Path planning method of unmanned bee colony aircraft under multi-constraint condition

Abstract

The invention discloses a path planning method for an unmanned bee colony aircraft under a multi-constraint condition, and the method comprises the following steps: building a mathematical model of the bee colony aircraft from an initial position to a target position according to the multi-constraint condition, and converting a multi-target path planning problem of the bee colony aircraft in a three-dimensional space into an optimization problem for solving a target function; a K-Means clustering method is adopted to classify all the target positions, and the distance between each target position and the initial position of each aircraft is obtained; based on the distance between each target position and each aircraft, obtaining an access sequence of the bee colony aircrafts to the target positions through a TSP dynamic programming mode; and performing optimization iteration according to the access sequence through an improved bidirectional adaptive A * algorithm to obtain an optimal path. According to the method, the path planning efficiency and the adaptive capacity of the unmanned bee colony aircraft in a complex dynamic environment are improved, and the global optimality of the path and the efficient cooperative execution of a flight task are realized.

Description

Technical Field

[0001] The present invention relates to the field of spatial configuration planning and self-organizing control of swarm aircraft, and more specifically to a method for collaborative path planning of swarm aircraft in a complex dynamic environment. Background Art

[0002] The traditional Boids model is a classic bionic control model that realizes the coordinated control of a group through three rules: separation, alignment, and cohesion. However, this model has insufficient environmental perception dimensions, especially in the identification and avoidance of dynamic obstacles. In addition, the fixed-parameter configuration rules of the Boids model are difficult to adapt to the real-time obstacle avoidance requirements in a high-density obstacle environment, and lack a task-driven dynamic behavior adjustment mechanism.

[0003] The traditional A* algorithm is a heuristic search algorithm widely used in path planning, which finds the optimal path by using the weighted sum of the estimated cost function and the actual cost function. Although the A* algorithm can efficiently handle single-path optimization problems, in multi-constraint collaborative path planning, both the optimization efficiency and the quality of the solution are lacking, and it is difficult to meet the requirements of real-time path replanning in complex task scenarios, lacking the overall optimization ability for the kinematic constraints of the group.

[0004] The above technologies have certain application values in the fields of group collaborative control and path planning, but there are still problems in realizing multi-constraint path planning for large-scale swarm aircraft in a dynamic environment: existing bionic models and path planning algorithms lack an efficient adaptive mechanism in a multi-constraint complex environment, and traditional technologies are also difficult to achieve complex formation control and path optimization of large-scale unmanned swarm aircraft.

[0005] Therefore, providing a method for collaborative path planning by integrating the advantages of bionic swarm intelligence and modern optimization algorithms is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a path planning method for unmanned swarm aircraft under multi-constraint conditions, which solves the path planning problem under multi-constraint conditions by combining an improved bidirectional adaptive A* algorithm with a multi-objective optimization strategy, and introducing TSP dynamic programming, path re-routing, and smoothing processing.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] The present invention provides a path planning method for unmanned swarm aircraft under multi-constraint conditions, including the following steps:

[0009] S1. Establish a mathematical model for the swarm of aerial vehicles from the initial position to the target position according to multiple constraints, and transform the multi-objective path planning problem of the three-dimensional swarm of aerial vehicles into an optimization problem of solving the objective function;

[0010] S2. Use the K-Means clustering method to classify all target positions and obtain the distance between each target position and the initial position of each aerial vehicle;

[0011] S3. Based on the distance between each target position and each aerial vehicle, obtain the access order of the swarm of aerial vehicles to the target positions through the TSP dynamic programming method;

[0012] S4. Perform optimization iteration in the access order through the improved bidirectional adaptive A* algorithm to obtain the optimal path of the swarm of aerial vehicles under multiple constraints.

[0013] Furthermore, the expression of the mathematical model in step S1 is:

[0014]

[0015] where p ij (x ij ,y ij ,z ij ,α ij ,β ij ) represents the initial position state of the jth aerial vehicle, and p tj (x tj ,y tj ,z tj ,α tj ,β tj ) represents the end position state of the jth aerial vehicle; x ij ,y ij ,z ij respectively represent the three-dimensional coordinates of the initial position of the jth aerial vehicle, α ij represents the course deviation angle of the initial position of the jth aerial vehicle, and β ij represents the track azimuth angle of the initial position of the jth aerial vehicle; x tj ,y tj ,z tj respectively represent the three-dimensional coordinates of the end position of the jth aerial vehicle, α tj represents the course deviation angle of the end position of the jth aerial vehicle, and β tj represents the track azimuth angle of the end position of the jth aerial vehicle; N is the total number of aerial vehicles, N≥2; r j (q) represents the path planning of the jth aerial vehicle from the starting point to the end point under multiple constraints, and q is the path constraint parameter.

[0016] Furthermore, step S2 specifically includes:

[0017] The K-Means algorithm is used to select the initial k centroids as the center points of clustering; and all target positions are assigned to the cluster corresponding to the nearest centroid. The distance between each target position and these k centroids is calculated one by one, and the centroid with the shortest distance is selected as the cluster to which the target position belongs.

[0018] Create a two-dimensional array DP to store the length of the shortest path. The size of the array is (1<<a)*a, and the initial value is infinity; then create a two-dimensional array D to store the index of the previous point in the path for restoring the shortest path. The size of the array is (1<<a)*a, and the initial value is -1.

[0019] Further, step S3 specifically includes:

[0020] Create an array initP to store the order of the shortest path, and add the starting point to the array; in a backtracking manner, restore the order of the shortest path according to the D array; add the ending point to the array to form the final shortest path.

[0021] By traversing the combinations of states and points, continuously update the length of the shortest path and the index of the previous point, and finally obtain the access order of the points on the shortest path.

[0022] Further, step S4 specifically includes:

[0023] Obtain the intermediate nodes in the flight space of the swarm of flying vehicles through a directional search method;

[0024] Obtain the step size between the current node and the intermediate node, and adaptively adjust the step size according to the position of the obstacle;

[0025] Based on the Dijkstra algorithm and the Best-First-Search algorithm, balance the convergence speed and path quality of the step size between the nodes in the flight space of the swarm of flying vehicles by dynamically adjusting the weights of the evaluation function and the heuristic function;

[0026] Delete the intermediate nodes that deviate from the path, and then perform path re-routing and smoothing on the remaining intermediate nodes of the path to obtain the optimal path of the swarm of flying vehicles under multiple constraint conditions.

[0027] Further, the obtaining of the intermediate nodes in the flight space of the swarm of flying vehicles through the directional search method in step S4 specifically includes:

[0028] Based on the maneuverability constraint conditions, find the intermediate nodes in the flight space of the swarm of flying vehicles through a directional search method;

[0029] Perform forward and backward searches simultaneously. When the forward search path starting from the initial position and the backward search path starting from the target position explore to the same intermediate node, delete the intermediate node.

[0030] Furthermore, in step S4, the step size is adaptively adjusted according to the obstacle position, and the formula is:

[0031]

[0032] where L is the step size, L max and L min represent the maximum step size and the minimum step size respectively, D represents the distance between the current node and the nearest obstacle warning area characterized by multiple constraints, and D eff is the effective radius range of the corresponding obstacle warning area.

[0033] Furthermore, in step S4, based on the Dijkstra algorithm and the Best-First-Search algorithm, the convergence speed and path quality of the step size between the flight space nodes of the swarm aircraft are balanced by dynamically adjusting the weights of the evaluation function and the heuristic function; specifically including:

[0034] 1) Based on multiple constraint conditions including the maneuverability, coverage area, group formation, and strike target of the aircraft, define the evaluation function g(P (n) ) as:

[0035] g(P (n) ) = ω1g obs (P (n) ) + ω2g alt (P (n) ) + ω3g cost (P (n) ) + ω4g cover (P (n) ) + ω5g target (P (n) )

[0036] where g obs (P (n) ) represents the cost from the current position P (n) to its nearest obstacle, g alt (P (n) ) represents the cost of the flight altitude from the current position P (n) to the reference altitude, g cost (P (n) ) represents the Euclidean distance cost from the current position P (n) to the initial position P init , and g cover (P (n)) represents the minimum distance cost adjusted from the current position to meet the area coverage constraint, g target (P (n) ) represents the minimum modification cost from the current position to meet the strike target constraint; ω1, ω2, ω3, ω4, and ω5 are the weights of each parameter respectively;

[0037] 2) The Manhattan distance is used to calculate the heuristic function, and the formula is:

[0038] h(P(n)) = |P x (n) - P goal_x | + |P y (n) - P goal_y | + |P z (n) - P goal_z |

[0039] Among them, P x (n), P y (n), P z (n) represents the coordinates of the current node P (n) , P goal_x , P goal_y , P goal_z are the coordinates of the target node P goal ;

[0040] 3) Dynamically adjust the weights of the evaluation function and the heuristic function;

[0041] f(P(n)) = ω g g(P(n)) + ω h h(P(n))

[0042]

[0043] ω h = 1 - ω g

[0044] Among them, f(P(n)) represents the dynamic adjustment formula, ω g and ω h are the weight values for dynamic adjustment, ω max and ω min are preset hyperparameters; the larger the value of ω h , the faster the convergence speed of the algorithm, and the larger the value of ω g , the higher the quality of the path generated by the algorithm.

[0045] Furthermore, in step S4, the intermediate nodes deviating from the path are deleted, and then path re-routing and smoothing processing are performed on the remaining intermediate nodes of the path to obtain the optimal path of the swarm aircraft under multiple constraints; specifically including:

[0046] 1) Taking three consecutive path nodes as a group, continuously check all path nodes. The three consecutive nodes in each group are: node A represents the currently checked node, node B represents the child node of the currently checked node A in the path set, and node C represents the child node of node B;

[0047] If A, B, and C satisfy the following formula, the original path A - B - C is re-routed to A - C, and the redundant node B will be deleted;

[0048]

[0049] ∥C - A∥≤∥C - B∥ + ∥B - A∥

[0050] where S obs represents the obstacle warning area, ||*|| is the Euclidean norm, ||C - A|| represents the Euclidean distance between C and A, ||C - B|| represents the Euclidean distance between C and B, and ||B - A|| represents the Euclidean distance between B and A;

[0051] 2) Using the B-spline method, the m-th B-spline of m + 1 control nodes (x0, y0, z0), (x1, y1, z1), …, (xm, ym, zm) is expressed as

[0052]

[0053]

[0054] where note_u = 0, u < k, note_u = u - k + 1, k ≤ u < m, note_u = m - k + 2, m < u;

[0055] 3) The optimized path coordinates are expressed as:

[0056]

[0057] where, 0 ≤ v ≤ 10, B u,k (v) represents the blending function of the curve;

[0058] Finally, the optimal path of the swarm of flying vehicles that meets multiple constraint conditions is obtained.

[0059] From the above technical solutions, it can be seen that compared with the prior art, the present invention discloses a path planning method for an unmanned swarm of flying vehicles under multiple constraint conditions. By introducing an improved A* algorithm and a dynamic programming strategy, combined with technologies such as the dynamic constraints of unmanned aerial vehicles and path direction constraints, the planning rules are dynamically adjusted, enhancing the adaptability of the swarm of flying vehicles in scenarios with dense obstacles and frequent environmental changes, thereby avoiding problems such as formation chaos or collisions, and improving the stability and flexibility of task execution.

[0060] Meanwhile, the present invention comprehensively applies TSP dynamic programming, an improved bidirectional adaptive A* algorithm, path re-routing, and smoothing processing to optimize the computational complexity, improve the path planning efficiency and accuracy of the unmanned swarm aircraft under multiple constraint conditions, and meet the requirements of real-time path re-planning.

[0061] The present invention significantly improves the global optimality of path planning and the coordination of flight missions, and improves the task execution efficiency and reliability of the existing technology in complex dynamic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the provided drawings.

[0063] Figure 1 It is a flowchart of a path planning method for an unmanned swarm aircraft under multiple constraint conditions provided by the present invention.

[0064] Figure 2 It is a schematic diagram of directional search for candidate nodes that meet the constraint conditions provided by an embodiment of the present invention.

[0065] Figure 3 It is a schematic diagram of step size adjustment when the unmanned aircraft is in different positions provided by an embodiment of the present invention.

[0066] Figure 4 It is a comparison diagram before and after re-routing and smoothing processing provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0067] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0068] An embodiment of the present invention discloses a path planning method for an unmanned swarm aircraft under multiple constraint conditions. Referring to Figure 1 as shown, the method includes the following steps:

[0069] S1. Establish a mathematical model of the swarm aircraft from the initial position to the target position according to multiple constraint conditions, and transform the multi-objective path planning problem of the three-dimensional space swarm aircraft into an optimization problem of solving the objective function;

[0070] S2. Classify the target positions using the K-Means clustering method and obtain the distances between each target position and each initial position of the aircraft;

[0071] S3. Based on the distances between each target position and each aircraft, obtain the visiting order of the swarm aircraft to the target positions through the TSP dynamic programming method;

[0072] S4. Perform optimization iteration according to the visiting order through the improved bidirectional adaptive A* algorithm to obtain the optimal path.

[0073] The following uses specific embodiments to describe the implementation steps of the present invention in detail:

[0074] Step S1. First, represent the planned path in the form of a spatial position point sequence. p i (x i , y i , z i , α i , β i ) represents the initial position state of the aircraft. p t (x t , y t , z t , α t , β t ) represents the end position state of the aircraft. x i , y i , z i respectively represent the three-dimensional coordinates of the initial position of the aircraft. α i represents the course deviation angle of the initial position, and β i represents the track azimuth angle of the initial position. r j (q) represents the path planning of the j-th aircraft from the starting point to the end point under multiple constraints. q is the path constraint parameter, which can represent either the length of a straight-line path restricted by a voyage or the curvature of a path that satisfies the turning angle limit. From the above, the path planning r ij from the starting point p tj to the end point p i of the j-th aircraft can be expressed by the mathematical formula:

[0075]

[0076] Secondly, during the process of the UAV performing path planning, it will be affected by multiple constraints, specifically including:

[0077] 1) Kinetic constraints: The constraints on the maneuverability of UAVs mainly include the flight altitude, flight speed, maximum elevation angle, and maximum turning angle of the UAV, etc. The maximum and minimum flight altitudes will directly affect the flight airspace of the UAV, and indirectly affect its path planning. It will affect the process of the A* directional search to confirm candidate nodes. For example, candidate nodes to be searched that exceed performance constraint conditions such as the maximum flight altitude, elevation angle, and turning angle will be abandoned.

[0078] The maximum pitch angle is used to characterize the ultimate ability of the UAV to climb or dive. The regulation of it will be reflected in the vertical attitude state of the UAV fuselage. The pitch angle will ensure that the height direction and glide angle of the planned path are within a certain optimal range.

[0079] 2) Mission constraints: Mission requirements such as the target points to be reached in the path and the flight range. It affects the process of the TSP planning to search for a path. For example, paths that cannot reach the path points sequentially completely will be abandoned.

[0080] 3) Target threat constraints: The threats faced by UAV path planning mainly include terrain threats and environmental threats. According to the intensity and coverage of the threats, they are divided into point threats and area threats. Point threats mean that the UAV is only affected by a single threat source. Area threats refer to a fan-shaped threat area composed of multiple threat sources. The size and intensity of the threats will directly affect the path selection and payload constraints of the UAV in the planning. It affects the A* heuristic search process of path planning. For example, autonomous search can avoid paths with obstacles and select paths with less threats. Paths that do not meet the target threat constraint conditions will not be heuristically searched

[0081] 4) Time limit constraints: It mainly means that the fuel carried by the UAV is limited when performing flight tasks, and there is a maximum flight time limit. If a tour path is too long, the calculated flight time will exceed the maximum flight time limit. The path planning will try to plan multiple tour paths (such as the number of clusters is 2) to ensure that a path that can still reach each target point in sequence is obtained.

[0082] The above-mentioned constraint conditions such as environmental constraints, UAV self-constraints, and mission constraints are all denoted as , then the mathematical expression of path planning is expressed as:

[0083]

[0084] Among them, p ij (x ij ,y ij ,z ij ,α ij ,β ij ) represents the initial position state of the jth aircraft, p tj (x tj ,y tj ,z tj, α tj , β tj ) represents the end - point position state of the j - th aircraft; x ij , y ij , z ij respectively represent the three - dimensional coordinates of the initial position of the j - th aircraft, α ij represents the course deviation angle of the initial position of the j - th aircraft, β ij represents the track azimuth angle of the initial position of the j - th aircraft; x tj , y tj , z tj respectively represent the three - dimensional coordinates of the end - point position of the j - th aircraft, α tj represents the course deviation angle of the end - point position of the j - th aircraft, β tj represents the track azimuth angle of the end - point position of the j - th aircraft; N is the total number of aircraft, N≥2; r j (q) represents the path planning of the j - th aircraft from the starting point to the end point under multiple constraint conditions, and q is the path constraint parameter.

[0085] Common mathematical programming methods, artificial potential field methods, graphics - based methods, and intelligent optimization algorithms usually do not consider kinematic constraints when planning paths; since the flight of unmanned aerial vehicles is a complex process and is affected by various factors, dynamic constraints are an essential part that must be considered in reliable path planning.

[0086] Now assume that the mass of the unmanned aerial vehicle is a constant, the curvature of the earth is ignored, and the gravitational acceleration does not change with the flight altitude. The dynamic equation of the unmanned aerial vehicle is as follows:

[0087]

[0088] Here p i (x i , y i , z i ) is the initial - point position, α i represents the course deviation angle, β i represents the track azimuth angle, α i and β i are control input variables, βmin≤β(i)≤βmax; the radius of curvature is

[0089]

[0090] where, Rmin≤R(s)≤Rmax.

[0091] Furthermore, the optimal control problem can be transformed into the shortest - path problem Here S t represents the length of the planned path, that is, the path - planning problem in three - dimensional space can be transformed into an optimization problem of solving the objective function.

[0092] Furthermore, the optimal control problem can be transformed into the shortest path problem, that is, the path planning problem in three-dimensional space can be transformed into the optimization problem of solving the objective function.

[0093] In step S2, the K-Means clustering method is used to classify the target points. The main purpose is to decompose a large-scale problem into independent sub-problems and solve the optimal path for each sub-problem separately.

[0094] First, use the K-Means++ algorithm to select the initial k centroids, which will be used as the center points of clustering. After determining these k initial centroids, all target points are assigned to the clusters corresponding to the nearest centroid. Calculate the distance between each target point and these k centroids one by one, and select the nearest centroid as the cluster to which the target point belongs.

[0095] Subsequently, create a two-dimensional array DP to store the length of the shortest path. The size of the array is (1<<a)*a, and the initial value is infinity. Additionally, create a two-dimensional array D to store the index of the previous point in the path for restoring the shortest path. The size of the array is (1<<a)*a, and the initial value is -1. Then initialize the direct distance from the starting radar point to each radar point and update the DP and D arrays.

[0096] Traverse the states and calculate the shortest path by means of dynamic programming. For each state, traverse each point. If the point has not been visited and is not in the traversed sequence, update the shortest path length and the index of the previous point.

[0097] In step S3, find the shortest loop, find the length of the shortest path and the index of the end point of the path. Create an array initP to store the order of the shortest path, and add the starting point to the array. By means of backtracking, restore the order of the shortest path according to the D array. Add the end point to the array to form the final shortest path.

[0098] By traversing the combinations of states and points, continuously update the length of the shortest path and the index of the previous point, and finally obtain the order of the points on the shortest path.

[0099] Finally, in step S4, use the improved bidirectional adaptive A* algorithm to optimize and iterate in sequence between these points to determine the optimal path, and finally connect them to form the overall optimal path.

[0100] In this embodiment, according to step S4, first obtain the intermediate nodes in the flight space of the swarm aircraft through the directional search method;

[0101] Based on the maneuverability constraint conditions, find the intermediate nodes through the directional search method; refer to Figure 2As shown, the candidate nodes obtained through the directional search strategy are all distributed on the arc surface that satisfies the constraints such as the pitch angle θ and yaw angle of the drone, and the coverage area. The directional search strategy not only ensures the exploration ability of the algorithm, but also improves the efficiency of finding candidate nodes. The directional search strategy not only ensures the exploration ability of the algorithm, but also improves the efficiency of finding candidate nodes.

[0102] To further improve the exploration ability of the A* algorithm, the bidirectional adaptive A* algorithm introduces a bidirectional search strategy, performing forward and backward searches simultaneously. When the forward search path A init originating from the initial node P a * and the backward search path A goal originating from the target node P b * explore to the same child node, the algorithm will return an optimal path that avoids threats such as obstacles and satisfies the constraint conditions. Subsequently, it is necessary to rewire and smooth it to further optimize it as the final path of the unmanned swarm aircraft.

[0103] As the number of candidate nodes increases, the performance simulation of the multi-directional traditional A* algorithm in terms of yaw angle can be improved, thereby effectively enhancing the exploration ability. However, due to the increase in the number of candidate nodes, the multi-directional A* algorithm will inevitably face the problems of a huge increase in computational complexity and out-of-memory.

[0104] Therefore, in this embodiment, the improved bidirectional adaptive A* algorithm introduces a more effective extended search method to balance the weight between computational complexity and performance, that is, by combining the direction vector from the current node to the target node in the obstacle-free space and the constraints such as the pitch angle, yaw angle, and coverage area of the drone, the candidate nodes of the current node are confirmed through the directional search strategy.

[0105] When the traditional A* algorithm connects the current node P (n) and the child node P child(n) , the step size is fixed short, which will delay its convergence under obstacle-free conditions to a certain extent. To overcome this defect, an adaptive step size strategy is introduced into the search process, and the step size is adaptively adjusted when the current node is in different environments (such as the obstacle area and the safe area). The calculation formula of the adaptive step size strategy is as follows:

[0106]

[0107] where L is the step size, L max and L min represent the maximum step size and the minimum step size respectively, D represents the distance between the current node and the nearest obstacle warning area characterized by various constraints, and D eff is the effective radius range of the corresponding obstacle warning area. Refer to Figure 4As shown, when the UAV is far from the obstacle, the UAV is considered to be in the safe area (D≥D eff ). In this case, a large step size is allowed to accelerate convergence. Correspondingly, if the UAV is estimated to be in the obstacle warning area (D<D eff ), a small step size is adopted to ensure the fine-grained feasibility of the path.

[0108] In this embodiment, based on the Dijkstra algorithm and the Best-First-Search algorithm, the convergence speed of the step size between nodes and the path quality are balanced by dynamically adjusting the weights of the evaluation function and the heuristic function; specifically, it includes:

[0109] 1) Under various constraint conditions such as maneuverability, coverage area, group formation, and strike target, the evaluation function g(P (n) ) of the UAV is expressed as:

[0110] g(P (n) ) = ω1g obs (P (n) ) + ω2g alt (P (n) ) + ω3g cost (P (n) ) + ω4g cover (P (n) ) + ω5g target (P (n) )

[0111] where g obs (P (n) ) represents the cost from the current position P (n) to its nearest obstacle, g alt (P (n) ) represents the cost of the flight altitude from the current position P (n) to the reference altitude, g cost (P (n) ) represents the Euclidean distance cost from the current position P (n) to the initial position P init , g cover (P (n) ) represents the minimum distance cost of adjusting from the current position to meet the area coverage constraint, g target (P (n) ) represents the minimum modification cost from the current position to meet the strike target constraint; ω1, ω2, ω3, ω4, and ω5 are the weights of each parameter respectively.

[0112] 2) Different from the traditional A* algorithm that uses the Euclidean distance to calculate the heuristic function h(P(n)), the bidirectional adaptive A* algorithm uses the Manhattan distance for calculation:

[0113] h(P(n)) = |P x (n) - P goal_x | + |P y (n) - P goal_y | + |P z (n) - P goal_z |

[0114] where (P x (n), P y (n), P z (n)) represents the coordinates of the current node P (n) , and (P goal_x , P goal_y , P goal_z ) are the coordinates of the target node P goal .

[0115] Due to the different geometric measurement methods adopted by the heuristic function h(P(n)), its specific optimization effects are also different. The heuristic function h(P(n)) based on the Euclidean distance will explore more adjacent node spaces. At the same time, there are a large number of square and square root operations in the Euclidean metric, which leads to exponential computational burden and memory consumption. On the contrary, the heuristic function h(P(n)) based on the Manhattan distance can conduct more accurate space exploration, reduce redundant nodes, thus ensuring the efficiency of the algorithm and the quality of the generated path.

[0116] 3) When the weight of h(P(n)) is small, the A* algorithm is close to the Dijkstra algorithm. It seeks the shortest path from the initial node P init to the target node P goal by exploring a large amount of space, and its convergence is relatively good, but the convergence speed is slow. When the weight of g(P (n) ) is small, the A* algorithm is transformed into the Best - First - Search algorithm (BFS). It will efficiently find a feasible path, but this path is usually not optimal for problem - solving.

[0117] Therefore, the A* algorithm based on the Dijkstra algorithm and the BFS algorithm can balance the convergence speed and the path quality by dynamically adjusting the weights of g(P (n) ) and h(P(n)). Thus, the evaluation function calculation method of the bidirectional adaptive A* algorithm is as follows:

[0118] f(P(n)) = ω g g(P(n)) + ω h h(P(n))

[0119] where ω g and ω his a weight value that can be dynamically adjusted, and the calculation method is as follows:

[0120]

[0121] ω h = 1 - ω g

[0122] where ω max and ω min are pre-set hyperparameters. The larger the value of ω h , the faster the convergence speed of the algorithm. Correspondingly, the larger the value of ω g , the higher the quality of the path generated by the algorithm. It can be seen from the formula that when the UAV plans a path in the obstacle-free area (D ≥ D eff ), ω h will be set to the maximum value, and ω g will be set to the minimum value to avoid useless exploration and accelerate the convergence speed. When the UAV enters the obstacle warning area (D < D eff ), more precise spatial exploration is required to generate a high-quality path. Therefore, the value of ω g will be adaptively increased to find an optimal path in the obstacle area. Thus, through the adaptive adjustment of the weights, the search space and the convergence speed are taken into account, and finally the performance and efficiency of the algorithm are balanced.

[0123] This embodiment then introduces a re-routing process to delete redundant nodes, reduce the total path length, eliminate redundant turning nodes, and smooth the path after re-routing to further optimize the generated path and enhance the flight feasibility of the final path. Referring to Figure 4 shown, an optimal path of A - B - C - D is obtained by bidirectional adaptive A* search. First, re-routing is performed to delete the redundant node B to get A - C - D, and then it is smoothed to obtain the final path of AD that is beneficial for the fixed-wing UAV to fly. The path generated by the above path planning algorithm consists of several short line segments, which connect a large number of path nodes, and may include some redundant nodes. Introducing a re-routing process to delete redundant nodes can reduce the total path length and eliminate redundant turning nodes. At the same time, path smoothing is the guarantee for the UAV to fly safely and smoothly along the planned path. The paths after the above re-routing are mostly connected by straight lines. Although they theoretically meet the requirements, the flight feasibility at the positions of the straight-line intersection inflection points in the actual flight path is usually low. To meet the flight requirements of the UAV, the generated path can be further optimized by smoothing the path after re-routing to enhance the flight feasibility of the final path.

[0124] The specific re-routing process is as follows:

[0125] 1) Taking three consecutive path nodes as a group, continuously check all path nodes. The three consecutive nodes in each group are: node A represents the currently checked node; node B represents the child node of the currently checked node A in the path set; node C represents the child node of node B.

[0126] If A, B, and C satisfy the following formula, the original path A - B - C will be re - routed as A - C, and the redundant node B will be deleted.

[0127]

[0128] ∥C - A∥≤∥C - B∥+∥B - A∥

[0129] where S obs represents the obstacle warning area, ||*|| is the Euclidean norm, ||C - A|| represents the Euclidean distance between C and A, ||C - B|| represents the Euclidean distance between C and B, and ||B - A|| represents the Euclidean distance between B and A;

[0130] 2) Then, using the B - spline method, the m - th B - spline of m + 1 control nodes (x0,y0,z0),(x1,y1,z1),…,(xm,ym,zm) is expressed as

[0131]

[0132] where note_u = 0, u < k, note_u = u - k + 1, k ≤ u < m, note_u = m - k + 2, m < u;

[0133] 3) The optimized path coordinates can be expressed as:

[0134]

[0135] where, 0 ≤ v ≤ 10, B u,k (v) represents the blending function of the curve;

[0136] Through the above process, the final smooth optimal path points that meet the constraint conditions can be obtained.

[0137] The present invention proposes an improved path planning algorithm under multiple constraints, comprehensively considering constraint conditions such as dynamic constraints, task constraints, target threat constraints, and time - limit constraints. It decomposes large - scale problems in the task into independent sub - problems and separately solves the optimal paths, that is, applying TSP dynamic programming and an improved bidirectional adaptive A* algorithm, and at the same time performing path re - routing and smoothing processing to achieve the global optimality and feasibility of path planning.

[0138] The present invention improves the path planning efficiency and real-time performance of unmanned swarm aircraft in complex dynamic environments, enabling rapid response to dynamic changes in the mission environment; it also enhances the global optimality and local feasibility of paths, achieving efficient collaborative optimization of path planning under multiple constraints; it improves the smoothness and adaptability of path execution, ensuring the mission execution efficiency and reliability of swarm aircraft in dynamic complex environments; at the same time, it overcomes the deficiencies of traditional algorithms in poor adaptability to multiple constraints and low planning efficiency, providing technical support for large-scale collaborative missions of unmanned swarm aircraft.

[0139] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and reference can be made to the description of the method part for related parts.

[0140] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A path planning method for unmanned swarm aircraft under multiple constraint conditions, characterized in that, It includes the following steps: S1. Establish a mathematical model for the swarm aircraft from the initial position to the target position according to multiple constraints, and transform the multi-objective path planning problem of the three-dimensional space swarm aircraft into an optimization problem of solving the objective function; S2. Use the K-Means clustering method to classify all target positions, and obtain the distance between each target position and each aircraft's initial position; S3. Based on the distance between each target position and each aircraft, obtain the access order of the swarm aircraft to the target positions by means of TSP dynamic programming; S4. Optimize and iterate according to the access order through the improved bidirectional adaptive A* algorithm to obtain the optimal path of the swarm aircraft under multiple constraints.

2. The path planning method of an unmanned swarm aircraft under multiple constraint conditions according to claim 1, characterized in that, The expression of the mathematical model in step S1 is: Among them, p ij (x ij , y ij , z ij , α ij , β ij ) represents the initial position state of the j-th aircraft, and p tj (x tj , y tj , z tj , α tj , β tj ) represents the end position state of the j-th aircraft; x ij , y ij , z ij respectively represent the three-dimensional coordinates of the initial position of the j-th aircraft, α ij represents the course deviation angle of the initial position of the j-th aircraft, and β ij represents the track azimuth angle of the initial position of the j-th aircraft; x tj , y tj , z tj respectively represent the three-dimensional coordinates of the end position of the j-th aircraft, α tj represents the course deviation angle of the end position of the j-th aircraft, and β tj represents the track azimuth angle of the end position of the j-th aircraft; N is the total number of aircraft, N ≥ 2; r j (q) represents the path planning of the j-th aircraft from the starting point to the end point under multiple constraint conditions, and q is the path constraint parameter.

3. A path planning method for an unmanned swarm aircraft under multiple constraint conditions according to claim 1, characterized in that, Step S2 specifically includes: Use the K-Means algorithm to select the initial k centroids as the center points of clustering; and assign all target positions to the clusters corresponding to the nearest centroids, calculate the distance between each target position and these k centroids one by one, and select the nearest centroid as the cluster to which the target position belongs; Create a two-dimensional array DP to store the length of the shortest path, with the array size of (1<<a)*a and the initial value being infinity; then create a two-dimensional array D to store the index of the previous point in the path for restoring the shortest path, with the array size of (1<<a)*a and the initial value being -1.

4. The path planning method of an unmanned swarm aircraft under multiple constraint conditions according to claim 3, wherein, Step S3 specifically includes: Create an array initP to store the order of the shortest path, and add the starting point to the array; restore the order of the shortest path according to the D array by means of backtracking; add the ending point to the array to form the final shortest path; By traversing the combination of states and points, continuously update the length of the shortest path and the index of the previous point, and finally obtain the access order of the points of the shortest path.

5. The path planning method of an unmanned swarm aircraft under multiple constraint conditions according to claim 1, characterized in that, Step S4 specifically includes: Obtain the intermediate nodes of the flight space of the swarm aircraft through the directional search method; Obtain the step length between the current node and the intermediate node, and adaptively adjust the step length according to the obstacle position; Based on the Dijkstra algorithm and the Best-First-Search algorithm, balance the convergence speed and path quality of the step lengths between the nodes in the flight space of the swarm aircraft by dynamically adjusting the weights of the evaluation function and the heuristic function; Delete the intermediate nodes deviating from the path, and then perform path re-wiring and smoothing processing on the remaining intermediate nodes of the path to obtain the optimal path of the swarm aircraft under multiple constraints.

6. The path planning method of an unmanned swarm aircraft under multiple constraint conditions according to claim 5, characterized in that In step S4, the intermediate nodes of the flight space of the swarm aircraft are obtained through the directional search method; specifically including: Based on the maneuverability constraint conditions, find the intermediate nodes of the flight space of the swarm aircraft through the directional search method; Perform forward and backward searches simultaneously. When the forward search path starting from the initial position and the backward search path starting from the target position explore to the same intermediate node, delete this intermediate node.

7. The path planning method of an unmanned swarm aircraft under multiple constraint conditions according to claim 6, wherein, In step S4, the formula for adaptively adjusting the step length according to the obstacle position is: where L is the step size, L max and L min represent the maximum step size and the minimum step size respectively, D represents the distance between the current node and the nearest obstacle warning area characterized by multiple constraints, D eff is the effective radius range of the corresponding obstacle warning area.

8. The path planning method of an unmanned swarm aircraft under multiple constraint conditions according to claim 7, characterized in that, In step S4, based on the Dijkstra algorithm and the Best-First-Search algorithm, the convergence speed of the step length between the flight space nodes of the swarm of flying vehicles and the path quality are balanced by dynamically adjusting the weights of the evaluation function and the heuristic function; specifically including: 1) Based on various constraint conditions including the maneuverability of the aircraft, the coverage area, the group formation, and the strike target, define the evaluation function g(P (n) ) as: g(P (n) ) = ω1g obs (P (n) ) + ω2gx lt (P (n) ) + ω3g cost (P (n) ) + ω4g cover (P (n) ) + ω5g target (P (n) ) Among them, g obs (P (n) ) represents the cost from the current position P (n) to its nearest obstacle, g alt (P (n) ) represents the cost of the flight altitude from the current position P (n) to the reference altitude, g cost (P (n) ) represents the Euclidean distance cost from the current position P (n) to the initial position P init , g cover (P (n) ) represents the minimum distance cost for adjusting from the current position to meet the area coverage constraint, g target (P (n) ) represents the minimum modification cost from the current position to meet the strike target constraint; ω1, ω2, ω3, ω4, and ω5 are the weights of each parameter respectively; 2) The Manhattan distance is used to calculate the heuristic function, and the formula is: h(P(n)) = |P x (n) - P goal_x | + |P y (n) - P goal_y | +|P z (n)-P goal_z | Among them, P x (n), P y (n), P z (n) represents the coordinates of the current node P (n) , P goal_x , P goal_y , P goal_z are the coordinates of the target node P goal ; 3) Dynamically adjust the weights of the evaluation function and the heuristic function; f(P(n)) = ω g g(P(n)) + ω h h(P(n)) ω h = 1 - ω g Among them, f(P(n)) represents the dynamic adjustment formula, ω g and ω h are the weight values for dynamic adjustment, ω max and ω min are the hyperparameters set in advance; the larger the value of ω h , the faster the convergence speed of the algorithm, and the larger the value of ω g , the higher the quality of the path generated by the algorithm.

9. A path planning method for an unmanned swarm aircraft under multiple constraint conditions according to claim 8, characterized in that In step S4, the intermediate nodes deviating from the path are deleted, and then the remaining intermediate nodes of the path are subjected to path re-routing and smoothing processing to obtain the optimal path of the swarm of flying vehicles under multiple constraints; specifically including: 1) Taking three consecutive path nodes as a group, all path nodes are continuously checked. The three consecutive nodes in each group are: node A represents the currently checked node, node B represents the child node of the currently checked node A in the path set, and node C represents the child node of node B; If A, B, and C satisfy the following formula, the original path A-B-C is re-routed to A-C, and the redundant node B will be deleted; ∥C-A∥≤∥C-B∥+∥B-A∥ Among them, S obs represents the obstacle warning area, ||*|| is the Euclidean norm, ||C - A|| represents the Euclidean distance between C and A, ||C - B|| represents the Euclidean distance between C and B, and ||B - A|| represents the Euclidean distance between B and A; 2) Using the B-spline method, the m-th B-spline of m+1 control nodes (x0,y0,z0),(x1,y1,z1),…,(xm,ym,zm) is expressed as where note_u = 0, u < k, note_u = u - k + 1, k ≤ u < m, note_u = m - k + 2, m < u; 3) The optimized path coordinates are expressed as: where 0 ≤ v ≤ 10, B ,k (v) represents the blending function of the curve; Finally, the optimal path of the swarm of flying vehicles that meets multiple constraints is obtained.

Images