A fixed-wing unmanned aerial vehicle formation landing auxiliary decision method

Abstract

The application discloses a fixed-wing unmanned aerial vehicle formation landing auxiliary decision method, which firstly sets unmanned aerial vehicle landing stage self-performance constraints and multi-unmanned aerial vehicle formation landing constraint criteria; secondly, improves the traditional ant colony algorithm by using a heuristic function combining a distance induction factor and an angle induction factor, a mechanism combining pheromone diffusion and pheromone random evaporation, and plans a single unmanned aerial vehicle landing track; then outputs multi-unmanned aerial vehicle formation landing tracks and smooth tracks; finally, according to the constraint criteria, it is judged whether the unmanned aerial vehicle formation can land smoothly and safely, and the auxiliary decision of the fixed-wing unmanned aerial vehicle formation landing is realized. The improved heuristic function and pheromone updating mechanism are adopted, the unmanned aerial vehicle landing constraint criteria are set, the unmanned aerial vehicle glide track deviation is effectively reduced, and the unmanned aerial vehicles do not collide and have good overall landing efficiency.

Description

Technical Field

[0001] This invention belongs to the field of multi-UAV trajectory planning, specifically involving a fixed-wing UAV formation landing auxiliary decision-making method for planning and deciding on UAV formation landing trajectories. Technical Background

[0002] Landing assistance decision systems for unmanned aerial vehicles (UAVs) provide a range of decision support functions, primarily assisting commanders and UAVs in environmental awareness and planning landing trajectories. Falling within the scope of UAV collaborative decision-making and control, this is one of the most critical aspects of UAV flight, determining whether the UAV can return to its starting point safely after completing its mission. Currently, decision-making technologies for collaboration among multiple UAVs are a key research focus, significantly contributing to improving the environmental perception and decision-making efficiency of multi-UAV systems.

[0003] Researching drone formation landing decisions requires obtaining drone descent altitude information and performing 3D spatial trajectory planning. 3D trajectory planning keeps the drone's flight area within a certain spatial range. Due to altitude limitations and 3D terrain constraints, threat models become more three-dimensional and intuitive, simulating realistic flight scenarios. However, this also increases the complexity of the trajectory planning algorithm. Many trajectory planning algorithms exist, but traditional classical algorithms such as dynamic programming and optimal control are not suitable for 3D space and have complex models requiring long planning times. Ant colony optimization (ACO) is an artificial intelligence algorithm with advantages such as distributed parallelism, positive feedback, and high robustness. Its model is easier to understand and implement, making it a popular research method. However, traditional ACO still suffers from drawbacks such as blind searching, susceptibility to local optima, and applicability only to discrete spaces. Further research is needed to optimize ACO by combining 3D airport environments and drone landing characteristics to achieve safe and reliable landing assistance decisions for drone formation entry and landing. Summary of the Invention

[0004] The purpose of this invention is to provide a method for assisting decision-making in the landing of fixed-wing unmanned aerial vehicles (UAVs) formations, which can solve the problems of low landing efficiency and low safety of UAV formations.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A method for assisting decision-making in the landing of fixed-wing unmanned aerial vehicle formations includes:

[0007] Step 1: To ensure the smooth and safe landing of fixed-wing UAVs, determine the performance constraints of the UAVs during the landing phase and set landing constraint criteria for multi-UAV formations.

[0008] Step 2: Establish a 3D airport environment spatial model D for drone formation landing;

[0009] Step 3, define the initial state P of a single drone. S The parameters include the approach and landing coordinates, the distance during the approach and capture phase, moving all ants to the initial point, and setting the ant colony algorithm initialization parameters, including the ant colony size N, the number of iterations K, the number of ants M, the starting point location information, the maximum initial value of pheromone τ0, the pheromone evaporation coefficient ρ, the pheromone diffusion factor δ, the pheromone importance parameter α, and the heuristic function importance parameter β.

[0010] Step 4, using distance induction factor D j_pt With angle-induced factor Ψ ij Combined heuristic function η ij Perform node probability calculation;

[0011] Step 5: Use the roulette wheel selection method to perform ant state transitions, and store the selected nodes in the tabu list TABU until the search reaches the destination and the feasible path is stored.

[0012] Step 6: Introduce a mechanism that combines pheromone diffusion and random pheromone evaporation, and use a global plus local pheromone update method to update pheromones;

[0013] Step 7: After reaching the maximum number of iterations, output the best path for landing a single UAV, and plan the UAV formation landing path based on the landing paths of multiple UAVs output in Steps 3 to 7.

[0014] Step 8: Determine whether the drone formation can land smoothly and safely based on the drone's own performance constraints during the landing phase and the multi-drone formation landing constraint criteria.

[0015] The present invention has the following advantages:

[0016] 1. This invention uses the distance-inducing factor D j_pt With angle-induced factor Ψ ij Combined heuristic function η ij Performing node probability calculations can effectively reduce the blind, step-by-step search by ants, reduce the deviation of the drone's descent trajectory, and avoid the problem of too many descent lines.

[0017] 2. This invention introduces a mechanism combining pheromone diffusion and random pheromone evaporation into the global and local pheromone update rules, simulating a more realistic pheromone environment in nature, so that pheromones will not be concentrated too much on the optimal path in a certain iteration, thus avoiding the local optimal solution problem caused by the traditional ant colony algorithm to a certain extent.

[0018] 3. The multi-UAV formation landing constraint criteria set by the present invention include the path non-intersection criterion and the UAV separation distance criterion, which are used to make landing safety judgments after UAV formation landing trajectory planning, and can ensure that UAVs do not collide with each other and the overall landing efficiency reaches the maximum.

[0019] Figure and Table Description

[0020] Figure 1 This is a flowchart of the method of the present invention.

[0021] Figure 2 This is a schematic diagram of trajectory smoothing based on cubic B-spline curves.

[0022] Figure 3 This is a three-dimensional environmental space model of the airport.

[0023] Figure 4 This is a 3D spatial planning route map based on the traditional ant colony algorithm.

[0024] Figure 5 This invention provides a three-dimensional spatial planning route map for the improved ant colony algorithm.

[0025] Figure 6 A comparison of iterative curves for 3D spatial route planning using the traditional and the improved ant colony algorithm of this invention.

[0026] Figure 7 This invention provides a three-dimensional plan of the landing routes for four unmanned aerial vehicles (UAVs). Detailed Implementation

[0027] The technical solution of the present invention will be described in detail with reference to the accompanying drawings and tables.

[0028] The present invention provides a method for assisting decision-making in the landing of fixed-wing unmanned aerial vehicle formations, which specifically includes the following steps:

[0029] Step 1: Determine the inherent performance constraints of the drones during the landing phase and set landing constraint criteria for multi-drone formations, specifically:

[0030] (1.1) The performance constraints of the UAV during the landing phase include four parts: UAV range constraint, maximum turning angle constraint, maximum glide angle constraint, and maximum glide curvature constraint, specifically:

[0031] (1.1.1) Define the UAV range constraint as follows: for each segment of the UAV's landing range l i The total range L cannot be less than a certain threshold, and the total flight distance L cannot be greater than a certain threshold, that is:

[0032]

[0033] In the formula, l minL represents the minimum distance limit for each segment of flight. max This represents the maximum limit on the total landing distance, and n represents the total number of flight segments.

[0034] (1.1.2) Define the maximum turning angle constraint as follows: Assuming the distance of each segment of the UAV's flight is represented by a vector, the maximum turning angle constraint at each node traversed is as follows:

[0035]

[0036] In the formula, This represents the range vector of the i-th segment. Let Ψ represent the range vector of the (i+1)th segment, and let Ψ represent the maximum turning angle of the UAV when turning.

[0037] (1.1.3) Define the maximum glide angle constraint for the UAV as follows: The UAV is generated by node (x i y i , z i ) to node (x i+1 y i+1 , z i+1 The maximum glide slope angle λ of this trajectory must satisfy the following condition:

[0038]

[0039] In the formula, x i y i z i x i+1 y i+1 z i+1 Let λ be the horizontal, vertical, and angular coordinates of the UAV at the i-th and (i+1)-th nodes, respectively, λ be the maximum glide angle, and n be the total number of nodes.

[0040] (1.1.4) Define the maximum curvature constraint for the drone's descent as follows: the curvature of the turn must be less than a certain threshold when the drone is descent and turning. The specific steps are as follows:

[0041] First, a cubic B-spline curve is used to smooth the planned landing polygonal trajectory. Four consecutive nodes N0, N1, N2, and N3 in the planned trajectory are selected, and N0N2 and N1N3 are connected. The midlines N1M1 and N2M2 of ΔN0N1N2 and ΔN1N2N3 are constructed, and the distance from point N1 on line segment N1M1 is taken. Point P1 is located at a distance of N2M2 from point N2. At point P2, when connecting P1 and P2, the tangent line of the trajectory at P1 should be parallel to N0N2, and the tangent line of the trajectory at P2 should be parallel to N1N3. Considering that the landing trajectory has an approach point and a landing point, boundary processing of the B-spline curve is required. The steps are as follows: Add a control point S and a control point T at each end of the polyline N0N1N2N3, replacing point N0 with point S and point N1 with point T, where point S lies on the extension of N1N0. Similarly This allows for the creation of a smooth path from N0 to N3, such as... Figure 2 As shown. The matrix expression for the connected smooth flight path satisfying the curve segment function N(t) is:

[0042]

[0043] Next, the B-spline curve undergoes boundary processing. The method is as follows: add a control point S and a control point T at each end of the polyline N0N1N2N3, replacing point N0 with point S and point N1 with point T. Point S lies on the extension of N1N0. Similarly This allows for the creation of a smooth flight path from N0 to N3, making it less likely for the drone to have a large turning radius when landing due to following sharp angles.

[0044] Then, the smoothed flight path is described as a three-dimensional spatial curve: x = x(t), y = y(t), z = z(t). Taking the first derivative of this three-dimensional spatial curve yields... Taking the second derivative again yields Treat the three first derivatives as a three-dimensional vector. Treat the three second derivatives as a three-dimensional vector. Then the maximum value of the glide curvature K of the drone max The conditions that must be met are:

[0045]

[0046] In the formula, k(t) represents the glide curvature of the UAV. This represents the first-order derivative three-dimensional vector of the UAV's flight path curve. K represents the second-order derivative three-dimensional vector of the UAV's trajectory curve. max This represents the maximum value of the drone's glide curvature.

[0047] (1.2) Set landing constraints for multi-UAV formations, including path non-intersection criteria and UAV separation distance criteria, specifically:

[0048] (1.2.1) Define the path non-intersection criterion as follows: Assuming each fixed-wing UAV is considered as a point mass and its safe range is considered as a circle, then the shortest distance d between the trajectories of two UAVs is... safe_minIt should be greater than the sum of the safe radii of the two drones, specifically satisfying the following:

[0049] d safe_min >r s_i +r s_j (6)

[0050] In the formula, r s_i r s_j Let be the safe radii of the i-th and j-th drones, respectively.

[0051] Assume points A and B are any two points on the flight path of the two drones, d A_to_B The distance between the flight paths of two drones is expressed as follows:

[0052]

[0053] In the formula, (x a y a , z a Let x be the coordinates of a drone at point A. b y b , z b () indicates the coordinates of another drone at point B.

[0054] According to equation (7), the shortest distance d between the flight paths of two UAVs in a formation can be obtained. safe_min .

[0055] (1.2.2) Define the UAV separation distance criterion as follows: Assuming the two UAVs are of the same type, their speeds are v and ... i and v j If the drones are descending at a constant speed, the separation distance between them must meet the following condition when they are about to reach the intersection of their tracks or are flying on overlapping tracks:

[0056] d int，i，j >2r s (8)

[0057] In the formula, d int，i，j The distance r represents the real-time distance between the two drones during landing. s Indicates the safe radius of the drone.

[0058] Step 2: Establish a 3D airport environment spatial model D for UAV formation landing using the grid method. The specific process is as follows:

[0059] (2.1) Simulate the airport terrain environment and establish the random terrain z1(x,y) as follows:

[0060]

[0061] In the formula, (x, y) are the horizontal projection coordinates of the ground point, and a, b, c, d, e, f, and g are all random terrain constants, with specific values ​​shown in Table 1.

[0062] Table 1 Random Terrain Parameter Settings

[0063]

[0064] (2.2) Seven mountain peak disturbances are established, and the mountain peak undulation height z2(x, y) is represented as:

[0065]

[0066] In the formula, h i Let x be the height of the i-th peak, (x) i y i Let x be the projection of the summit of the i-th mountain peak onto the horizontal plane. is y is The attenuation rates along the x and y directions of the i-th peak are respectively used to provide peak slope information, and the specific values ​​are shown in Table 2:

[0067] Table 2 Mountain Peak Model Parameter Settings

[0068]

[0069]

[0070] (2.3) The point containing the maximum value of the random terrain z1(x, y) and the mountain peak undulation height z2(x, y) is selected as the final height z(x, y), specifically:

[0071] z(x,y)=max(z1(x,y),z2(x,y)) (11)

[0072] (2.4) Divide the horizontal plane into 100×100 grids, with each intersection point corresponding to a z value, and store a total of 101×101 height information.

[0073] (2.5) Let the airport space D = {(x, y, z) | 0m≤x, y≤1000m, 0m≤z≤800m}. Through MATLAB simulation, the final three-dimensional environmental space model of the airport is established as follows: Figure 3 As shown.

[0074] Step 3: Move all ants to the initial point and define the initial state P of the drone. sThe parameters include the drone's approach point coordinates (0m, 850m, 500m), landing point coordinates (1000m, 200m, 0m), and the drone's approach flight capture phase, with the range controlled within [200m, 300m]. Initial parameters for the ant colony algorithm are set, including ant colony size N = 100, number of iterations K = 150, number of ants M = 50, maximum initial pheromone value τ0 = 10, pheromone evaporation coefficient ρ ∈ [0.2, 0.6], pheromone diffusion factor δ = 0.3, pheromone importance parameter α = 3, and heuristic function importance parameter β = 7.

[0075] Step 4, using distance induction factor D j_pt With angle-induced factor Ψ ij Combined design of heuristic function η ij Specifically:

[0076] η ij =D j_pt Ψ ij H j (12)

[0077] In the formula, D j_pt To measure the distance to the destination, the deviation of the drone's descent trajectory can be reduced, specifically:

[0078]

[0079] In the formula, r is a constant, assuming P is the current node, T is the landing point, and d j_pt Let j be the perpendicular distance from the next node j to the line PT, satisfying:

[0080]

[0081] Here, (x) p y p , z p ), (x J y J , z J ), (x T y T , z T These are the coordinates of the current node, the next node, and the landing point, respectively.

[0082] Ψ ij To measure the quality of corner turns, information that effectively avoids excessive downward curves can be obtained, specifically:

[0083]

[0084] here, Let be the turning angle of the drone from the current node i to the next node j.

[0085] H j The factor representing whether the next node j is a feasible node is as follows:

[0086]

[0087] Step 5, using the heuristic function η ij and pheromone τ ij Node probabilities are calculated, and the ant state transition is performed using the roulette wheel selection method. Selected nodes are stored in the tabu list (TABU) until the destination is reached and feasible paths are stored. The specific process is as follows:

[0088] (5.1) Calculate the selection probability of candidate nodes using the following formula.

[0089]

[0090] In the formula, τ ij η represents the pheromone concentration values ​​from current node i to node j. ij τ represents the heuristic function value from the current node i to node j. iq Let η be the pheromone intensity from the current node i to the candidate node q. iq Let α be the heuristic function value from the current node i to the candidate node q, β be the pheromone importance parameter, and β be the heuristic function importance parameter. m It is the set of nodes that have not been visited.

[0091] (5.2) Selection probability of candidate nodes The cumulative probability P(l) of the l-th candidate node is calculated as follows:

[0092]

[0093] Then, the roulette wheel selection method is used to select candidate nodes. The specific process is as follows: According to the roulette wheel selection method, a random number between [0, 1] is generated and compared with P(l), where the value of l ranges from 1 to F, and F is the total number of candidate nodes. When the random number is less than P(l), the individual l is selected.

[0094] (5.3) Starting from N=0, select path nodes according to step (5.2), that is, use the distance induction factor D. j_pt With angle-induced factor Ψ ij Combined heuristic function η ijNode probabilities are calculated, and a roulette wheel selection method is used to determine the drone's next direction. To avoid excessive gliding turns, a path node is selected every 5 steps, and the selected node is stored in the tabu list TABU. This continues until the ant reaches the destination (N=100), at which point a feasible path is stored in TABU.

[0095] Step 6: Starting from M=0, repeat steps 4 and 5 until all 50 ants have completed one iteration. After one iteration, pheromone updates are performed, introducing a mechanism combining pheromone diffusion and random pheromone evaporation. A global plus local pheromone update method is used for pheromone updates, specifically:

[0096] (6.1) Assume the random evaporation coefficient of pheromones is ρ m (N), satisfying:

[0097]

[0098] In the formula, N is the current iteration value of the m-th ant, ρ is the initial value of the pheromone evaporation coefficient, set to 0.4, ρ min ρ max α represents the upper and lower limits of the pheromone evaporation coefficient, which are 0.2 and 0.6 respectively, and β represents the upper and lower limits of the random number, which are 0.5 and 1.5 respectively.

[0099] (6.2) Assume that the pheromone update amount from node i to node j on the optimal path after the Nth iteration is Δτ. ij (N) min Introducing the Δτ mechanism of pheromone diffusion ij (N) min The expression is:

[0100]

[0101] In the formula, δ represents the pheromone diffusion factor, with a value of 0.3, S represents the pheromone increase on the optimal path, with a value of 10, and L... min This represents the optimal path length.

[0102] (6.3) The pheromone diffusion mechanism and random pheromone evaporation are introduced into the global plus local pheromone update method. Here, local pheromone update refers to updating pheromones on all feasible paths after completing one iteration, and global pheromone update refers to updating pheromones on the shortest path after completing one iteration. The specific update method is as follows:

[0103] Assume that the pheromone content from node i to node j is τ after the (N+1)th iteration. ij (N+1), specifically represented as:

[0104]

[0105] In the formula, ρ m (N) is the random evaporation coefficient of the pheromone after the Nth iteration, with a value range of [0.2, 0.6], τ ij (N) represents the pheromone content from node i to node j after the Nth iteration, and M represents the total number of ants. Let Δτ be the increase in pheromone produced by the m-th ant after the N-th iteration, representing the local update of the pheromone. ij (N) min Δτ represents the increase in pheromone from node i to node j on the optimal path after the Nth iteration, and Δτ represents the global update of pheromone. ij (N) min A pheromone diffusion mechanism was introduced.

[0106] Step 7: After reaching the maximum number of iterations K=150, output the best path for a single UAV landing. Then repeat steps 3 to 7 until multiple UAV landing paths are output for UAV formation landing path planning. The specific process is as follows:

[0107] (7.1) Use steps 3 to 7 to plan the best landing path for a single UAV.

[0108] (7.2) The landing trajectory of a single UAV is smoothed using a cubic B-spline curve. The specific smoothing method is described in step (1.1.4).

[0109] (7.3) Set the approach and landing points for four UAVs in the fixed-wing UAV formation, and plan the UAV formation landing trajectory according to the improved ant colony algorithm. The trajectory is smoothed as follows: Figure 7 As shown in Table 3, the approach and landing coordinates of the four UAVs are as follows:

[0110] Table 3. Coordinates of the approach and landing positions of the four UAVs

[0111]

[0112] Step 8: Determine whether the drone formation can land smoothly and safely based on the drone's own performance constraints and the multi-drone formation landing constraint criteria. The specific process is as follows:

[0113] (8.1) Based on the performance constraints of the UAV itself, determine whether the four UAVs meet the performance requirements in terms of total range, minimum flight segment, algebraic sum of turning angles, and maximum glide curvature according to equations (1) to (3) and (5).

[0114] (8.2) Based on the UAV formation landing constraint criteria (6) to (8), determine whether the four UAVs can satisfy the path non-intersection criterion. If this condition is met, the multi-UAV landing trajectory reaches the optimal efficiency, and both safety and reliability can be satisfied. This invention can then be used for auxiliary decision-making for high-efficiency UAV formation landing.

[0115] To verify the feasibility and effectiveness of the method, the present invention will be described in further detail below with reference to examples.

[0116] This invention is based on a 64-bit Windows 10 operating system with an Intel Core i7-8565 processor, a 1.80GHz CPU, and 8GB of RAM, and uses Matlab 2016a simulation software for simulation analysis. The simulation environment uses a 1000m × 1000m × 500m three-dimensional airport environment model, as shown below. Figure 3 As shown in the figure. In this three-dimensional environment, the airport ground is simulated using random terrain, with 7 mountain peaks representing 7 atmospheric disturbances. The planned flight path needs to avoid intersecting with these disturbances.

[0117] Figure 4 This is a 3D spatial planning route map based on the traditional ant colony algorithm. Figure 5 This is a 3D spatial planning route diagram for the improved ant colony algorithm of this invention. In the diagram, green asterisks represent approach points, yellow squares represent landing points, and red / magenta dashed lines represent the flight path. (Comparison) Figure 4 and Figure 5 It can be seen that the landing path planned by the traditional ant colony algorithm has many twists and turns, which reduces the landing efficiency of the UAV, while the landing path under the improved ant colony algorithm of the present invention is smoother.

[0118] Figure 6 This figure compares the iteration curves of the traditional ant colony algorithm and the improved ant colony algorithm of this invention for 3D spatial route planning. The orange solid line and the magenta dashed line in the figure represent the iteration curves of the traditional ant colony algorithm and the improved ant colony algorithm, respectively. It can be seen that the improved ant colony algorithm of this invention significantly reduces the minimum number of iterations and can converge quickly compared to the traditional ant colony algorithm, thus greatly improving the running efficiency.

[0119] Table 4 shows the comparison of the shortest flight distance, first-generation shortest flight distance, minimum number of iterations, and number of turning angles for route planning using the traditional and the improved ant colony algorithm of this invention in three-dimensional space.

[0120] Table 4. Performance comparison of traditional and improved ant colony algorithms in three-dimensional space.

[0121]

[0122] As shown in Table 4, the improved ant colony algorithm of this invention has better performance in all aspects of the UAV landing route than the traditional ant colony algorithm. The shortest flight distance is shortened by 3.5% compared with the traditional ant colony algorithm, the first-generation shortest flight distance is shortened by 7.5%, the number of turning angles is reduced by 37.3%, and the minimum flight segment is increased by 5.7%. Moreover, all performance aspects can meet the landing constraints.

[0123] Figure 7 This diagram illustrates the landing planning of four UAVs using the method described in this invention in three-dimensional space. All tracks have been smoothed using cubic B-spline curves. The diagram assumes UAV1 and UAV2 approach at an altitude of 500m, and UAV3 and UAV4 at 450m. The yellow dashed line represents the landing track of UAV1, the blue solid line represents the landing track of UAV2, the magenta dotted line represents the landing track of UAV3, and the red dotted line represents the landing track of UAV4. As shown in the diagram, the shortest distance between the tracks of UAV1 and UAV2 is approximately at x = 668.8m (17.5m); the shortest distance between the tracks of UAV3 and UAV4 is approximately at x = 400m (21.9m). Based on the 5m safety radius setting for fixed-wing UAVs, this distance satisfies the path non-intersection criterion and does not require replanning. The total range, minimum segment, algebraic sum of turning angles, and maximum curvature of the four UAVs are shown in Table 5.

[0124] Table 5. Landing planning results for four UAVs in three-dimensional space.

[0125]

[0126] As shown in Table 5, the total flight distance, minimum flight segment, algebraic sum of turning angles, and maximum curvature of the three-dimensional approach routes of the four UAVs can all meet the landing constraints of the UAVs, and the UAVs can achieve safe landing.

Claims

1. A method for assisting decision-making in the landing of fixed-wing unmanned aerial vehicle (UAV) formations, characterized in that, The method includes the following steps: Step 1: To ensure the smooth and safe landing of fixed-wing UAVs, determine the performance constraints of the UAVs during the landing phase and set landing constraint criteria for multi-UAV formations. Step 2: Establish a 3D airport environment spatial model D for drone formation landing; Step 3, define the initial state P of a single drone. S This includes the approach and landing coordinates, the distance during the approach and capture phase, moving all ants to the initial point, and setting the ant colony algorithm initialization parameters, including the ant colony size N, the number of iterations K, the number of ants M, the starting point location information, the maximum initial value of pheromone τ0, the pheromone volatility coefficient ρ, the pheromone diffusion factor δ, the pheromone importance parameter α, and the heuristic function importance parameter β. Step 4, using distance induction factor D j_pt With angle-induced factor Ψ ij Combined heuristic function η ij Node probability calculation is performed, specifically using the heuristic function η. ij satisfy: or ij =D j_pt P ij H j (1) In the formula, D j_pt To measure the distance to the destination, the specific information is as follows: In the formula, r is a constant, assuming P is the current node, T is the landing point, and d j_pt Let j be the perpendicular distance from the next node j to the line PT, satisfying: Here, (x) P y P , z P ), (x j y j , z j ), (x T y T , z T These are the coordinates of the current node, the next node, and the landing point, respectively. Ψ ij To measure the quality of corner turns, information that effectively avoids excessive downward curves can be obtained, specifically: here, Let $\mathbf$ be the turning angle of the drone from the current node $i$ to the next node $j$. H j The factor representing whether the next node j is a feasible node is as follows: Step 5: Use the roulette wheel selection method to perform ant state transitions, and store the selected nodes in the tabu list TABU until the search reaches the destination and the feasible path is stored. Step 6: Introduce a mechanism combining pheromone diffusion and random pheromone evaporation, and use a global plus local pheromone update method for pheromone updating, specifically: (6.1) Assume the random evaporation coefficient of pheromones is ρ m (N), satisfying: In the formula, N is the current iteration value of the m-th ant, ρ min ρ max ρ represents the upper and lower limits of the pheromone evaporation coefficient, a and b represent the upper and lower limits of random numbers, and ρ represents the initial value of the pheromone evaporation coefficient. (6.2) Assume that the pheromone update amount from node i to node j on the optimal path after the Nth iteration is Δτ. ij (N) min This introduces the Δτ mechanism of pheromone diffusion. ij (N) min Specifically, it can be expressed as follows: In the formula, δ represents the pheromone diffusion factor, S represents the increase in pheromone along the optimal path, and L... min The optimal path length; (6.3) The pheromone diffusion mechanism and random pheromone evaporation are introduced into the global plus local pheromone update method. Here, local pheromone update refers to updating pheromones on all feasible paths after completing one iteration, and global pheromone update refers to updating pheromones on the shortest path after completing one iteration. The specific update method is as follows: Assume that the pheromone content from node i to node j is τ after the (N+1)th iteration. ij (N+1), specifically represented as: In the formula, ρ m (N) is the random evaporation coefficient of the pheromone after the Nth iteration, τ ij (N) represents the pheromone content from node i to node j after the Nth iteration, and M represents the total number of ants. Let Δτ be the pheromone update amount generated by the m-th ant after the N-th iteration, representing the local update of the pheromone. ij (N) min Δτ represents the pheromone update from node i to node j on the optimal path after the Nth iteration, and Δτ represents the global pheromone update. ij (N) min A pheromone diffusion mechanism was introduced; Step 7: After reaching the maximum number of iterations, output the best path for landing a single UAV, and plan the UAV formation landing path based on the landing paths of multiple UAVs output in Steps 3 to 7. Step 8: Determine whether the drone formation can land smoothly and safely based on the drone's own performance constraints during the landing phase and the multi-drone formation landing constraint criteria.

2. The fixed-wing UAV formation landing auxiliary decision-making method according to claim 1, characterized in that, Step 1: Determine the inherent performance constraints of the UAVs during the landing phase and set UAV formation landing constraint criteria, specifically: (2.1) Performance constraints of the UAV during the landing phase, including four parts: UAV range constraint, maximum turning angle constraint, maximum glide angle constraint, and maximum glide curvature constraint, specifically: (2.1.1) The UAV range constraint refers to the range l of the UAV's landing distance. i The total range L cannot be less than a certain threshold, and the total flight distance L cannot be greater than a certain threshold, that is: In the formula, l min L represents the minimum range limit for each segment. max This indicates the maximum limit on the total landing distance, and n represents the total number of flight segments; (2.1.2) The maximum turning angle constraint of the UAV refers to the requirement that the maximum turning angle ψ at each node it passes through must satisfy the following condition: In the formula, This represents the range vector of the i-th segment. Let Ψ represent the range vector of the (i+1)th segment, and let Ψ represent the maximum turning angle of the UAV when turning. (2.1.3) The maximum glide angle constraint for the UAV refers to the constraint that the UAV can only glide from node (x) at a maximum glide angle of 100°. i y i , z i ) to node (x i+1 y i+1 , z i+1 The maximum glide slope angle λ of this trajectory must satisfy the following condition: In the formula, x i y i z i x i+1 y i+1 z i+1 λ represents the horizontal, vertical, and angular coordinates of the i-th and (i+1)-th nodes of the UAV, respectively, λ represents the maximum glide angle, and n represents the total number of nodes; (2.1.4) The maximum glide curvature constraint for the UAV refers to the requirement that the curvature of the UAV during glide turns must be less than a certain threshold. First, the planned landing polygonal trajectory is smoothed using a B-spline curve. Then, the smoothed trajectory is described as a three-dimensional spatial curve: x = x(t), y = y(t), z = z(t). The first derivative of this three-dimensional spatial curve is obtained. Taking the second derivative again yields Treat the three first derivatives as a three-dimensional vector. Treat the three second derivatives as a three-dimensional vector. The maximum curvature K of the drone during its descent max The constraints that must be met are: In the formula, k(t) represents the glide curvature of the UAV. This represents the first-order derivative three-dimensional vector of the UAV's flight path curve. K represents the second-order derivative three-dimensional vector of the UAV's trajectory curve. max This represents the maximum bending rate of the drone during its descent. (2.2) Set landing constraints for multi-UAV formations, including path non-intersection criteria and UAV separation distance criteria, specifically: (2.2.1) Path non-intersection criterion Assuming each fixed-wing UAV is considered a point mass and its safe range is considered a circle, then the shortest distance d between the trajectories of two UAVs is... safe_min It should be greater than the sum of the safe radii of the two drones, specifically satisfying the following: d safe_min ＞r s_i +r s_j (13) In the formula, r s_i r s_j Let be the safe radii of the i-th and j-th drones, respectively; (2.2.2) UAV separation distance criterion Assuming the two drones are the same model, and each has a speed v i and v j If the drones are descending at a constant speed, the separation distance between them must meet the following condition when they are about to reach the intersection of their tracks or are flying on overlapping tracks: d int，i，j ＞2r s (14) In the formula, d int，i，j The distance r represents the real-time distance between the two drones during landing. s Indicates the safe radius of the drone.

Images