Unmanned aerial vehicle real-time path planning method for environment exploration

Abstract

The invention discloses an environment exploration-oriented unmanned aerial vehicle real-time path planning method, and relates to the field of unmanned aerial vehicle path planning, and the method comprises the following steps: S1, extracting topological nodes; s2, carrying out Delou inner triangulation and collision detection; s3, characterizing the topological map in a numeralization manner; and S4, path planning and trajectory optimization. According to the unmanned aerial vehicle real-time path planning method for environment exploration, the calculation bottleneck of a traditional graph search algorithm in a high-dimensional space can be overcome, redundant map preprocessing operation can be avoided, intelligent multiplexing of existing information is achieved, path safety and optimality are guaranteed, the planning efficiency and the system response speed are greatly improved, and the real-time path planning method for the unmanned aerial vehicle is suitable for being popularized and applied. Through intelligent multiplexing of structured information in a historical global map and collection of topological features and past successful path experience, the unmanned aerial vehicle can rapidly adapt to and optimize path decisions in similar scenes, repeated calculation is reduced, and the self-adaptive optimization effect that the unmanned aerial vehicle is more intelligent than the unmanned aerial vehicle is used is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle path planning, in particular to an unmanned aerial vehicle real-time path planning method for environment exploration. BACKGROUND

[0002] In the field of unmanned aerial vehicle autonomous navigation and environment exploration, real-time path planning is one of the core technologies to ensure the safety and efficiency of the task completion of the unmanned aerial vehicle. In particular, in an unknown or partially known environment, the unmanned aerial vehicle needs to generate a safe and feasible path from the current position to the target point quickly while dynamically perceiving the surrounding obstacles. With the development of sensor technology, modern unmanned aerial vehicles are usually equipped with laser radars, depth cameras or stereo vision systems, which can construct a high-resolution grid map of the local environment in real time, and maintain a constantly updated historical global map in combination with the SLAM (Simultaneous Localization and Mapping) technology. These prior information provides valuable data basis for path planning.

[0003] The patent with the authorized publication number CN114138006A discloses a kind of unmanned aerial vehicle path planning method and equipment for urban low-altitude environment, first obtain the area that unmanned aerial vehicle base station signal cannot cover, then utilize the planning space model of MAKLINK two-dimensional path construction, in combination with Dijkstra algorithm and ant colony algorithm, quickly build the shortest flight path of unmanned aerial vehicle, so that the cost of unmanned aerial vehicle task execution is reduced, and the safety problem brought by unmanned aerial vehicle entering the area without base station signal in complex urban low-altitude environment is avoided.

[0004] The patent with the authorized publication number CN110146085B discloses a kind of unmanned aerial vehicle real-time avoidance re-planning method based on mapping and fast exploration random tree, on the basis of local map, adopts the path planning method of fast exploration random tree based on sampling, finds out the key position point of unmanned aerial vehicle obstacle avoidance path, and utilizes uniform Bézier curve to convert the path composed of path key position point line segment into smooth and dynamic flyable trajectory. The method proposed in the present application is simple, and can be used in real-time obstacle avoidance of unmanned aerial vehicle on-board system with limited computing resources.

[0005] The above patents and traditional path planning methods face significant challenges in processing such large-scale, high-dimensional perception data. Although classical graph search algorithms such as Dijkstra perform well in small-scale or simple structured environments, their computational complexity increases exponentially with the resolution of the map when dealing with dense grid maps, resulting in significant planning delays and difficulty in meeting real-time requirements for unmanned aerial vehicles in high-speed flight or complex dynamic environments. To ensure the safety of the path, an inflation process is usually performed on the original grid map, which marks the free space within a certain range around the obstacle as an impassable area to provide collision buffering. This preprocessing step not only increases the overall computational overhead but also is repeatedly executed during frequent replanning, further reducing system efficiency.

[0006] More critically, existing methods often fail to fully exploit and utilize structured information in historical global maps, such as environmental topological features accumulated in multiple exploration tasks, common obstacle distribution patterns, and experience knowledge from past successful paths. These are usually simply discarded or only used for map initialization, lacking effective mechanisms to integrate them into the current path decision-making process, which leads to repeated extensive computation by the system in similar scenarios and prevents the realization of adaptive optimization capabilities that become smarter with use.

[0007] To address the above problems, it is necessary to innovate and design based on existing planning methods. SUMMARY

[0008] The present application aims to provide an unmanned aerial vehicle real-time path planning method for environmental exploration to solve the problem of lacking effective mechanisms to integrate them into the current path decision-making process, which leads to repeated extensive computation by the system in similar scenarios and prevents the realization of adaptive optimization capabilities that become smarter with use.

[0009] To achieve the above-mentioned purpose, the present application provides the following technical solution: An unmanned aerial vehicle real-time path planning method for environmental exploration, comprising the following steps:

[0010] Step S1: Topological node extraction;

[0011] During the flight of the unmanned aerial vehicle, first, its equipped laser radar or visual sensor is used to obtain environmental information, and these data are combined with the existing global octree grid map to update the local perception information in real time;

[0012] Step S2: Delaunay triangulation and collision detection;

[0013] The Delaunay triangulation algorithm is used to process the collected topology node information set, and a continuous triangular network structure is created using the topology node information. When a drone collides, the information of the corresponding edge in the topology graph is deleted.

[0014] Step S3: Numericalize the topological map;

[0015] The topology map is converted into a numerical form, and a feature matrix is ​​built using the topology vertex information, which makes it easy to quickly view the route through the feature matrix.

[0016] Step S4: Path planning and trajectory optimization;

[0017] The Dijkstra algorithm is used to find the optimal route forward, and inflection points with large information differences in the path trajectory are smoothed to improve the stability of the flight path and ensure the safe arrival of the UAV.

[0018] Preferably, in step S1, when the UAV identifies new boundary points, these points are added to the current topology as potential key topology nodes.

[0019] Preferably, in step S1, the connectivity of the topology graph is further enhanced by identifying inflection points in the path planning process.

[0020] Preferably, in step S1, the UAV samples the optimal boundary points within its sensor range to guide the subsequent exploration direction. After each global path planning is completed, only the data verified as key nodes are retained, ensuring the conciseness and effectiveness of the topology node set.

[0021] Preferably, in step S2, the collected set of topological nodes is processed using the Delaunay triangulation algorithm to create a continuous triangular network structure. Whenever a new node is added to the topological structure, the mesh layout is adjusted by point-by-point insertion to ensure the quality of the triangulation.

[0022] Preferably, in step S2, each topological edge is expanded according to the physical size of the UAV in response to obstacle constraints in the actual environment, and a global octree grid map is used for accurate collision detection. If any collision is found, the corresponding edge will be considered invalid and removed from the topological graph, thereby ensuring that the retained edges all represent feasible collision-free paths.

[0023] Preferably, in step S3, the constructed topological map is converted into a numerical form to facilitate subsequent path search. Each topological vertex contains information such as its coordinate position, connected edges, and the Delaunay triangle to which it belongs.

[0024] Preferably, in step S3, a feature matrix is ​​established using the information collected from the topological vertices to represent the connection relationship between nodes. This matrix stores the information of all adjacent nodes in the form of a two-dimensional array. When performing path search, all possible forward routes can be quickly found by querying the corresponding part of a specific node in the feature matrix, which greatly improves the efficiency of graph traversal.

[0025] Preferably, in step S4, Dijkstra's algorithm or other suitable shortest path algorithm is applied to the optimized topology map to find the best path from the current location to the target point.

[0026] Preferably, in step S4, considering the problem that there may be too many inflection points in actual flight leading to a non-optimal flight trajectory, the preliminary planning result is smoothed to reduce unnecessary turning times, improve the smoothness of the flight path and the stability of motion control, and ensure that the UAV can safely and efficiently reach its destination.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] 1. It can overcome the computational bottleneck of traditional graph search algorithms in high-dimensional space, avoid redundant map preprocessing operations, and realize the intelligent reuse of existing information, thereby significantly improving planning efficiency and system response speed while ensuring path safety and optimality.

[0029] 2. It can effectively integrate local perception information with prior knowledge in historical global maps, overcoming the problems of low computational efficiency, long map preprocessing time, and insufficient utilization of prior information in traditional graph search algorithms under large-scale dense grid maps. This enables efficient and safe real-time path planning in complex and unknown environments, improving the intelligence level and response speed of UAV autonomous navigation.

[0030] 3. Furthermore, it not only improves the efficiency and real-time performance of UAV path planning in large-scale unknown environments, but also enhances the safety and reliability of the entire exploration process;

[0031] 4. Furthermore, by intelligently reusing structured information from historical global maps, collecting topological features and past successful path experiences, the UAV can quickly adapt to and optimize path decisions in similar scenarios, reducing redundant calculations and achieving an adaptive optimization effect that "gets smarter the more it is used."

[0032] 5. Furthermore, the Delaunay triangulation and collision detection mechanism ensures that the UAV can perceive and avoid obstacles in real time in dynamic environments. At the same time, the mesh layout is adjusted by point-by-point insertion, which ensures the quality of triangulation and further improves the accuracy of path planning.

[0033] 6. Furthermore, the establishment of numerical representation topology maps and feature matrices provides an efficient data structure for path search, enabling UAVs to quickly find all possible routes forward, greatly improving graph traversal efficiency and providing strong support for real-time path planning. Attached Figure Description

[0034] Fig. 1 This is a schematic diagram of the structure of the present invention;

[0035] Fig. 2 This is a schematic diagram illustrating the principle and flow of the method of the present invention;

[0036] Fig. 3 This is a schematic diagram of the graphical state in the method of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] This implementation example Figs. 1-3 As shown, to address the issues of low efficiency and insufficient utilization of prior knowledge in existing technologies for large-scale dense maps, the process is disclosed, demonstrating that this method can improve planning efficiency and system response speed. The specific content includes the following steps:

[0039] Step S1: Extract topology nodes;

[0040] During the flight of the drone, the environmental information is first acquired using the lidar or visual sensors it is equipped with, and this data is combined with the existing global octree grid map to update the local perception information in real time.

[0041] In step S1, when the UAV identifies new boundary points, these points are added to the current topology as potential key topology nodes. In step S1, the connectivity of the topology graph is further enhanced by identifying inflection points in the path planning process. In step S1, the UAV samples the optimal boundary points within its sensor range to guide the subsequent exploration direction. After each global path planning is completed, only the data verified as key nodes are retained, ensuring the conciseness and effectiveness of the topology node set.

[0042] Step S2: Delaunay triangulation and collision detection;

[0043] The Delaunay triangulation algorithm is used to process the collected set of topological node information, and a continuous triangular network structure is created using the topological node information. When a UAV collides, the information of the corresponding edge in the topological graph is deleted.

[0044] In step S2, the collected set of topological nodes is processed using the Delaunay triangulation algorithm to create a continuous triangular network structure. Whenever a new node is added to the topological structure, the mesh layout is adjusted by point-by-point insertion to ensure the quality of the triangulation. In step S2, for obstacle constraints in the actual environment, each topological edge is expanded according to the physical size of the UAV, and a global octree grid map is used for accurate collision detection. If any collision is found, the corresponding edge will be considered invalid and removed from the topological graph, thereby ensuring that the retained edges all represent feasible collision-free paths.

[0045] Step S3: Numericalize the topological map;

[0046] The topology map is converted into a numerical form, and a feature matrix is ​​built using the topology vertex information, which makes it easy to quickly view the route through the feature matrix.

[0047] In step S3, the constructed topology map is converted into a numerical form to facilitate subsequent path search. Each topology vertex contains information such as its coordinate position, connected edges, and the Delaunay triangle to which it belongs. In step S3, a feature matrix is ​​built using the information collected from the topology vertices to represent the connection relationship between nodes. This matrix stores the information of all adjacent nodes in the form of a two-dimensional array. When performing path search, all possible forward routes can be quickly found by querying the corresponding part of a specific node in the feature matrix, which greatly improves the efficiency of graph traversal.

[0048] Step S4: Path planning and trajectory optimization;

[0049] The Dijkstra algorithm is used to find the optimal route forward, and inflection points with large information differences in the path trajectory are smoothed to improve the stability of the flight path and ensure the safe arrival of the UAV.

[0050] In step S4, Dijkstra's algorithm or other suitable shortest path algorithm is applied to the optimized topology map to find the best path from the current position to the target point. In step S4, considering that there may be too many turning points in actual flight that lead to non-optimal flight trajectories, the preliminary planning results are smoothed to reduce unnecessary turning times, improve the smoothness of the flight path and the stability of motion control, and ensure that the UAV can safely and efficiently reach the destination.

[0051] When using this UAV real-time path planning method for environmental exploration, an efficient topology map is dynamically constructed and maintained by combining the established global octree grid map with local perception information during the environmental exploration process. In the boundary detection module, asymptotically growing local candidate boundary points in the workspace are extracted in real time as key topology nodes. To enhance the connectivity of the topology, inflection points generated during path optimization are identified and extracted as supplementary topology nodes. Within the sensor's perception range, local sampling of the globally optimal boundary points is performed to guide the exploration direction. After each global path planning is completed, the above-mentioned local sampling points are automatically deleted, and only the verified key topology nodes are retained, thereby achieving dynamic simplification of the topology node set and avoiding the accumulation of redundant nodes.

[0052] The Delaunay triangulation algorithm was then used to mesh the topological node set, constructing a continuous triangular network structure. The topological mesh was dynamically updated using a point-by-point insertion method: when a new topological node was added, its triangle was located and the vertices were connected. Subsequently, the mesh structure was adjusted through a local optimization algorithm to ensure the quality of the triangulation. For obstacle constraints in the actual environment, each generated topological edge was subjected to line segment interpolation dilation based on the physical dimensions of the UAV, and accurate collision detection was performed using the established global octree grid map. If any collision point exists on the path of the dilated topological edge, the edge is deemed invalid and removed from the topological map, thus ensuring that the retained topological edges represent feasible collision-free paths. Delaunay triangulation is a core algorithm in computational geometry, used to generate non-overlapping triangular meshes that satisfy specific geometric properties from a discrete point set.

[0053] Secondly, the constructed topology map is numerically represented to support efficient path search. The topology map consists of a finite non-empty set of vertices and undirected edges between them. Each topology vertex contains its position information in the world coordinate system, associated topology edges, and the Delaunay triangle to which it belongs. An N-order feature matrix (N is the total number of nodes in the current topology map) is constructed to represent the geometric connection relationship between topology nodes. This matrix stores the node adjacency information in the form of a two-dimensional array. For any topology node, all adjacent nodes can be quickly obtained by querying its eigenvector in the feature matrix, which significantly improves the efficiency of graph traversal in path planning.

[0054] Finally, path planning is performed on the optimized topology map. The Dijkstra algorithm is used to search for the shortest path from the current position of the UAV to the target point (preferably the globally optimal boundary point). Given that the topology path may have many inflection points and is not geometrically optimal, the planning results are further optimized by using smoothing algorithms such as "converting points into lines" to reduce the path length and number of turns, improve the smoothness of the flight trajectory and the stability of motion control, and ensure that the UAV can safely and efficiently reach the target area.

[0055] Dijkstra's algorithm formula is: ifdist[u] + weight(u,v) <dist[v]:

[0056] dist[v] = dist[u] + weight(u,v)

[0057] prev[v] = u# Record the predecessor node

[0058] Where weight(u,v) is the edge weight; Termination condition: Ends when the priority queue is empty or the target node is marked as processed; Algorithm characteristics: Time complexity: Naive implementation: O(V 2 (V is the number of nodes), priority queue optimization: O[(V+E)logV] (E is the number of edges).

[0059] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0060] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time path planning method for unmanned aerial vehicles (UAVs) for environmental exploration, characterized in that: Includes the following steps: Step S1: Extract topology nodes; During the flight of the drone, the environmental information is first acquired using the lidar or visual sensors it is equipped with, and this data is combined with the existing global octree grid map to update the local perception information in real time. Step S2: Delaunay triangulation and collision detection; The Delaunay triangulation algorithm is used to process the collected set of topological node information, and a continuous triangular network structure is created using the topological node information. When a UAV collides, the information of the corresponding edge in the topological graph is deleted. Step S3: Numericalize the topological map; The topology map is converted into a numerical form, and a feature matrix is ​​built using the topology vertex information, which makes it easy to quickly view the route through the feature matrix. Step S4: Path planning and trajectory optimization; By using Dijkstra's algorithm to find the optimal route, and smoothing out inflection points with large information differences in the path trajectory, the stability of the flight path is improved and the drone arrives safely.

2. The real-time path planning method for UAVs oriented towards environmental exploration according to claim 1, characterized in that: In step S1, when the UAV identifies new boundary points, these points are added to the current topology as potential key topology nodes.

3. The real-time path planning method for UAVs oriented towards environmental exploration according to claim 1, characterized in that: In step S1, the connectivity of the topology graph is further enhanced by identifying inflection points in the path planning process.

4. The real-time path planning method for UAVs oriented towards environmental exploration according to claim 1, characterized in that: In step S1, the UAV samples the optimal boundary points within its sensor range to guide the subsequent exploration direction. After each global path planning is completed, only the data verified as key nodes are retained, ensuring the conciseness and effectiveness of the topology node set.

5. The real-time path planning method for UAVs for environmental exploration according to claim 1, characterized in that: In step S2, the collected set of topological nodes is processed using the Delaunay triangulation algorithm to create a continuous triangular network structure. Whenever a new node is added to the topology, the mesh layout is adjusted by point-by-point insertion to ensure the quality of the triangulation.

6. The real-time path planning method for UAVs for environmental exploration according to claim 1, characterized in that: In step S2, each topological edge is expanded according to the physical size of the UAV in response to obstacle constraints in the actual environment, and a global octree grid map is used for accurate collision detection. If any collision is found, the corresponding edge will be considered invalid and removed from the topological graph, thereby ensuring that the retained edges all represent feasible collision-free paths.

7. The real-time path planning method for UAVs for environmental exploration according to claim 1, characterized in that: In step S3, the constructed topology map is converted into a numerical form to facilitate subsequent path search. Each topology vertex contains information such as its coordinates, connected edges, and the Delaunay triangle to which it belongs.

8. The real-time path planning method for UAVs for environmental exploration according to claim 1, characterized in that: In step S3, a feature matrix is ​​established using the information collected from the topological vertices to represent the connection relationships between nodes. This matrix stores the information of all adjacent nodes in the form of a two-dimensional array. When performing path search, all possible forward routes can be quickly found by querying the corresponding part of a specific node in the feature matrix, which greatly improves the efficiency of graph traversal.

9. The real-time path planning method for UAVs for environmental exploration according to claim 1, characterized in that: In step S4, Dijkstra's algorithm is applied to the optimized topology map to find the best path from the current location to the target point.

10. A real-time path planning method for unmanned aerial vehicles (UAVs) for environmental exploration according to claim 1, characterized in that: In step S4, considering that there may be too many inflection points in actual flight leading to a non-optimal flight trajectory, the preliminary planning results are smoothed to reduce unnecessary turning times, improve the smoothness of the flight path and the stability of motion control, and ensure that the UAV can safely and efficiently reach its destination.

Images