A global path planning method for autonomous vehicles

Abstract

This invention discloses a global path planning method for autonomous vehicles, belonging to the field of autonomous vehicle path planning technology. It addresses the problems of low search efficiency, slow convergence speed, high memory consumption, and sharp corners in the planning results of existing path planning methods. First, the invention simplifies the global grid map, retaining only larger obstacles. Then, it uses the simplified map to guide the initial path. After determining the initial guiding path, it expands it into an initial sampling region. Introducing the sampling region improves the algorithm's search efficiency and convergence speed. Simultaneously, iterative search is performed using the unsimplified global grid map, ensuring the algorithm can plan a more reasonable path while reducing the number of iterations, thus reducing memory consumption. Finally, the generated path is smoothed to produce a path with continuity and smoothness. This invention can be applied to autonomous vehicle path planning.

Description

Technical Field

[0001] This invention belongs to the field of unmanned vehicle path planning technology, specifically relating to a global path planning method for unmanned vehicles. Background Technology

[0002] Autonomous vehicles are currently a hot topic in automotive technology development, and path planning is a crucial element in achieving autonomous driving. Path planning in autonomous driving is divided into two categories: global path planning based on known global prior information and local path planning based on changes in the vehicle's surrounding environment (GONZALEZ D, PEREZ J, MILANES V et al. A review of motion planning techniques for automated vehicles[J]. IEEE Transactions on Intelligent Transportation Systems, 2015, 17(4): 1-11.).

[0003] Global path planning requires mastering all environmental information and performing path planning based on all information in the environmental map, such as Dijkstra's algorithm, A* algorithm, genetic algorithm (GA), and fast random search tree algorithm (RRT), etc. I,Jajszczyk A, B. Generic Dijkstra for optical networks[J]. Journal of Optical Communications and Networking, 2019, 11(11): 568-577.). The main feature of Dijkstra's algorithm is that it expands outward from the center until it reaches the destination, but it traverses and calculates many nodes, resulting in low efficiency. The A* algorithm introduces a heuristic function to avoid a large number of invalid search paths that are often generated in path planning problems. The path obtained by the genetic algorithm is more in line with the path selection method of intelligent agents in reality, but it has a large amount of computation and takes a long time, which cannot meet the real-time requirements. LaValle et al. (H.Umari and S.Mukhopadhyay. Autonomous robotic exploration based on multiple rapidly-exploring randomized trees[C], 2017 IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS), Vancouver, BC, Canada, 2017:1396-1402.) proposed a node-sampling-based Rapidly-exploring Random Tree (RRT) algorithm. Compared with other algorithms, the advantage of the RRT algorithm is that it does not require precise global map environment information and can find a collision-free global path by direct search. Researchers have proposed improved algorithms to address the shortcomings of RRT. Karaman et al. (Lau B, Sprunk C, Burgard W. Kinodynamic motionplanning for mobile robots using splines[C]. IEEE / RSJ International Conference on Intelligent Robots and Systems. Louis, 2009:2427-2433.) proposed the asymptotically optimal (shortest distance) RRT* algorithm (Rapidly-exploring Random Tree* algorithm), which adds random geometry and pruning optimization theory to the RRT algorithm. The probabilistically complete RRT* algorithm guarantees that the optimal path can be found with a sufficient number of iterations.However, because the RRT* algorithm performs an indiscriminate search of the entire space during path planning, it suffers from low search efficiency, slow convergence speed, and high memory consumption. Furthermore, the planned paths have shortcomings such as sharp corners.

[0004] In conclusion, it is essential to propose a new global path planning method to address the problems of low search efficiency, slow convergence speed, high memory consumption, and sharp corners in the planning results of existing methods. Summary of the Invention

[0005] The purpose of this invention is to address the problems of low search efficiency, slow convergence speed, high memory consumption, and sharp corners in the planning results of existing path planning methods, and to propose a global path planning method for unmanned vehicles.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0007] A global path planning method for autonomous vehicles, the method specifically includes the following steps:

[0008] Step 1: After dividing the global map into grids, binarize the information in the global grid map to obtain the binarized global grid map.

[0009] Obstacles with areas smaller than a threshold are removed from the binarized global raster map to obtain the processed global raster map.

[0010] Step 2: In the global grid map processed in Step 1, a graph search algorithm is used to search for a path for the autonomous vehicle from the starting point to the destination.

[0011] Step 3: Put the obstacles deleted in Step 1 back into the global map and determine whether the path found in Step 2 passes through the obstacles.

[0012] Step 4: If the searched path passes through obstacles, establish a sampling area based on the searched path, delete the obstacles deleted in Step 1, and then use a graph search algorithm to search for a new path for the autonomous vehicle from the starting point to the end point within the sampling area before returning to Step 3.

[0013] Otherwise, if the searched path does not pass through obstacles, the searched path will be used directly as the initial guiding path, and step five will be executed.

[0014] Step 5: In the global raster map processed in Step 1, determine the sampling area based on the initial guiding path;

[0015] Step 6: Within the determined sampling area, use the RRT* algorithm to search for a path from the starting point to the ending point, and put the obstacles deleted in Step 1 back into the global map;

[0016] Determine whether the path found by the RRT* algorithm passes through obstacles. If the path found by the RRT* algorithm passes through obstacles, proceed to step seven; otherwise, proceed directly to step eight.

[0017] Step 7: Adjust the expansion factor Then, return to step five;

[0018] Step 8: Determine whether the searched path meets the iteration stopping condition;

[0019] If the iteration stopping condition is met, the searched path is the final path planning result;

[0020] Otherwise, if the iteration stopping condition is not met, the sampling region is established based on the searched path, and then the process returns to step six.

[0021] The beneficial effects of this invention are:

[0022] This invention first simplifies the global grid map, retaining only the larger obstacles. Then, it uses the simplified map to guide the initial path. After determining the initial guiding path, it expands it into an initial sampling region. Introducing the sampling region improves the algorithm's search efficiency and convergence speed. Simultaneously, it uses the unsimplified global grid map for iterative search. This ensures that the algorithm can plan a more reasonable path while reducing the number of iterations, thereby reducing memory usage. Finally, a dynamic smoothing optimization method is introduced to post-process the generated path, thereby generating a path with continuity and smoothness, solving the problem of sharp corners in the planning results, and improving the path tracking effect. Attached Figure Description

[0023] Figure 1 This is a flowchart of a global path planning method for an unmanned vehicle according to the present invention;

[0024] Figure 2(a) shows the global map after binarization;

[0025] Figure 2(b) is a schematic diagram of the identified obstacle area;

[0026] Figure 2(c) shows the global map after removing small obstacles;

[0027] Figure 3 This is a schematic diagram of the sampling area on the simplified map;

[0028] Figure 4 This is a schematic diagram of the sampling area in the original map;

[0029] Figure 5 This is a schematic diagram of the smoothing process;

[0030] Figure 6(a) is a schematic diagram of the planning results of the RRT algorithm in an accessible environment;

[0031] Figure 6(b) is a schematic diagram of the planning results of the RRT* algorithm in an accessible environment;

[0032] Figure 6(c) is a schematic diagram of the planning results of the method of the present invention in an accessible environment;

[0033] Figure 7(a) is a schematic diagram of the planning results of the RRT algorithm in a simple obstacle environment;

[0034] Figure 7(b) is a schematic diagram of the planning results of the RRT* algorithm in a simple obstacle environment;

[0035] Figure 7(c) is a schematic diagram of the planning results of the method of the present invention in a simple obstacle environment;

[0036] Figure 8(a) is a schematic diagram of the global map;

[0037] Figure 8(b) is a schematic diagram of the initial guidance path of the RRT* algorithm after the map is simplified;

[0038] Figure 8(c) is a schematic diagram of the initial guidance path of the Smooth-RRT* algorithm with restricted sampling after map simplification;

[0039] Figure 8(d) shows the path planning results of the RRT* algorithm;

[0040] Figure 8(e) shows the path planning results of the restricted sampling Smooth-RRT* algorithm;

[0041] Figure 8(f) shows the curvature of the search path for the restricted sampling Smooth-RRT* algorithm. Detailed Implementation

[0042] Specific Implementation Method 1: Combination Figure 1 This embodiment describes a global path planning method for an autonomous vehicle, which specifically includes the following steps:

[0043] Step 1: After dividing the global map into grids, binarize the information in the global grid map to obtain the binarized global grid map.

[0044] Remove obstacles with areas smaller than a threshold (the threshold size can be set according to the actual situation) from the binarized global raster map to obtain the processed global raster map.

[0045] Step 2: In the global grid map processed in Step 1, a graph search algorithm is used to search for a path for the autonomous vehicle from the starting point to the destination.

[0046] Step 3: Put the obstacles deleted in Step 1 back into the global map and determine whether the path found in Step 2 passes through the obstacles.

[0047] Step 4: If the searched path passes through obstacles, establish a sampling area based on the searched path, delete the obstacles deleted in Step 1, and then use a graph search algorithm to search for a new path for the autonomous vehicle from the starting point to the end point within the sampling area before returning to Step 3.

[0048] Otherwise, if the searched path does not pass through obstacles, the searched path will be used directly as the initial guiding path, and step five will be executed.

[0049] Step 5: In the global raster map processed in Step 1, determine the sampling area based on the initial guiding path;

[0050] Step 6: Within the determined sampling area, use the RRT* algorithm to search for a path from the starting point to the ending point, and put the obstacles deleted in Step 1 back into the global map;

[0051] Determine whether the path found by the RRT* algorithm passes through obstacles. If the path found by the RRT* algorithm passes through obstacles, proceed to step seven; otherwise, proceed directly to step eight.

[0052] Step 7: Adjust the expansion factor Then, return to step five; that is, redetermine the sampling area by adjusting the expansion factor (after determining a path l1 that does not pass through obstacles based on the initial guiding path, determine the sampling area based on path l1, and then search for a path again based on the sampling area until the next path l2 that does not pass through obstacles is found, and so on. It should be noted that if a path that passes through obstacles is found during the process of searching for path l2 based on path l1, the sampling area needs to be readjusted based on path l1 until a path that does not pass through obstacles is found).

[0053] Step 8: Determine whether the searched path meets the iteration stopping condition;

[0054] If the iteration stopping condition is met, the searched path is the final path planning result;

[0055] Otherwise, if the iteration stopping condition is not met, the sampling region is established based on the searched path, and then the process returns to step six.

[0056] This invention introduces intelligent sampling regions and combines them with an obstacle reproduction and overlay strategy to achieve iterative planning of the global path, which can effectively reduce the performance loss caused by the undifferentiated search space treatment of the RRT* algorithm.

[0057] Specific Implementation Method Two: This implementation method differs from Specific Implementation Method One in that, in step one, the information in the global raster map is binarized, specifically as follows:

[0058] 0 represents the non-obstacle area (i.e., the feasible area) in the global grid map, and 1 represents the area where the obstacle is located in the global grid map.

[0059] This implementation classifies the occupancy information in the raster map and segments the obstacle information in connected graticles to ensure that each obstacle region can be identified individually. Obstacles are sorted according to their occupancy data size for easier subsequent processing. As shown in Figure 2(a), the global raster map information is first binarized, where 0 represents a feasible area and 1 represents an obstacle in the graticle. Then, obstacles in the map are segmented according to the numerical information in the graticles, and their corresponding sizes are calculated, as shown in Figure 2(b). Finally, obstacles below a certain threshold are deleted, completing the simplification process of the global map, as shown in Figure 2(c). Global raster map simplification is of great significance for path planning algorithms.

[0060] The other steps and parameters are the same as in Specific Implementation Method 1.

[0061] Specific Implementation Method 3: This implementation method differs from Specific Implementation Method 1 or 2 in that the graph search algorithm used in step 2 is the A* algorithm.

[0062] Other steps and parameters are the same as in specific implementation method one or two.

[0063] Specific Implementation Method Four: This implementation method differs from Specific Implementation Methods One to Three in that the specific process of determining the sampling area based on the initial guidance path is as follows:

[0064] Step 1: Calculate the path dilation distance:

[0065]

[0066] Where: S is the path expansion distance, and E is the side length of the global map. For expansion factors;

[0067] Step 2: Expand the sampling area with the initial guide path as the center (the shape of the two boundary lines of the sampling area is the same as that of the initial guide path), and the distance between the initial guide path and the two boundaries of the sampling area is the path expansion distance S.

[0068] like Figure 3 As shown, the line from the starting point to the ending point represents the initial guiding path found during the search, and the surrounding gray area is its expanded sampling region. The distance between this path and the gray boundary is S. To ensure that the width of this sampling region is large enough, and to ensure search efficiency, generally... Take 4. For example... Figure 4 As shown, if the sampling area passes through small obstacles, the sampling area needs to be adjusted by adjusting the expansion factor.

[0069] The other steps and parameters are the same as those in one of the specific implementation methods one to three.

[0070] Specific Implementation Method Five: This implementation method differs from Specific Implementation Methods One to Four in that the iteration stopping condition is: during N consecutive iterations (each execution of step eight is considered the end of an iteration), the searched paths are the same.

[0071] By iteratively searching along this path, new sampling areas are defined. The algorithm continues searching until the found path no longer changes after a certain number of iterations, at which point the iteration ends. This process effectively improves search efficiency and reduces memory usage. The number of searches is related to the size of the environment map; the larger the environment, the more searches are required. In this invention, N is set to 10.

[0072] The other steps and parameters are the same as those in one of the specific implementation methods one to four.

[0073] Specific Implementation Method Six: This implementation method differs from Specific Implementation Methods One to Five in that the new node P generated by the RRT* algorithm... new The following constraints must be satisfied:

[0074] C new ≤C path -C goal

[0075] Where: C new For the new node P new The current cost, C path C represents the total cost of the current optimal path. goal This is the Euclidean distance from the new node to the target point.

[0076] By adding constraints to the traditional RRT* algorithm, this implementation method can ensure that the searched paths have better connectivity and improve search efficiency.

[0077] The other steps and parameters are the same as those in one of the specific implementation methods one to five.

[0078] Specific implementation method seven: Combination Figure 5 This embodiment is described below. This embodiment differs from any of the specific embodiments one through six in that the method further includes step nine, which specifically comprises:

[0079] The final path planning result in step eight is then smoothed. The specific process of smoothing is as follows:

[0080] Let the new state node be represented as point P. new The current vertex is represented as point P. i Point P i The extension of the path direction intersects point P new Let point P be the intersection of the extended lines of the path directions. i The straight-line distance between point P and point S is denoted as S. i Point P new The straight-line distance between point P and point S is denoted as S. n At point P and point P i Take a virtual node P on the connection line s Satisfying the virtual node P s The straight-line distance between point P and point S is n And virtual node P s With point P i The directions are the same;

[0081] At point P new With virtual node P s Between these points, the unmanned vehicle moves in an arc with a constant speed v1 and a yaw rate ω1;

[0082]

[0083] Where Δθ is the arc The angle (e.g.) Figure 5 As shown, point O represents the center of the arc, and ρ represents the curvature.

[0084] At virtual node P s With point P i Between these points, the unmanned vehicle moves in a straight line at a constant speed v2 and a yaw rate ω2, i.e., ω2 equals 0.

[0085] This implementation takes into account motion dynamics constraints for smoothing, which solves the problem of insufficient stability and easy mechanical wear caused by the tortuous, uneven, and excessive number of nodes in the path of the unmanned vehicle.

[0086] The other steps and parameters are the same as those in one of the specific implementation methods one to six.

[0087] Specific Implementation Method Eight: This implementation method differs from Specific Implementation Methods One to Seven in that the velocities v1 and v2 are:

[0088] Define velocity profile v n It's an autonomous vehicle at point P. new The driving speed, v i It's an autonomous vehicle at point P. i The driving speed, v n and v iDetermined by the state transition equation in random tree expansion;

[0089] The velocities v1 and v2 are randomly selected from the velocity profile V, meaning there are nine possible combinations of v1 and v2.

[0090] The other steps and parameters are the same as those in any of the specific implementation methods one to seven.

[0091] Specific Implementation Method Nine: This implementation method differs from Specific Implementation Methods One to Eight in that the speed and yaw rate satisfy the following constraints:

[0092]

[0093] Where, α min α is the minimum acceleration of the autonomous vehicle along the direction of travel. max β is the maximum acceleration of the autonomous vehicle along the direction of travel. min It is the minimum yaw acceleration of the unmanned vehicle, β. max Δt is the maximum value of the yaw angle acceleration of the autonomous vehicle, and Δt is the planned time step.

[0094] The other steps and parameters are the same as those in one of the specific implementation methods one to eight.

[0095] Experimental simulation

[0096] A comparative experiment was conducted in a simulation environment to compare the planning results of the restricted sampling Smooth-RRT* algorithm (the method of this invention) with those of traditional algorithms. This experimental scenario consisted of two parts: one with no obstacles between the starting point and the ending point, and the other with some obstacles. The map was primarily a 500×500 grid map with a grid resolution of 1. The global target area was a rectangular region with the bottom left corner at coordinates (50, 48) as the starting point and the top right corner at coordinates (430, 450) as the center point. The curve represented the final planned path.

[0097] First, the parameters of the Smooth-RRT* algorithm for restricted sampling in an accessible environment are set. Since there are no obstacles on the map, the threshold for simplifying the raster map can be set to any value; here, it is set to 0. The initial sampling number n = 500. The basic parameters of the RRT* algorithm are set as follows: iteration step size is 5m, and trimming radius is 10m. The path obtained by limiting the maximum number of iterations to 3000 is selected. Due to the randomness of the algorithm, 30 simulation experiments are conducted.

[0098] As shown in Figures 6(a), 6(b), 6(c) and Table 1, a comparison reveals that, although the RRT* algorithm has a relatively smaller number of inflection points, it still contains a large number of meaningless sampling points in the paths planned by the traditional RRT algorithm and the RRT* algorithm. The restricted sampling Smooth-RRT* algorithm of this invention obtains the initial guiding path by optimizing the raster map, and then performs sampling search within this path area, resulting in a straight line segment that is essentially close to the optimal solution.

[0099] Table 1 Comparison of path planning results in an obstacle-free environment (optimal path length 553.18)

[0100]

[0101] As shown in Figures 7(a), 7(b), and 7(c) and Table 2, a comparison reveals that the improved algorithm of this invention exhibits a more significant advantage in simple obstacle environments. Traditional RRT and RRT* algorithms have numerous meaningless inflection points. The path obtained by this invention based on the restricted sampling Smooth-RRT* algorithm has virtually no meaningless inflection points, and the entire route presents a quasi-polyline shape. Compared to the traditional RRT algorithm, the average path length of this algorithm is shortened by 7.83%, enabling more efficient generation of global paths and demonstrating relative superiority, thus meeting the global path optimization requirements of autonomous vehicles. Therefore, this algorithm has high practical value and application prospects, providing strong support and assurance for global path planning of autonomous vehicles.

[0102] Table 2. Path planning results in a simple obstacle environment (optimal path length 567.76).

[0103]

[0104] Effectiveness verification of global path planning algorithm in multiple scenarios

[0105] The proposed method and the RRT* algorithm were compared and simulated on two different maps. Both maps were 500×500 raster maps with a raster resolution of 1. The global target area was a rectangular region with the lower left corner coordinates (50, 48) as the starting point and the upper right center point coordinates (430, 450). The curve represented the final planned path. First, the parameters of the restricted sampling Smooth-RRT* algorithm were set. The threshold for simplifying the raster map was set to 100, the initial sampling number was set to n = 500, and the basic parameters of the RRT* algorithm were: iteration step size of 5m and trimming radius of 10m.

[0106] In the map shown in Figure 8(a), after removing smaller obstacles based on a threshold, 30 simulations were run. The simulation results are shown in Figures 8(a), 8(b), 8(c), and Table 3 (Figures 8(b) and 8(c) are path maps with a higher number of iterations in the 30 simulations, and Table 3 shows the average results of the 30 simulations). When using the RRT* algorithm to calculate the initial path, the average number of samples was 834, the average time was 0.227s, and the average path length was 695.39m. When the algorithm of this invention searches for the shortest path, the average number of samples was 118, the average time was 0.108s, and the average path length was 582.74m. The number of iterations decreased by 85.9%, the convergence time of the algorithm decreased by 52.4%, and the path length decreased by 16.2%.

[0107] Table 3 Simulation data for each algorithm after map simplification

[0108]

[0109] Since the simplified path in the global map passes through smaller obstacles that have been removed, sampling optimization was performed after the initial sampling results. The restricted sampling Smooth-RRT* algorithm performed iterative sampling search within the sampling area to optimize the impact of smaller obstacles on the final path. Simulation results are shown in Figures 8(d) and 8(e) and Table 4 (Figures 8(d) and 8(e) are path maps with a relatively large number of iterations from 30 simulations, and Table 4 is the average of the results from 30 simulations). After verification through multiple simulation experiments, the restricted sampling Smooth-RRT* algorithm search method used in this invention has a faster search speed and fewer iterations compared to the RRT* algorithm. Furthermore, due to the significant reduction in the number of sampling points, the memory required for the search is also smaller, the number of iterations is reduced by 78.8%, the algorithm convergence time is reduced by 56.5%, and the path length is reduced by 21.9%.

[0110] Table 4

[0111]

[0112] Based on the simulation experiments above, it can be seen that the restricted sampling Smooth-RRT* algorithm improves the convergence speed and sampling efficiency. As shown in Figure 8(f), after path post-processing, the obtained path curvature is continuous, so the global path is smoother and more suitable for global path planning of autonomous vehicles.

[0113] The above examples of the present invention are merely illustrative of the computational model and process of the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is impossible to exhaustively list all possible implementations here. Any obvious variations or modifications derived from the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A global path planning method for unmanned vehicles, characterized in that, The method specifically includes the following steps: Step 1: After dividing the global map into grids, binarize the information in the global grid map to obtain the binarized global grid map. Obstacles with areas smaller than a threshold are removed from the binarized global raster map to obtain the processed global raster map. Step 2: In the global grid map processed in Step 1, a graph search algorithm is used to search for a path for the autonomous vehicle from the starting point to the destination. Step 3: Put the obstacles deleted in Step 1 back into the global map and determine whether the path found in Step 2 passes through the obstacles. Step 4: If the searched path passes through obstacles, establish a sampling area based on the searched path, delete the obstacles deleted in Step 1, and then use a graph search algorithm to search for a new path for the autonomous vehicle from the starting point to the end point within the sampling area before returning to Step 3. Otherwise, if the searched path does not pass through obstacles, the searched path will be used directly as the initial guiding path, and step five will be executed. Step 5: In the global raster map processed in Step 1, determine the sampling area based on the initial guiding path; Step 6: Within the determined sampling area, use the RRT* algorithm to search for a path from the starting point to the ending point, and put the obstacles deleted in Step 1 back into the global map; Determine whether the path found by the RRT* algorithm passes through obstacles. If the path found by the RRT* algorithm passes through obstacles, proceed to step seven; otherwise, proceed directly to step eight. Step 7: Adjust the expansion factor Then, return to step five; Step 8: Determine whether the searched path meets the iteration stopping condition; If the iteration stopping condition is met, the searched path is the final path planning result, and then step nine is executed; Otherwise, if the iteration stopping condition is not met, the sampling region is established based on the searched path, and then the process returns to step six. Step 9: Smooth the final path planning result from Step 8. The specific process of smoothing is as follows: Represent the new state node as a point. Represent the current vertex as a point , will point The extension of the path direction and the point Let point P be the intersection of the extended lines of the path directions. The straight-line distance between point P and point P is expressed as , will point The straight-line distance between point P and point P is expressed as At point P and point Take a virtual node on the connection. To satisfy virtual nodes The straight-line distance between point P and point P is And virtual nodes With point The directions are the same; At point With virtual nodes Between, the driverless car maintained a constant speed and yaw rate Make an arc motion; in, It is an arc The angle it faces; In virtual nodes With point Between, the driverless car maintained a constant speed and yaw rate To move in a straight line, that is... It equals 0.

2. The global path planning method for unmanned vehicles according to claim 1, characterized in that, In step one, the information in the global raster map is binarized, specifically as follows: 0 represents the non-obstacle area in the global grid map, and 1 represents the obstacle area in the global grid map.

3. The global path planning method for unmanned vehicles according to claim 2, characterized in that, The graph search algorithm used in step two is the A* algorithm.

4. The global path planning method for unmanned vehicles according to claim 3, characterized in that, The specific process for determining the sampling area based on the initial guidance path is as follows: Step 1: Calculate the path dilation distance: in, This represents the path dilation distance. This represents the side length of the global map. For expansion factors; Step 2: Expand the sampling area using the initial guide path as the center, and the distance between the initial guide path and the two boundaries of the sampling area is the path expansion distance. .

5. The global path planning method for an unmanned vehicle according to claim 4, characterized in that, The iteration stopping condition is: in continuous The same path was found during each iteration.

6. The global path planning method for an unmanned vehicle according to claim 5, characterized in that, The new node generated by the RRT* algorithm The following constraints must be satisfied: in: For new nodes The current cost, The total cost of the current optimal path. This is the Euclidean distance from the new node to the target point.

7. The global path planning method for an unmanned vehicle according to claim 6, characterized in that, The speed and speed for: Define velocity profile , It's an autonomous vehicle at the point. driving speed, It's an autonomous vehicle at the point. The speed of travel; speed and Randomly selected from the velocity profile V.

8. The global path planning method for an unmanned vehicle according to claim 7, characterized in that, The speed and yaw rate satisfy the following constraints: in, It is the minimum acceleration of the autonomous vehicle along the direction of travel. It is the maximum acceleration of the driverless car along the direction of travel. It is the minimum value of the yaw angle acceleration of the autonomous vehicle. Δt is the maximum value of the yaw angle acceleration of the autonomous vehicle, and Δt is the planned time step.

Images