A high safety margin global path planning method based on optimized Voronoi diagram

Abstract

The present invention discloses a high-safety-margin global path planning method based on an optimized Voronoi diagram, belonging to the field of robot path planning. The method comprises: first, applying a Voronoi diagram algorithm to generate an initial Voronoi road network; then, identifying nodes with at least three neighboring nodes, treating these nodes as transportation hubs, and constructing a first-layer sparse environment topology map; second, searching for neighboring hub nodes directly adjacent to and connected to the hub node, obtaining path information between adjacent transportation hub nodes, and performing pruning and cubic quasi-uniform B-spline curve smoothing to construct a second-layer path information map between adjacent and connected hub nodes; finally, performing a global path search between the starting point and the end point based on the Voronoi road network after hierarchical optimization modeling, initially obtaining corresponding path point sequence information, and performing Bezier curve smoothing to obtain a smooth, motion-constrained, high-safety-margin global path.

Description

Technical Field

[0001] The invention relates to a high safety margin global path planning method based on an optimized Voronoi diagram, and belongs to the field of robot path planning. Background Art

[0002] Robotic path planning technology is widely used in aerospace, military, industrial, and civilian applications. Autonomous robots, including satellites, rockets, self-driving vehicles, and sweeping robots, all incorporate path planning technology. The path planning module receives environmental maps, obstacle perception, and positioning information derived from front-end sensor data fusion and outputs a sequence of feasible path points to the control and guidance module, solving the "how to get there" problem. It serves as the decision-making brain within the entire autonomous navigation system architecture. Path planning essentially aims to find a motion path within a map that satisfies a series of constraints, minimizing the cost of moving the mobile vehicle from its starting point to its destination.

[0003] Global path planning technology requires a large amount of environmental information to be known in advance. It mainly performs path planning in a large-scale static map of the known environment to provide a rough route for the actual movement of the mobile robot. Among them, the commonly used global path planning algorithms are as follows: graph search-based path planning algorithms represented by grid graph method, visibility graph method, and topology graph method; sampling-based path planning algorithms represented by rapid traversal random tree (RRT) and probabilistic map search (PRM); intelligent bionics-based path planning algorithms represented by ant colony algorithm, particle swarm optimization algorithm, and genetic algorithm; reinforcement learning path planning algorithms represented by Q-learning.

[0004] In practical applications of autonomous robots, the path planning algorithm's priorities often need to be determined by considering the environmental context, task requirements, and the robot's own performance. For example, in environments with numerous hazardous radiation sources, dense obstacles, and insufficiently robust positioning accuracy, it is necessary to balance optimal navigation map modeling, algorithm runtime, path safety margins, and the robot's motion constraints. Therefore, it is essential to develop a global path planning algorithm with a simple environmental model, short runtime, high path safety margins, and smooth operation. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a global path planning algorithm with short running time, high path safety margin and smoothness based on further hierarchical optimization modeling of the environmental Voronoi road network. It can quickly plan a safe and smooth global path from the starting point to the end point under the conditions of multiple dangerous radiation sources in the environment, dense obstacles, low robustness and poor accuracy of the robot's own positioning.

[0006] To solve the above technical problems, the present invention provides a high safety margin global path planning method based on an optimized Voronoi diagram, the method comprising the following steps:

[0007] S1: Generate a global obstacle occupancy grid map through the current mainstream LiDAR-based SLAM mapping and occupancy grid map construction algorithm, further process it into a global binary grid map based on the probability threshold, and load the generated global binary grid map into the planning module;

[0008] S2: Applying the Voronoi graph algorithm to the environmental obstacle grid map input in step S1 to generate an initial Voronoi road network;

[0009] S3: All nodes of the initial Voronoi road network generated in step S2 are judged, and nodes with at least three neighboring nodes are identified. The nodes are used as transportation hub nodes to construct a first-layer sparse environment topology map;

[0010] S4: Based on the initial Voronoi road network node database generated in step S2, path search is performed between the transportation hub nodes generated in step S3, and neighboring hub nodes directly adjacent to and connected to the hub node are found, path information between adjacent transportation hub nodes is obtained, and a second-layer path information graph between adjacent and connected transportation hub nodes is constructed;

[0011] S5: Pruning and cubic quasi-uniform B-spline curve smoothing are performed on the path information between adjacent hub nodes obtained in step S4 to remove redundant nodes between adjacent transportation hub nodes and obtain a smooth path;

[0012] S6: Based on the path information between adjacent transportation hub nodes smoothed in step S5, further adaptive length path point interpolation is performed to obtain a complete hierarchical optimization modeled Voronoi road network;

[0013] S7: performing a global path search between the starting point and the end point based on the Voronoi road network after hierarchical optimization modeling generated in step S6, and preliminarily obtaining corresponding path point sequence information;

[0014] S8: Perform Bezier curve smoothing on the path point sequence information obtained by the preliminary search in step S7, and finally obtain a smooth global path with a high safety margin.

[0015] Preferably, the step S1 specifically includes the following process:

[0016] (1-1) A global obstacle occupancy grid map is generated through the current mainstream lidar-based SLAM mapping and positioning algorithm and the occupancy grid map construction algorithm, and is further processed into a global binary grid map based on the probability threshold.

[0017] (1-2) Load the generated global binary raster map into the planning module.

[0018] Preferably, step S2 specifically includes the following process:

[0019] (2-1) Based on the binary grid image of step S1, generate an obstacle connected domain;

[0020] (2-2) Based on the obstacle connectivity domain, the center of mass of the obstacle is extracted, and the center of mass point set C = {P i |i=1,2,3,…,n},P i Represents the centroid of the polygonal obstacle, then uses the centroid set C to construct a triangulation (Delaunay), and finally uses the relationship between the same triangle edges connected by the centroid to obtain the neighbor pair set Pair of the current polygonal obstacle feature i ={(A i ,A j )∣i≠j,i=1,2,…,n,j=1,2,…,n}, A i Indicates the obstacle element with element number i, A j Indicates A i Neighbor elements;

[0021] (2-3) After calculating the neighbor pairs of each obstacle element, it is necessary to classify the neighbor pairs according to the distance between the neighbor elements so as to interpolate the boundary points according to different strategies;

[0022] (2-4) Discretize the edges of each connected domain, construct a Voronoi graph based on the discrete edge point set, and obtain a road network node set consisting of Voronoi vertices;

[0023] (2-5) According to the vertex common edge relationship, each road network node is associated with the adjacent point to obtain the original map Voronoi road network structure node.

[0024] Preferably, the step S3 specifically includes the following process:

[0025] (3-1) Traverse all nodes of the Voronoi network structure node generated in step S2, and count the nodes of each node i The number of neighbor nodes no i _count;

[0026] (3-2) If no i _count≥3, the node is regarded as a transportation hub node and stored in the new transportation hub node structure new_node.

[0027] Preferably, step S4 specifically includes the following process:

[0028] (4-1) Setting the distance threshold D thresholdFor the transportation hub node structure new_node obtained in step S3, based on the initial Voronoi road network structure node, the Dijkstra path planning algorithm is applied to find the path sequence points between the transportation hub nodes within the distance threshold range;

[0029] (4-2) In order to identify the neighboring nodes of the transportation hub node, if the searched path sequence points contain other transportation hub nodes, it means that starting from the starting transportation hub node m Go to the target transportation hub node goal n The path between them passes through other transportation hub nodes, so there is no neighbor relationship between the starting transportation hub node and the target transportation hub node;

[0030] (4-3) On the contrary, if a neighbor relationship is formed, the path information between the transportation hub node and the corresponding node is stored as the road network information between the hub nodes of the second layer of new_node, and finally a hierarchical transportation hub node road network is generated.

[0031] Preferably, the step S5 specifically includes the following process:

[0032] (5-1) For the hierarchical transportation hub node network generated by S4, prune the paths between adjacent transportation hub nodes;

[0033] (5-2) The pruning process is divided into two steps. The first step is to remove redundant collinear points;

[0034] (5-3) Based on the path of eliminating collinear points, redundant turning points are further eliminated;

[0035] (5-4) For the key path points, perform B-spline curve smoothing. Suppose PO0, PO1, PO2,…, PO k , there are k+1 control points in total. These control points are used to define the direction and limit range of the spline curve. The definition of the k-order B-spline curve is:

[0036]

[0037] Where B i,k (u) is the i-th k-order B-spline basis function, and the control point PO i Correspondingly, k≥1;

[0038] The basis function has the following de Boer-Cox recursion:

[0039]

[0040] Because the quasi-uniform cubic B-spline curve has good locality and can achieve second-order derivative continuity, it is convenient for the robot to calculate the control instructions for smooth tracking. Therefore, the quasi-uniform cubic B-spline curve is finally selected to smooth the path curve.

[0041] Preferably, step S6 specifically includes the following process:

[0042] Based on the path information between adjacent transportation hub nodes smoothed in step S5, adaptive length path point interpolation is further performed to obtain a Voronoi road network after complete hierarchical optimization modeling.

[0043] Preferably, step S7 specifically includes the following process:

[0044] (7-1) Based on the Vinoy network modeled after hierarchical optimization, global path planning between any starting point START and end point GOAL is carried out. First, the starting point START and end point GOAL are integrated into the Vinoy network at the nearest location, and the intersection node FP is determined. START FP GOAL , the merged road segment identification number ID START 、ID GOAL , starting point START to converging node FP START The path is Loc_Path start , Fusion Node FP GOAL The path to the destination GOAL is Loc_Path goal ;

[0045] (7-2) Based on road segment ID START The endpoints determine the two traffic hub nodes ND1 and ND2 based on the road segment ID GOAL The endpoints determine two transportation hub nodes ND3 and ND4;

[0046] (7-3) Based on the first-layer transportation hub topology road network, Dijkstra algorithm is applied to obtain the path Path1 from ND1 to ND4, the path Path2 from ND2 to ND3, and the path from ND2 to ND4;

[0047] (7-4) The path length len_Loc_Path from the starting point to the intersection of the road network start , the path length from the end point to the road network intersection len_Loc_Path goal , the length between the intersection node and the transportation hub node, and the path length len_Path between transportation hub nodes i (i∈[1,2,3,4]) are superimposed, and according to the principle of the shortest planned path, a shortest path sequence Path_final is determined.

[0048] Preferably, the step S8 specifically includes the following process:

[0049] (8-1) Calculate the length of the global path sequence Len obtained in step S7 and determine the number of Bezier curve control points Count b =floor(2*Len), where floor is the floor function.

[0050] (8-2) Based on the number of control points Count b Determine the Bezier curve order (Count b -1) order, generate Bezier points according to the Bernstein basis function, form the Bezier curve global path sequence points, and obtain the final smooth, high safety margin global path res_path. Among them, the Bezier curve formula is:

[0051]

[0052] Among them, B i,n (t) is the Bernstein basis function, which has the advantages of non-negativity, unit decomposition, recursion, and derivative function. Its derivative is

[0053]

[0054] The derivative of the nth order Bernstein basis function can be obtained by linearly combining two n-1th order Bernstein basis functions, based on which the function recursion function can be realized.

[0055] Compared with the prior art, the present invention has the following advantages:

[0056] First, the optimized Voronoi road network modeling method proposed in this invention can solve the problems of rough road network models generated by the initial Voronoi diagram algorithm, redundant path nodes, and tortuous paths between nodes. It can make the environmental modeling of the entire map clearer and more concise, and fully utilize the high security advantage of the Voronoi road network.

[0057] Secondly, the proposed hierarchical modeling concept based on transport hub nodes and the method of pre-searching and storing path information between adjacent transport hub nodes can greatly shorten the search time in the path planning algorithm. The path search algorithm only needs to focus on how to integrate the starting point and the end point into the modeled road network node nearby, while the path sequence between the transport hub nodes that the intermediate path must pass through can be directly extracted.

[0058] Thirdly, the method of smoothing the global path points obtained based on hierarchical Voronoi road network planning proposed in the present invention can make the global path smooth and meet the motion constraints of the mobile robot while taking into account a high path safety margin. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Schematic diagram of the method flow of the present invention;

[0060] Figure 2 Schematic diagram of the basic Voronoi road network results;

[0061] Figure 3 Schematic diagram of the defects of basic Voronoi road network environment modeling, Figure 3 (a) There are many local nodes and they are not smooth, (b) The nodes are tortuous and redundant;

[0062] Figure 4 Road network topology after transportation hub nodes are extracted;

[0063] Figure 5 Schematic diagram of path search, pruning, and smoothing results between transportation hub nodes;

[0064] Figure 6 Global path planning results with high safety margin based on hierarchical Voronoi road network optimization modeling. DETAILED DESCRIPTION

[0065] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

[0066] Example 1: Figure 1 As shown, a high safety margin global path planning method based on optimized Voronoi diagram includes the following steps:

[0067] Step S1: Load a pre-generated global binary raster map.

[0068] (1-1) A global obstacle occupancy grid map is generated through the current mainstream lidar-based SLAM mapping and positioning algorithm and the occupancy grid map construction algorithm, and is further processed into a global binary grid map based on the probability threshold.

[0069] (1-2) Load the generated global binary raster map into the planning module.

[0070] Step S2: Apply the Voronoi graph algorithm to the environmental obstacle grid map input in step S1 to generate an initial Voronoi road network.

[0071] (2-1) Based on the binary grid image of step S1, generate an obstacle connected domain;

[0072] (2-2) Based on the obstacle connectivity domain, the center of mass of the obstacle is extracted, and the center of mass point set C = {P i |i=1,2,3,…,n},P iRepresents the centroid of the polygonal obstacle, then uses the centroid set C to construct a triangulation (Delaunay), and finally uses the relationship between the same triangle edges connected by the centroid to obtain the neighbor pair set Pair of the current polygonal obstacle feature i ={(A i ,A j )∣i≠j,i=1,2,…,n,j=1,2,…,n}, A i Indicates the obstacle element with element number i, A j Indicates A i Neighbor elements;

[0073] (2-3) After calculating the neighbor pairs of each obstacle element, it is necessary to classify the neighbor pairs according to the distance between the neighbor elements so as to interpolate the boundary points according to different strategies;

[0074] (2-4) Discretize the edges of each connected domain, construct a Voronoi graph based on the discrete edge point set, and obtain a road network node set consisting of Voronoi vertices;

[0075] (2-5) According to the vertex common edge relationship, each road network node is associated with the adjacent point to obtain the original map Voronoi road network structure node.

[0076] In this embodiment, the above steps are completed to generate the following effect: Figure 2 The basic Voronoi network diagram is shown.

[0077] Step S3: All nodes of the initial Voronoi road network generated in step S2 are judged, and nodes with at least three neighboring nodes are identified. The nodes are used as transportation hub nodes to construct the first-layer sparse environment topology map.

[0078] (3-1) Traverse all nodes of the Voronoi network structure node generated in step S2, and count the nodes of each node i The number of neighbor nodes no i _count;

[0079] (3-2) If no i _count≥3, the node is regarded as a transportation hub node and stored in the new transportation hub node structure new_node.

[0080] In this embodiment, the above steps are completed to generate the following effect: Figure 4 The road network topology diagram after the transportation hub nodes are extracted is shown.

[0081] Step S4: Based on the initial Vinoy road network node database generated in step S2, perform path search between the transportation hub nodes generated in step S3, find neighboring hub nodes that are directly adjacent to and connected to the hub nodes, obtain path information between adjacent transportation hub nodes, and construct a second-layer path information graph between adjacent and connected transportation hub nodes.

[0082] (4-1) Setting the distance threshold D threshold For the transportation hub node structure new_node obtained in step S3, based on the initial Voronoi road network structure node, the Dijkstra path planning algorithm is applied to find the path sequence points between the transportation hub nodes within the distance threshold range;

[0083] (4-2) In order to identify the neighboring nodes of the transportation hub node, if the searched path sequence points contain other transportation hub nodes, it means that starting from the starting transportation hub node m Go to the target transportation hub node goal n The path between them passes through other transportation hub nodes, so there is no neighbor relationship between the starting transportation hub node and the target transportation hub node;

[0084] (4-3) On the contrary, if a neighbor relationship is formed, the path information between the transportation hub node and the corresponding node is stored as the road network information between the hub nodes of the second layer of new_node, and finally a hierarchical transportation hub node road network is generated.

[0085] Step S5: Pruning and cubic quasi-uniform B-spline curve smoothing are performed on the path information between adjacent hub nodes obtained in step S4 to remove redundant nodes between adjacent transportation hub nodes and obtain a smooth path.

[0086] (5-1) For the hierarchical transportation hub node network generated by S4, prune the paths between adjacent transportation hub nodes;

[0087] (5-2) Pruning is a two-step process. The first step is to remove redundant collinear points. Starting from the starting node, traverse all path points in the path. If the current node, the parent node of the current node, and the parent node of its parent node are all on the same straight line, then these three path points are collinear. The parent node of the current node is a redundant collinear path point and is deleted.

[0088] (5-3) Based on the path after eliminating collinear points, redundant turning points are further eliminated. Suppose that the path after eliminating redundant collinear points has k path points {P m |m=1,2,3,…,k}, traverse all path points and connect the current path point P in sequence i (1≤i≤k-2) and other path points P j (i+2≤j≤k). If Pi P j If the line segment does not pass through any obstacles on the map, then the path point P is considered i and P j Can be adjacent, remove path point P i and P j The remaining corner path points between the two paths are the key path points after pruning;

[0089] (5-4) For the key path points, perform B-spline curve smoothing. Suppose PO0, PO1, PO2,…, PO k , there are k+1 control points in total. These control points are used to define the direction and limit range of the spline curve. The definition of the k-order B-spline curve is:

[0090]

[0091] Where B i,k (u) is the i-th k-order B-spline basis function, and the control point PO i Correspondingly, k≥1;

[0092] The basis function has the following de Boer-Cox recursion:

[0093]

[0094] Because the quasi-uniform cubic B-spline curve has good locality and can achieve second-order derivative continuity, it is convenient for the robot to calculate the control instructions for smooth tracking. Therefore, the quasi-uniform cubic B-spline curve is finally selected to smooth the path.

[0095] In this embodiment, the above steps are completed to generate the following effect: Figure 5 The diagram shows the results of path search, pruning, and smoothing between transportation hub nodes.

[0096] Step S6: Based on the path information between adjacent transportation hub nodes smoothed in step S5, further adaptive length path point interpolation is performed to obtain the Voronoi road network after complete hierarchical optimization modeling.

[0097] Step S7: Based on the Voronoi road network after hierarchical optimization modeling generated in step S6, a global path search is performed between the starting point and the end point to preliminarily obtain the corresponding path point sequence information.

[0098] (7-1) Based on the Vinoy network modeled after hierarchical optimization, global path planning between any starting point START and end point GOAL is carried out. First, the starting point START and end point GOAL are integrated into the Vinoy network at the nearest location, and the intersection node FP is determined. START FP GOAL , the merged road segment identification number ID START 、IDGOAL , starting point START to converging node FP START The path is Loc_Path start , Fusion Node FP GOAL The path to the destination GOAL is Loc_Path goal ;

[0099] (7-2) Based on road segment ID START The endpoints determine the two traffic hub nodes ND1 and ND2 based on the road segment ID GOAL The endpoints determine two transportation hub nodes ND3 and ND4;

[0100] (7-3) Based on the first-layer transportation hub topology road network, Dijkstra algorithm is applied to obtain the path Path1 from ND1 to ND4, the path Path2 from ND2 to ND3, and the path from ND2 to ND4;

[0101] (7-4) The path length len_Loc_Path from the starting point to the intersection of the road network start , the path length from the end point to the road network intersection len_Loc_Path goal , the length between the intersection node and the transportation hub node, and the path length len_Path between transportation hub nodes i (i∈[1,2,3,4]) are superimposed and a shortest path sequence Path_final is determined based on the principle of the shortest planned path;

[0102] Step S8: Perform Bezier curve smoothing on the path point sequence information obtained by the preliminary search in step S7 (ensuring at least 4 path control points), and finally obtain a smooth global path with a high safety margin.

[0103] (8-1) Calculate the length of the global path sequence Len obtained in step S7 and determine the number of Bezier curve control points Count b =floor(2*Len), where floor is the floor function.

[0104] (8-2) Based on the number of control points Count b Determine the Bezier curve order (Count b -1) order, generate Bezier points according to the Bernstein basis function, form the Bezier curve global path sequence points, and obtain the final smooth, high safety margin global path res_path. The Bezier curve formula is:

[0105]

[0106] Among them, B i,n(t) is the Bernstein basis function, which has the advantages of non-negativity, unit decomposition, recursion, and derivative function. Its derivative is:

[0107]

[0108] In this embodiment, the above steps are completed to generate the following effect: Figure 6 The figure shows a schematic diagram of the high safety margin global path planning results based on hierarchical Voronoi road network optimization modeling.

[0109] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention, and equivalent changes or substitutions made on the basis of the above technical solutions fall within the scope of protection of the claims of the present invention.

Claims

1. A high safety margin global path planning method based on optimized Voronoi diagram, characterized in that: The method comprises the following steps: S1: Load the pre-generated global binary raster map; S2: Applying the Voronoi graph algorithm to the environmental obstacle grid map input in step S1 to generate an initial Voronoi road network; S3: All nodes of the initial Voronoi road network generated in step S2 are judged, and nodes with at least three neighboring nodes are identified. The nodes are used as transportation hub nodes to construct a first-layer sparse environment topology map; S4: Based on the initial Voronoi road network node database generated in step S2, path search is performed between the transportation hub nodes generated in step S3, and neighboring hub nodes directly adjacent to and connected to the hub node are found, path information between adjacent transportation hub nodes is obtained, and a second-layer path information graph between adjacent and connected transportation hub nodes is constructed; S5: Pruning and cubic quasi-uniform B-spline curve smoothing are performed on the path information between adjacent hub nodes obtained in step S4 to remove redundant nodes between adjacent transportation hub nodes and obtain a smooth path; S6: Based on the path information between adjacent transportation hub nodes smoothed in step S5, further adaptive length path point interpolation is performed to obtain a complete hierarchical optimization modeled Voronoi road network; S7: performing a global path search between the starting point and the end point based on the Voronoi road network after hierarchical optimization modeling generated in step S6, and preliminarily obtaining corresponding path point sequence information; S8: Perform Bezier curve smoothing on the path point sequence information obtained by the preliminary search in step S7, and finally obtain a smooth global path with a high safety margin.

2. The high safety margin global path planning method based on optimized Voronoi diagram according to claim 1, characterized in that: Specifically in step S1 The following processes are included: (1-1) Generate a global obstacle occupancy grid map using the current mainstream LiDAR-based SLAM mapping and occupancy grid map construction algorithm, and further process it into a global binary grid map based on the probability threshold; (1-2) Load the generated global binary raster map into the planning module.

3. The high safety margin global path planning method based on optimized Voronoi diagram according to claim 2, characterized in that: Specifically in step S2 The following processes are included: (2-1) Based on the binary grid image of step S1, generate an obstacle connected domain; (2-2) Based on the obstacle connectivity domain, the center of mass of the obstacle is extracted, and the center of mass point set C = {P i |i=1,2,3,…,n},P i Represents the centroid of the polygonal obstacle, then uses the centroid set C to construct the triangulation network, and finally uses the relationship between the same triangle edges connected by the centroid points to obtain the neighbor pair set Pair of the current polygonal obstacle feature i ={(A i ,A j )∣i≠j,i=1,2,…,n,j=1,2,…,n}, A i Indicates the obstacle element with element number i, A j Indicates A i Neighbor elements; (2-3) After calculating the neighbor pairs of each obstacle element, it is necessary to classify the neighbor pairs according to the distance between the neighbor elements so as to interpolate the boundary points according to different strategies; (2-4) Discretize the edges of each connected domain, construct a Voronoi graph based on the discrete edge point set, and obtain a road network node set consisting of Voronoi vertices; (2-5) According to the vertex common edge relationship, each road network node is associated with the adjacent point to obtain the original map Voronoi road network structure node.

4. The high safety margin global path planning method based on optimized Voronoi diagram according to claim 3, characterized in that: Specifically in step S3 The following processes are included: (3-1) Traverse all nodes of the Voronoi network structure generated in step S2 and count the number of neighboring nodes of each node; (3-2) If the number of neighbor nodes of the node is greater than or equal to 3, the node is regarded as a transportation hub node and stored in a new transportation hub node structure.

5. The high safety margin global path planning method based on optimized Voronoi diagram according to claim 4, characterized in that: Specifically in step S4 The following processes are included: (4-1) Setting a distance threshold, for the transportation hub node structure obtained in step S3, based on the initial Voronoi road network structure, applying the Dijkstra path planning algorithm to find path sequence points between transportation hub nodes within the distance threshold range; (4-2) To identify the neighboring nodes of a transportation hub node, if the searched path sequence points include other transportation hub nodes, it means that the path from the starting transportation hub node to the target transportation hub node passes through other transportation hub nodes, so there is no neighbor relationship between the starting transportation hub node and the target transportation hub node; (4-3) On the contrary, if a neighbor relationship is formed, the path information between the transportation hub node and the corresponding node is stored as the road network information between the second-level hub nodes, and finally a hierarchical transportation hub node road network is generated.

6. The high safety margin global path planning method based on optimized Voronoi diagram according to claim 5, characterized in that: Specifically in step S5 The following processes are included: (5-1) For the hierarchical transportation hub node network generated by S4, prune the paths between adjacent transportation hub nodes; (5-2) The pruning process is divided into two steps. The first step is to remove redundant collinear points; (5-3) Based on the path with collinear points removed, redundant turning points are further removed, and the remaining path points are the key path points after pruning; (5-4) For the key path points remaining after pruning and smoothing, quasi-uniform cubic B-spline curve smoothing is selected.

7. The high safety margin global path planning method based on optimized Voronoi diagram according to claim 1, characterized in that: Specifically in step S6 The following processes are included: Based on the path information between adjacent transportation hub nodes smoothed in step S5, adaptive length path point interpolation is further performed to obtain a Voronoi road network after complete hierarchical optimization modeling.

8. The high safety margin global path planning method based on optimized Voronoi diagram according to claim 1, characterized in that: Specifically in step S7 The following processes are included: (7-1) To perform global path planning between any starting point and end point based on the Vinoy network modeled after hierarchical optimization, the starting point and end point must first be merged into the Vinoy network as close as possible, the intersection node is determined, and the identification number of the merged road segment is recorded; (7-2) Determine two transportation hub nodes adjacent to the starting point based on the endpoints of the merging starting point segment, and determine two transportation hub nodes adjacent to the ending point based on the endpoints of the merging ending point segment; (7-3) Based on the first-level transportation hub topology road network, the Dijkstra algorithm is applied to obtain the paths between the transportation hub nodes close to the starting point and the transportation hub nodes close to the end point respectively; (7-4) The path lengths from the starting point to the road network intersection, the path lengths from the end point to the road network intersection, the lengths from the intersection node to the transportation hub node, and the path lengths between transportation hub nodes are superimposed and compared to determine the shortest path sequence Path_final.

9. The high safety margin global path planning method based on optimized Voronoi diagram according to claim 1, characterized in that: Specifically in step S8 The following processes are included: (8-1) Calculate the length of the global path sequence Len obtained in step S7 and determine the number of Bezier curve control points Count b =floor(2*Len), where floor is the floor function; (8-2) Based on the number of control points Count b Determine the Bezier curve order (Count b -1) order, Bezier points are generated according to the Bernstein basis function to form the sequence points of the Bezier curve global path, and the final smooth global path with high safety margin is obtained.

Images