Complex mountainous area coverage search method based on hybrid strategy

Abstract

The invention relates to the technical field of unmanned aerial vehicle path planning, and provides a complex mountainous area coverage search method based on a hybrid strategy, which comprises the following steps: step 1, initialization and fixed search: performing area coverage by taking a reciprocating path as a reference, detecting passable points of a current node neighborhood in real time, and storing the passable points in a queue; step 2, dead point processing and path optimization: when no unsearched point exists in the neighborhood, triggering a queue dequeue operation, taking the recently recorded dead point as a new starting point, and calling an improved particle swarm algorithm to plan a cross-regional safety path; the improved particle swarm algorithm balances global exploration and local development through adaptive inertia weight, ensures that the path is shortest and avoids terrain shielding; and step 3, loop termination: the steps are repeated until the queue is empty, and global coverage is completed. According to the invention, efficient and safe complex terrain coverage search can be realized.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) path planning technology, and more specifically, to a complex mountainous area coverage search method based on a hybrid strategy. Background Technology

[0002] In recent years, with the widespread application of UAV technology in emergency rescue, disaster monitoring, and other fields, its coverage search capability in complex terrain has become a research hotspot. Traditional fixed search methods (such as "Z-shaped" and "spiral" paths) are prone to path redundancy and terrain obstacles in mountainous environments due to their fixed trajectories. Especially when encountering dead points, they need to backtrack, resulting in a large number of repeated paths. At the same time, simply using heuristic algorithms such as genetic algorithms, RRT, and artificial bee colony algorithms is subject to significant computational power and planning timeliness issues. However, their high computational complexity and sensitivity to parameters make them difficult to adapt to the dynamic changes in mountainous terrain.

[0003] For example, the existing literature "UAV Full Coverage Path Planning in 3D Terrain and Wind Field Environment" proposes a wind field and terrain adaptive full coverage path planning algorithm. This method first models the environment using a grid method, constructing airway points based on elevation data points within the grid. Next, it analyzes the maneuvering actions and energy functions of the rotor-driven UAV in the wind field environment to construct a UAV wind model and an energy consumption model. Then, based on rolling optimization and global concepts, it proposes a neighbor-point rolling exploration-global prediction fusion strategy as the core of the algorithm, planning a near-optimal full coverage path for the UAV that adapts to the wind field and undulating terrain. Finally, simulation experiments were conducted in different environments. The experimental results demonstrate the necessity of considering wind field and terrain, and compared to single reciprocating coverage, it can adapt to environmental factors, thus significantly reducing energy consumption during full coverage. However, its consideration of obstacles or dead points is based on avoidance. While the terrain complexity is relatively low, it achieves full coverage through avoidance without considering the situation where the UAV has to backtrack when entering local dead zones.

[0004] The existing literature, "A Vehicle-Mounted Multi-UAV Cooperative Multi-Area Coverage Path Planning Method," proposes a vehicle-mounted multi-UAV cooperative mode for large-area multi-area coverage scanning tasks. In this mode, the vehicle can act as a mobile base station for UAVs, cooperating with multiple UAVs to complete coverage scanning tasks across multiple large areas. After fully analyzing the characteristics of the new problem, a 0-1 integer programming model is established to optimize the vehicle's ground driving path and the cooperative aerial flight path of the multi-UAVs. A three-stage intelligent optimization algorithm is proposed, which first plans the multi-UAV area coverage path and the vehicle cooperative path to quickly construct feasible solutions, and then optimizes the feasible solutions based on an adaptive large-scale neighborhood search algorithm. However, this method requires the use of other vehicles for the UAVs to scan line by line according to traditional rules of coverage search and escape local dead points through heuristic search.

[0005] Patent document CN106406346A discloses a method for planning a multi-UAV cooperative rapid coverage search trajectory. First, based on prior information about the battlefield environment and the geometric features of the target area, the gray area is extracted into three types of key search targets: point, line, and area targets. Then, a dual-encoding strategy integer genetic algorithm is used to determine the access order of each key search target. Finally, considering the allocation order of key search targets and UAVs, a Dubins path and greedy strategy are used to obtain the local shortest connection trajectory between the end point of the current target coverage search trajectory and the starting point of the next target coverage search trajectory, thus deriving the coverage search trajectory for the next target, thereby obtaining the multi-UAV cooperative coverage search trajectory. This invention can plan a cooperative rapid coverage search trajectory for multiple UAVs in a specific gray area with short coverage time and high algorithm robustness. However, it requires consideration of the gray area, making it relatively complex.

[0006] In summary, the existing technology has the following problems:

[0007] 1) Fixed search mode suffers from severe path redundancy and poor terrain adaptability.

[0008] Fixed-path searches rely on preset trajectories (such as "Z-shaped" or "spiral" paths), and their path planning logic is rigid, making them unable to dynamically respond to complex terrain changes. When a drone encounters mountain obstructions or unconnected areas (dead points), it can only backtrack to the previous node along the original path, resulting in high path repetition rates and low search efficiency. This is especially problematic in nested dead-point scenarios, where it severely wastes energy and mission time.

[0009] 2) Heuristic algorithms lack real-time performance and have weak robustness in dynamic environments.

[0010] Existing heuristic algorithms (such as genetic algorithms and particle swarm optimization) require frequent iterative calculations and complex parameter tuning, resulting in high computational complexity and slow response speed. In dynamic mountainous environments (such as sudden obstacles and airflow disturbances), these algorithms struggle to quickly generate safe and feasible paths and are prone to path oscillations or planning failures due to parameter sensitivity issues.

[0011] 3) Imbalance between local and global optimization, easily leading to suboptimal solutions.

[0012] Traditional methods often focus on local path optimization (such as obstacle avoidance and turning), lacking a comprehensive consideration of global coverage efficiency. For example, some algorithms can avoid individual obstacles, but because they do not consider the continuity of the overall path, the global path length increases significantly; other algorithms converge too early and get stuck in local optima, failing to achieve efficient global coverage.

[0013] 4) The dead-point backtracking mechanism is rigid, and the path replanning capability is lacking.

[0014] Existing methods typically employ a single backtracking rule (such as returning in a fixed order) when dealing with complex dead points, lacking the ability to dynamically optimize cross-regional paths. When a UAV needs to jump from the current dead point to a distant, unexplored area, traditional backtracking mechanisms only support returning along the original path or random exploration, resulting in detours, reduced safety, and even mission interruption.

[0015] Therefore, a complex mountainous area coverage search method based on a hybrid strategy is needed to solve the above problems. Summary of the Invention

[0016] The present invention provides a hybrid strategy-based method for complex mountainous terrain cover search, which can solve the problems of redundancy and terrain obstruction in traditional fixed search paths. This method integrates reciprocating fixed path search with the Improved Particle Swarm Optimization (IPSO) algorithm, combined with queue and stack backtracking mechanisms, to achieve efficient and safe complex terrain cover search.

[0017] A complex mountainous area coverage search method based on a hybrid strategy according to the present invention includes the following steps:

[0018] Step 1, Initialization and Fixed Search: Based on the previous compound path, perform area coverage, detect passable points in the neighborhood of the current node in real time, and enqueue and store the passable points;

[0019] Step 2, Dead Point Handling and Path Optimization: When there are no unsearched points in the neighborhood, a queue dequeue operation is triggered, the most recently recorded dead point is used as the new starting point, and the improved particle swarm algorithm (heuristic algorithm) is called to plan a safe cross-region path; the improved particle swarm algorithm balances global exploration and local development through adaptive inertia weights to ensure the shortest path and avoid terrain occlusion.

[0020] Step 3, Loop Termination: Repeat the above steps until the queue is empty, completing the full coverage.

[0021] As a preferred option, the improved particle swarm optimization algorithm stipulates that the line connecting two adjacent path points of the UAV cannot pass through obstacles formed by mountains exceeding a specified height, with a penalty value C. O As shown in the following formula:

[0022]

[0023] Where C O The penalty value of the current constraint, p i For points in the path, M(p) i Let t be a point on the binarized map, and 255 represent the white, passable area. i , t i+1() represents the coordinates of the interpolation insertion point between two points;

[0024] Path points cannot be inside obstacles; the penalty value is C. p As shown in the following formula:

[0025]

[0026] Where C p This represents the penalty value for the current constraint.

[0027] During drone flight, it is necessary to ensure that the drone maintains a safe distance from obstacles and to apply a corresponding penalty, the penalty value of which is C. s As shown in the following formula:

[0028]

[0029] Where C s The penalty value for the constraint, d(p) i o j ) represents the distance between the drone's position coordinates and the obstacle. If the distance is greater than the safe distance R, no penalty value is accumulated; otherwise, the distance difference is multiplied by the penalty coefficient λ.

[0030] The constraints are integrated into an objective function, as shown in the following equation:

[0031] ρ=α·L path +β·C o +ε·C s +δ·C p

[0032] Where α, β, ε, and δ represent the penalty weights of each penalty term, ρ is the sum of the terms in the objective function, and L... path This outputs the total length of the path points.

[0033] As a preferred approach, the improved particle swarm optimization algorithm applies an adaptive inertia weight based on exponential decay. In the early stages of iteration, a larger inertia value is assigned, enabling it to quickly explore feasible paths and overcome local constraints. In the later stages of iteration, a smaller inertia value is assigned, guiding the algorithm to focus on a refined exploration of the current optimal region. Its mathematical expression is shown below:

[0034]

[0035] Where ω t ω is the inertia value for the current iteration number. min ω max The minimum and initial values ​​of the inertia weights are given by t. max This represents the maximum number of iterations.

[0036] The beneficial effects of this invention are as follows:

[0037] 1) Significantly reduces path redundancy. Experiments show that the average path length is shortened by 21.1%, and the coverage rate steadily increases to 100%.

[0038] 2) This invention can improve terrain adaptability. Traditional UAV coverage search methods are prone to getting stuck in local dead points. If the traditional backtracking method is used, it will cause route redundancy and waste overall energy consumption. However, this invention can be applied to complex mountainous terrain with many obstacles and many local areas. This invention can achieve fast convergence by improving the PSO algorithm.

[0039] 3) This invention considers complex map environments under various working conditions and compares them with traditional heuristic algorithms. The fluctuation amplitude is significantly reduced and the gap between the maximum and minimum values ​​is also greatly compressed, indicating that this invention has better robustness and convergence stability, and can effectively avoid obstacles even under extreme terrain conditions.

[0040] 4) Optimize the backtracking mechanism and automatically select queue or stack backtracking strategies based on different terrain features to improve overall search efficiency.

[0041] This invention is applicable to scenarios such as emergency rescue and topographic mapping, significantly improving the search efficiency and safety of UAVs in complex mountainous areas, and has broad application prospects. Attached Figure Description

[0042] Figure 1 This is a flowchart of a complex mountainous area coverage search method based on a hybrid strategy, as described in this embodiment.

[0043] Figure 2 This is a comparison of the path lengths of the two basic methods, queue backtracking (T1) and stack backtracking (T2), after the introduction of heuristic algorithm optimization in the embodiment.

[0044] Figure 3 This is a schematic diagram illustrating the coverage growth of the hybrid strategy in the embodiment;

[0045] Figure 4(a) shows the planned path result of the custom terrain region 1 in the embodiment;

[0046] Figure 4(b) shows the planned path result of the custom terrain region 2 in the embodiment;

[0047] Figure 4(c) shows the planned path result of the custom terrain region 3 in the embodiment;

[0048] Figure 5(a) shows the coverage trend of queue backtracking (T1) in the embodiment;

[0049] Figure 5(b) shows the coverage trend of stack backtracking (T2) in the embodiment;

[0050] Figure 6This is a comparison chart of path lengths in the examples. Detailed Implementation

[0051] To further understand the content of this invention, a detailed description of the invention will be provided in conjunction with the accompanying drawings and embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0052] Example

[0053] like Figure 1 As shown, this embodiment provides a method for complex mountainous area coverage search based on a hybrid strategy, which includes the following steps:

[0054] Step 1, Initialization and Fixed Search: Based on the previous compound path, perform area coverage, detect passable points in the neighborhood of the current node in real time, and enqueue and store the passable points;

[0055] Step 2, Dead Point Handling and Path Optimization: When there are no unsearched points in the neighborhood, a queue dequeue operation is triggered, the most recently recorded dead point is used as the new starting point, and the improved particle swarm algorithm (heuristic algorithm) is called to plan a safe cross-region path; the improved particle swarm algorithm balances global exploration and local development through adaptive inertia weights to ensure the shortest path and avoid terrain occlusion.

[0056] Step 3, Loop Termination: Repeat the above steps until the queue is empty, completing the full coverage.

[0057] In the improved particle swarm optimization algorithm, it is stipulated that the line connecting two adjacent path points of the UAV cannot pass through obstacles formed by mountains exceeding a specified height, with a penalty value C. o As shown in the following formula:

[0058]

[0059] Where C O The penalty value of the current constraint, p i For points in the path, M(p) i Let t be a point on the binarized map, and 255 represent the white, passable area. i , t i+1 () represents the coordinates of the interpolation insertion point between two points;

[0060] Path points cannot be inside obstacles; the penalty value is C. p As shown in the following formula:

[0061]

[0062] Where C p This represents the penalty value for the current constraint.

[0063] During drone flight, it is necessary to ensure that the drone maintains a safe distance from obstacles and to apply a corresponding penalty, the penalty value of which is C. s As shown in the following formula:

[0064]

[0065] Where C s The penalty value for the constraint, d(p) i o j ) represents the distance between the drone's position coordinates and the obstacle. If the distance is greater than the safe distance R, no penalty value is accumulated; otherwise, the distance difference is multiplied by the penalty coefficient λ.

[0066] The constraints are integrated into an objective function, as shown in the following equation:

[0067] ρ=α·L path +β·C o +ε·C s +δ·C p

[0068] Where α, β, ε, and δ represent the penalty weights of each penalty term, ρ is the sum of the terms in the objective function, and L... path This outputs the total length of the path points.

[0069] In the improved particle swarm optimization algorithm, an adaptive inertia weight based on exponential decay is applied. A larger inertia value is assigned in the early stages of iteration to enable it to quickly explore feasible paths and overcome local constraints. A smaller inertia value is assigned in the later stages of iteration to guide the algorithm to focus on a refined exploration of the current optimal region. Its mathematical expression is shown below:

[0070]

[0071] Where ω t ω is the inertia value for the current iteration number. min ω max The minimum and initial values ​​of the inertia weights are given by t. max This represents the maximum number of iterations.

[0072] This embodiment achieves efficient and secure complex terrain coverage search by integrating reciprocating fixed-path search with an improved particle swarm optimization algorithm, combined with a queue and stack backtracking mechanism. The queue and stack backtracking mechanism works as follows: during fixed-path search, the neighborhood of the flight path is searched; if there are still uncovered paths, the current path point is pushed onto a stack / enqueued. When a dead point is encountered, it is popped from the stack / dequeued in a last-in-first-out (LIFO) order to serve as the starting point for the next search region. A heuristic algorithm is used to determine the starting point for the next search region.

[0073] The hybrid strategy coverage search method proposed in this embodiment requires initializing the queue S to be empty and marking the first starting point of the search matrix book as True during the initialization phase. For the current node P i Iterate through its adjacent nodes, and if there is a blank node, then add P. i Enqueue. When the current area is explored and a dead end is encountered, preventing further search, the element is dequeued according to the first-in, first-out (FIFO) rule and becomes the starting point for the next search stage. Traditional fixed-path search requires backtracking from the path Ab due to terrain obstacles. The hybrid strategy proposed in this embodiment introduces a heuristic path planning algorithm to improve backtracking efficiency when dealing with such cross-regional connectivity issues. Finally, when queue S is empty, it means that the entire connected region has been searched, and the algorithm stops execution. The complete implementation logic and process framework of the algorithm are shown below:

[0074]

[0075] Experimental Analysis

[0076] Figure 2 This paper presents a comparison of path lengths for two basic methods, queue backtracking (T1) and stack backtracking (T2), after incorporating heuristic algorithm optimization. The T1 hybrid strategy, after optimizing the backtracking path using the IPSO algorithm, reduces the total path length by 2.41% compared to the queue backtracking method. The T2 hybrid strategy, however, shows relatively limited path optimization for the stack backtracking method, achieving only a slight improvement of 0.048%. The coverage growth of the hybrid strategies is shown below. Figure 3 As shown in the figure, the coverage rate increases slowly and steadily as the drone flies.

[0077] To verify the quality of the paths planned by the coverage search under more complex conditions, three more complex map environments were customized, and the planned paths are shown in Figures 4(a), 4(b), and 4(c). As can be seen from the figures, the planned paths can still completely cover the search area when the map becomes more complex (the complexity of dead point locations increases). However, from the coverage change trend graphs shown in Figures 5(a) and 5(b), it is found that when the map becomes more complex, the coverage no longer increases steadily but becomes more tortuous. This is because the algorithm needs to repeatedly fly through and explore blank areas in some narrow dead point spaces. In the first 70 steps, the algorithm does not encounter dead point backtracking, resulting in extremely similar coverage growth curves in the three regions. In subsequent paths, the different backtracking mechanisms lead to changes in the search strategy, resulting in significant differences.

[0078] from Figure 6The path length comparison results show that, regardless of whether it's stack backtracking or queue backtracking, the path length fluctuates significantly under different terrain scenarios, indicating that the fixed search pattern lacks universal adaptability to complex terrain. When the IPSO algorithm is introduced to optimize cross-regional backtracking paths, the total path length of the hybrid strategy is shortened by an average of 21.1%. Further combining... Figure 3 and Figure 6 Comparative analysis reveals that under the hybrid strategy framework, the path optimization capability of queue backtracking is significantly weakened, with an average optimization rate of only 0.66%, far lower than the 31.8% optimization effect of stack backtracking. This difference may stem from the first-in, first-out (FIFO) characteristic of queue backtracking. Since the next search starting point is always located in the vicinity of the current dead point, the cross-regional path optimization space planned by the IPSO algorithm is limited.

[0079] The present invention and its embodiments have been described above illustratively. This description is not restrictive, and the figures shown are only one embodiment of the present invention; the actual structure is not limited thereto. Therefore, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the present invention, such designs should fall within the protection scope of the present invention.

Claims

1. A hybrid strategy based complex mountainous coverage search method, characterized in that: Comprising the following steps: Step 1, initialization and fixed search: cover the area based on the reciprocating path, detect the passable points of the current node neighborhood in real time, and store the passable points in the queue; Step 2, dead point processing and path optimization: when there is no unsearched point in the neighborhood, trigger the queue dequeue operation, take the recently recorded dead point as a new starting point, and call the improved particle swarm algorithm to plan a safe path across the region; the improved particle swarm algorithm balances global exploration and local development through adaptive inertia weight, ensuring the shortest path and avoiding terrain obstructions; Step 3, loop termination: repeat the above steps until the queue is empty, completing global coverage.

2. The method of claim 1, wherein the method is a hybrid strategy based complex mountainous coverage search method. In the improved particle swarm algorithm, it is stipulated that the line connecting two adjacent path points of the UAV cannot pass through the obstacle formed by the mountain with a height exceeding a specified height, and the penalty value C of the line is O as shown in the following formula: where C O the current constraint penalty value, p i for a point in the path, M(p i ) is a certain point of the binarized map, 255 is white, i.e. a passable area, L(t i , t i+1 ) is the coordinate of the interpolated point between two points. The path point cannot be inside the obstacle, its penalty value C p As shown in the following formula: where C p is the current constraint penalty value result; In the process of UAV flight, the UAV needs to keep a safe distance from obstacles and apply corresponding penalties, and the penalty value C s As shown in the following formula: where C s is a penalty value for the constraint, d(p i , o j ) is the distance of the position coordinates of the drone from the obstacle, if the distance is greater than the safety distance R then the penalty value is not accumulated, otherwise the distance difference is multiplied by the penalty coefficient λ; Integrate the constraints into the objective function as follows: p = a - L path + β - C o + ε - C s + δ - C p where a, b, e, d represent the penalty weight of each penalty term, p is the sum of each term of the objective function, L path is the total length of the output path points.

3. The method of claim 2, wherein: In the improved particle swarm algorithm, an adaptive inertia weight based on exponential decay is applied, giving it a larger inertia value at the beginning of iteration to break through local constraints and quickly explore feasible paths; a smaller inertia value is given in the later iteration to guide the algorithm to focus on fine exploration of the current optimal area, which is mathematically expressed as follows: where ω t is the inertia value for the current iteration, ω min , ω max are the minimum and initial values of the inertia weight, t max is the maximum number of iterations.

Images