be-planned scene model is a multi-target planning for multiple unmanned aerial vehicles; the to-be-planned scene model includes an unmanned aerial vehicle target distribution result and map environment information; the map environment information includes target positions, coordinates and threat area ranges;

[0026] A constraint determination module is configured to determine a maximum rotation angle constraint according to a curvature radius constraint of an unmanned aerial vehicle path turning;

[0027] A path planning module is configured to use an RRT* algorithm to plan a path for an unmanned aerial vehicle in the to-be-planned scene model according to the curvature radius constraint of the unmanned aerial vehicle path turning, the maximum rotation angle constraint and the threat area range.

[0028] An unmanned aerial vehicle path planning device, comprising: at least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method.

[0029] Optionally, the memory is a computer-readable storage medium.

[0030] According to the specific embodiments provided by the application, the following technical effects are disclosed.

[0031] The unmanned aerial vehicle path planning method, system and device provided by the application analyze the maximum rotation angle constraint and the requirement that the unmanned aerial vehicle can flexibly pass through a threat area in a to-be-planned scene model, improve the reconnection process of the neighbor set in RRT*, and add the maximum rotation angle constraint in the RRT* framework, so that the generated path strictly satisfies the constraint. In addition, the above process also reduces the time consumption of RRT*, and the speed advantage of the algorithm provided by the application becomes more and more obvious with the increase of the iteration number. Then, the obstacle information is taken as one of the components of the node cost of the path, and the original obstacle is taken as a soft constraint. This makes the path be able to adaptively pass through the threat area, and is more in line with the actual situation. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0033] Figure 1 A flowchart of the unmanned aerial vehicle path planning method provided by the application is shown in the figure.

[0034] Figure 2 A schematic diagram of the to-be-planned scene model is shown in the figure.

[0035] Figure 3 A schematic diagram of the minimum turning radius is shown in the figure.

[0036] Figure 4 A schematic diagram of selecting a suitable path to pass through a threat area is shown in the figure.

[0037] Figure 5 A schematic diagram of screening the neighbor set is shown in the figure.

[0038] Figure 6 A schematic diagram of the reconnection process is shown in the figure.

[0039] Figure 7 A schematic diagram of the maximum rotation angle constraint simulation result is shown in the figure.

[0040] Figure 8 This is a diagram illustrating the increase in time consumption with each iteration.

[0041] Figure 9 This is a schematic diagram of the simulation results for adaptive traversal of the threat zone.

[0042] Figure 10 A schematic diagram of the simulation results for planning unmanned aerial vehicles (UAVs). Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and 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.

[0044] The purpose of this invention is to provide a method, system, and device for UAV path planning, which can improve the efficiency and accuracy of UAV path planning and enable the path to adaptively traverse threat areas, making it more in line with actual conditions.

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0046] like Figure 1 As shown, the UAV path planning method provided by the present invention includes:

[0047] S101, Construct a scenario model to be planned; the scenario model to be planned is a multi-target planning for multiple UAVs; the scenario model to be planned includes: UAV target distribution results and map environment information; the map environment information includes: target location, coordinates and threat zone range.

[0048] like Figure 2 As shown, the attacking side's weapons start from the origin and strike the target point. The weapon target distribution and map environment information (target location, coordinates, and threat zone range, etc.) are all known. All weapons and target points are considered as point masses, and the weapon strike range is d. hit When the distance d from the target satisfies d ≤ d hit At that time, the planning is complete. Because the planning scenario involves multiple weapons targeting multiple targets, based on actual needs, it is expected that each weapon will attack the same target randomly, meaning the planning result is not fixed.

[0049] S102, determine the maximum rotation angle constraint based on the curvature radius constraint of the UAV path turning.

[0050] S103, according to the curvature radius constraint of the unmanned aerial vehicle path turning, the maximum rotation angle constraint and the threat area range, the RRT* algorithm is adopted to plan the path of the unmanned aerial vehicle in the to-be-planned scene model.

[0051] The RRT* algorithm is as shown in Algorithm 1:

[0052]

[0053] In Algorithm 1, Map, p start start and p end end are input map environment information, unmanned aerial vehicle start point and end point position respectively; T is a search tree, and the structure of the search tree includes its position, index, father node position, father node index and node cost. p rand is a node obtained by random sampling in the map; p near ←Near(p rand ,T) is a node closest to p rand in the search tree T; p near ←Steer(p new ,p rand ,StepSize) is a node obtained by extending p near to p near by a specified step size StepSize; P rand ←NearSet(T,p new ,R nears ) is all nodes in the search tree T within a distance R new from p nears , which are taken as a neighbor set P new ; p nears is a node in the neighbor set P nears , and p min is a node with the minimum cost from the start point to p new , and p min is taken as the father node of p new .

[0054] In the planning, the curvature radius of the unmanned aerial vehicle path turning must satisfy R > R min . The curvature radius of the bending is as shown in FIG. 3. Figure 3

[0055] In fact, since the research object of the present application lacks an accurate model, in order to facilitate path planning, the minimum turning radius and the maximum rotation angle constraint can be mapped as: θ rotation =f(R min ). According to the physical principle, this is a negative correlation mapping, and the greater the included angle between two paths, the smaller the corresponding curvature radius. Therefore, the planning result can be directly strictly constrained by the rotation angle of the path point. ​

[0056] In the scenario established by the present application, in addition to the conventional obstacle area (such as a wall), there is also the concept of a threat area (for example, an area covered by radar). In actual requirements, the weapon needs to avoid the threat area as much as possible. However, when the threat area is completely covered or bypassed, the weapon should choose an appropriate path to pass through the threat area (such as the desired path shown in FIG. 1). Figure 4 However, the above RRT and RRT* only contain the definition of the hard constraint of the obstacle, and cannot achieve the function of adjusting the path according to the will and actual situation of the person.

[0057] S103 specifically includes:

[0058] S301, acquiring a starting point position and a target position of a unmanned aerial vehicle to be planned in a scene model to be planned.

[0059] S302, performing secondary screening on a neighbor set generated in the RRT* algorithm by using a curvature radius constraint and a maximum rotation angle constraint of a path turning of the unmanned aerial vehicle, and as shown in FIG. 2. Figure 5

[0060] The key of the RRT* to increase the maximum rotation angle constraint lies in the selection of the surrounding neighbor nodes and the pruning and reconnection of the random tree, that is, the corresponding optimization of the NearSet and Rewire functions. This step first screens the neighbor set in the RRT*.

[0061] The NearSet starts from the first extended node, selects the node with the lowest cost in the radius R nears circle to the starting point as the parent node, and all subsequent nodes follow this rule to ensure the optimality of the result. However, in this process, there are actually many nodes in the neighbor circle that do not meet the maximum rotation angle constraint. However, they will be added to the neighbor set P nears , such as nodes c and d in FIG. 5. In order to strictly guarantee the maximum rotation angle constraint, secondary selection is performed in the neighbor circle, and only the nodes that meet the condition θ rotation (p.Prev, p, p new )≤θ limit (p=a, b, c, d) are considered to be added to the new neighbor set. θ rotation (p.Prev, p, p new ) is defined as:

[0062]

[0063] where p.Prev represents the parent node of p. Therefore, in the subsequent process of adding p new to the random tree, the node p nears in P new ​The angle of the path formed by p and its parent must satisfy the maximum rotation angle constraint. A special case is that there is no node in the neighbor circle that satisfies the condition, so the neighbor set is empty. If the neighbor set is empty, delete the expanded node p new , and as shown in Algorithm 2.

[0064]

[0065] NearSet (T, p new , R nears ) is a function name NearSet, which determines the new neighbor set, the input is the search tree, p new , and the neighbor set radius, and the output is the neighbor set P nears , and the judgment flag of the existence of the neighbor set.

[0066] As shown in Figure 6 , the original tree satisfies the maximum rotation angle constraint everywhere, but after reconnection, θ rotation (b, p new , a) no longer satisfies the maximum rotation angle constraint. Therefore, in the rewiring process, it is necessary to judge the results of pruning and reconnection. If the new tree does not satisfy the maximum rotation angle constraint, even if the cost of reconnecting the node is smaller, the reconnection process will be abandoned, and the nodes in the other adjacent circles will be directly considered.

[0067] S303, according to the threat area range, determine the risk of crossing the threat area; and add the risk of crossing the threat area to the cost function of the node to obtain an optimized cost function; wherein, a suitable path crossing the threat area is shown in Figure 4 .

[0068] The function of adaptive crossing of the threat area is analyzed and improved. In the original RRT*, CollisionCheck is directly performed on the newly expanded node. If the connection path of the node and the random tree passes through the obstacle, the node is directly deleted, which cannot meet the requirements of adaptive traversal. In order to control the aggressiveness of the path through the threat area or bypass it directly through parameters, the definition of the threat area in the map and the definition of the cost function of each node need to be changed. First, modify the properties of the threat area in the algorithm, and regard it as a region that can be crossed but has a higher crossing cost than the free space. Therefore, when generating the random tree, define the optimized cost function of each node:

[0069] J i = Tree (i). Prev. dist + cs_weight * dist_in_cs.

[0070] Where J iThe `Tree(i).Prev.dist` represents the optimized cost function for node `i`, `Cs_weight` represents a parameter controlling the aggressiveness or conservatism of the path, and `dist_in_cs` represents the distance between the connection path of the extended node and its corresponding parent node and the obstacle region. A value close to 1 indicates the most aggressive approach. Larger values ​​indicate a more conservative path, with the threat region closer to the absolute obstacle. Therefore, after the above definitions, the above cost function must be used instead of the simple distance calculation when calculating node costs in the subsequent `ChooseParent` and `Rewire` sections.

[0071] Incorporating the maximum rotation angle constraint, the new Rewire function is as described in Algorithm 3.

[0072]

[0073]

[0074] Among them, Rewire(T,P) nears ,p new The function named Rewire (cs_weight, map) performs a reconnection that comprehensively considers the maximum turning angle constraint and the cost of the node threat zone. Inputs are a search tree, a neighbor set, and p. new The threat zone risk parameter cs_weight and map environment information are used; the output is the new search tree after reconnection.

[0075] S303 was preceded by:

[0076] Based on the starting point, ending point, and map environment information of the path, a numerical sampling method is used to determine the total distance dist_in_cs across the threat zone. The dist_in_cs uses the original logic for determining whether a collision has occurred to calculate the crossing distance, as shown in Algorithm 4.

[0077]

[0078] Among them, CollisonCheck(p start ,p end The function `CollisonCheck` (map) calculates the total length of a path traversing a threat zone (which is a soft constraint). The input includes the start and end points (endpoints) of a path segment, map environment information, and the output is the total distance traversed within the threat zone. Note: However, actual distance calculation is difficult; therefore, this function simulates a line segment using densely sampled points and calculates the number of sampled points within the threat zone. checkSampling the line segment from the start point to the end point according to a certain step length, and traversing the traversal points one by one;

[0079] S304, according to the optimized cost function, the neighbor set after secondary screening, the search tree and the map environment information, a new search tree with a new connection;

[0080] Based on the above-mentioned adding process of the maximum rotation angle constraint and the design of the adaptive traversal of the threat area, combined with the path planning framework of RRT*, the final MA-RRT* algorithm can be obtained, and is shown in Algorithm 5.

[0081]

[0082] S305, determining the planned path according to the new search tree after reconnection.

[0083] In order to verify the effectiveness of the above-mentioned MA-RRT* algorithm, relevant numerical simulation is carried out on the MATLAB 2021b simulation platform. The scene information contained in the simulation only contains the basic information of the weapon, the start point, the end point and the threat area information. In order to show the effect of the present application, it is assumed that there is only one threat area in the scene, and there is no other obstacle.

[0084] This section will show the simulation results from three aspects: MA-RRT* strictly satisfies the maximum rotation angle constraint, time cost and adaptive traversal of the threat area.

[0085] Figure 7 The planning effect of the MA-RRT* algorithm under a specific scene (θ limit = 60°) is shown, and the RRT* algorithm and the QS-RRT algorithm are compared. From the simulation results, it can be seen that the RRT* algorithm is easy to produce a larger corner in this scene, resulting in uneven path. Although QS-RRT limits the rotation angle, the algorithm is not optimal, and the path is easy to be too tortuous. However, the method proposed in the present application strictly restricts the rotation angle of the path based on the framework of RRT*, while ensuring the smoothness and gradualness of the generated path.

[0086] In addition, the present application also verifies the improvement of the time cost of the algorithm after adding the maximum rotation angle constraint. Table 1 shows the time cost comparison of RRT* and MA-RRT* of 10 unmanned aerial vehicles under 1000 iterations. The results show that the time cost of MA-RRT* is nearly 40% lower than that of RRT*.

[0087] Table 1

[0088]

[0089] The present application also statistics the time cost of MA-RRT* algorithm and RRT* algorithm with the growth of iteration number. As shown inFigure 8 As shown in the figure, the more the number of iterations, the more obvious the acceleration effect of MA-RRT* on RRT*.

[0090] Finally, the function of the algorithm to adaptively cross the threat area is verified. The simulation tests the planning results of MA-RRT* under three conditions cs_weight=1, cs_weight=2, cs_weight=10. In the Figure 9 , the path planning results are different for different cs_weight values. When the value of cs_weight is appropriate, the path can achieve the effect of not being too aggressive and not being too conservative. Therefore, in the actual application scene, the path can be adaptively generated by changing the parameters.

[0091] The final application effect of MA-RRT* in the above multi-unmanned aerial vehicle task scene is as shown in the figure. Figure 10 From the simulation results, it can be seen that each unmanned aerial vehicle can select a suitable path from the starting point to the target according to the actual threat area.

[0092] The MA-RRT* algorithm provided by the application has the following advantages:

[0093] 1. MA-RRT* has the function of strictly meeting the maximum turning angle constraint and ensuring the dynamics feasibility. The method proposed in the application can add a maximum turning angle constraint in the continuous state space when the object has a strict turning angle limit, so that the path strictly meets the constraint condition.

[0094] 2. After screening the nodes in the neighbor set, the nodes that need to be traversed to find the parent node are actually reduced, which obviously reduces the time cost of the RRT* algorithm, and the effect is more obvious with the increase of the number of iterations. new

[0095] 3. MA-RRT* can flexibly control the aggressiveness and conservatism of the path according to the actual situation by using the parameter cs_weight, so that in the threat area scene similar to the one described above, it can adaptively choose to cross directly, partially bypass and completely bypass, and is more in line with the actual application requirements.

[0096] Corresponding to the above method, the application also provides an unmanned aerial vehicle path planning system, comprising:

[0097] A to-be-planned scene model construction module is configured to construct a to-be-planned scene model; the to-be-planned scene model is a multi-unmanned aerial vehicle planning for multiple targets; the to-be-planned scene model comprises: unmanned aerial vehicle target distribution results and map environment information; the map environment information comprises: target positions, coordinates and threat area ranges;

[0098] ​The constraint determination module is configured to determine the maximum rotation angle constraint according to a curvature radius constraint of a turn of the UAV path.

[0099] The path planning module is configured to plan a path for the UAV in the to-be-planned scene model according to the curvature radius constraint of the turn of the UAV path, the maximum rotation angle constraint, and the threat area range, and using an RRT* algorithm.

[0100] In order to implement the above method to achieve the corresponding functions and technical effects, the present application further provides a UAV path planning device, comprising at least one processor, at least one memory and computer program instructions stored in the memory, when the computer program instructions are executed by the processor, the method is realized.

[0101] The memory is a computer readable storage medium.

[0102] Based on the above description, the technical solution of the present application or the part of the prior art that essentially contributes or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes instructions to make a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the method of each embodiment of the present application. The aforementioned computer storage medium includes: U disk, mobile hard disk, read-only memory, random access memory, magnetic disk or optical disk and various program code storage media.

[0103] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts of each embodiment can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part.

[0104] The principles and implementation modes of the present application are described by using specific examples in this paper. The above embodiment description is only used to help understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed. In conclusion, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for unmanned aerial vehicle (UAV) path planning, characterized in that, include: Construct a model of the scenario to be planned; The scenario model to be planned involves multi-objective planning using multiple drones. The scenario model to be planned includes: the distribution results of UAV targets and map environment information; the map environment information includes: target location, coordinates, and threat zone range. The maximum rotation angle constraint is determined based on the radius of curvature constraint of the UAV path turning. Based on the curvature radius constraint, maximum rotation angle constraint and threat zone range of the UAV path turning, the RRT* algorithm is used to plan the path of the UAV in the scenario model to be planned. Based on the curvature radius constraint, maximum rotation angle constraint, and threat zone range of the UAV path turning, the RRT* algorithm is used to perform path planning for the UAV in the scenario model to be planned, specifically including: Obtain the starting position and target position of the drone to be planned in the scene model; The neighbor set generated in the RRT* algorithm is further filtered by using the radius of curvature constraint and the maximum rotation angle constraint of the UAV path turning. The risk of crossing the threat zone is determined based on the threat zone's extent; and this risk is then added to the node's cost function to obtain the optimized cost function. Based on the optimized cost function, the neighbor set after secondary filtering, the search tree, and the map environment information, a new search tree after reconnection is obtained; The planned path is determined based on the new search tree after reconnection.

2. The UAV path planning method according to claim 1, characterized in that, The risk of crossing the threat zone is determined based on the extent of the threat zone. The risk of traversing the threat zone is then added to the node's cost function, resulting in an optimized cost function. Previously, this also included: Based on the starting and ending points of the path and map environment information, a numerical sampling method is used to determine the total distance across the threat zone.

3. The UAV path planning method according to claim 2, characterized in that, The risk of crossing the threat zone is determined based on the extent of the threat zone. The risk of traversing the threat zone is then incorporated into the node's cost function, resulting in an optimized cost function, which includes: Using formula Determine the optimized cost function; in, Represents a node The optimized cost function Represents a node in RRT* The cost of the parent node, Parameters representing the risk of traversing a threat zone, and parameters characterizing the aggressiveness or conservatism of a control path. This indicates the total distance traveled across the threat zone.

4. A drone path planning system for implementing the drone path planning method as described in any one of claims 1-3, characterized in that, include: The module for building the scenario model to be planned is used to build the scenario model to be planned. The scenario model to be planned involves multi-objective planning using multiple drones. The scenario model to be planned includes: the distribution results of UAV targets and map environment information; the map environment information includes: target location, coordinates, and threat zone range. The constraint determination module is used to determine the maximum rotation angle constraint based on the radius of curvature constraint of the UAV path turning. The path planning module is used to plan the path of the UAV in the scenario model based on the curvature radius constraint, maximum rotation angle constraint and threat zone range of the UAV path turning.

5. A drone path planning device, characterized in that, include: The system comprises at least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the UAV path planning method as described in any one of claims 1-3.

6. The UAV path planning device according to claim 5, characterized in that, The memory is a computer-readable storage medium.

Images