A short-time bidirectional search algorithm automatic parking path planning method
By integrating bidirectional search algorithm and grid-based optimization of hybrid A* algorithm, the problems of excessively long search time and difficulty in node overlap in path planning in automatic parking system are solved, achieving more efficient path planning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHAANXI SCI TECH UNIV
- Filing Date
- 2023-02-14
- Publication Date
- 2026-04-21
AI Technical Summary
In existing automated parking systems, the hybrid A* algorithm has an excessively long search time and the expansion nodes of forward and reverse searches are not easily overlapped, resulting in low path planning efficiency.
A hybrid A* algorithm, which integrates bidirectional search, is adopted. The node search process is optimized by combining the grid method. By interrupting the search in the grid and merging paths through forward and reverse searches, node management is optimized and search efficiency is improved.
It shortens the search time for automatic parking path planning, improves the efficiency and accuracy of path planning, reduces the search for invalid nodes, and enhances the optimization effect of the path.
Smart Images
Figure CN116295483B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence, and in particular to a method for automatic parking path planning using a short-time bidirectional search algorithm. Background Technology
[0002] The widespread use of automobiles has expanded people's travel range and made transportation more convenient. However, this convenience has also brought many safety issues, with parking being a major challenge. The emergence of automatic parking systems has largely avoided traffic problems caused by limited visibility and other factors. Current automatic parking systems mainly consist of five components: environmental perception, parking space detection and recognition, parking path planning, parking path following control, and simulation display. This patent focuses on optimizing the parking path planning component.
[0003] Current automated parking systems divide parking path planning into two main parts: global planning and local planning. In global planning algorithms, most automated parking systems use algorithms such as Dubins' algorithm and A* algorithm, while a few use a hybrid A* algorithm. Only by combining a superior global path planning algorithm with the system can the vehicle achieve a better automated parking effect in subsequent local planning and control tracking. Summary of the Invention
[0004] Purpose of the invention: To provide a more effective short-time bidirectional search algorithm for automatic parking path planning, the specific purpose of which is described in the detailed implementation section for several substantial technical effects.
[0005] To achieve the above objectives, the present invention adopts the following technical solution:
[0006] A method for automatic parking path planning using a short-time bidirectional search algorithm is characterized by the following steps: incorporating a hybrid A* algorithm that integrates a bidirectional search algorithm, using a grid method to optimize the bidirectional search algorithm to solve the problem that the expanded nodes of the forward and reverse searches in the bidirectional search are not easy to overlap, optimizing the node search process of the hybrid A* algorithm, and integrating the bidirectional search algorithm to accelerate the search time of the hybrid A* algorithm in the automatic parking path planning process.
[0007] A further technical solution of the present invention is that, in the global path planning stage of automatic parking, under the scheduling of the automatic parking system via vehicle sensors, the starting position, final parking position, and map information of the vehicle are known, and path planning for the global path of automatic parking begins: initializing the grid and four lists; calculating the Euclidean distance from the starting point to the ending point, and using this as the heuristic value for the hybrid A* algorithm; forward search using the current position of the vehicle as the starting point and the final parking position as the ending point to select the node with the minimum cost value as the parent node; using the hybrid A* algorithm to search for neighboring points and obtain the cost of this action; determining the cost of the parent node and the ending point to obtain the estimated cost h, determining the cost of the parent node and the starting point to obtain the actual cost g, and determining whether the forward search and the reverse search have the same expansion point; repeating the operation until the forward and reverse searches have the same expansion point, merging the lists and outputting curve information.
[0008] A further technical solution of the present invention lies in the following key steps of improving the bidirectional search hybrid A* algorithm:
[0009] 1) Create four lists: openlist1, openlist2, closelist1, and closelist2. Openlist1 stores nearby expanded points in the forward search, openlist2 stores nearby expanded points in the reverse search, closelist1 stores expanded points in the forward search, and closelist2 stores expanded points in the reverse search; set a judgment variable T to determine whether it is a forward search or a reverse search.
[0010] 2) Put the starting point into closelist1 and the ending point into closelist2, and set D = 0;
[0011] 3) If D = 0, proceed to step 4); if D = 1, proceed to step 11.
[0012] 4) Traverse openlist1, find the minimum point of the cost estimation function, which represents the Euclidean distance from the point to the destination, add it as the current point to closelist1, and delete it from openlist1; otherwise, there is no path.
[0013] 5) Check the index value of the current grid cell to determine if there is only one node in the grid cell with that index value. If so, skip this step. Otherwise, compare the heuristic function values of the nodes in the grid cell, take the node with the minimum value, and delete the node with the larger value.
[0014] 6) Determine if the current point is the destination; if yes, execute 18); otherwise, execute 7).
[0015] 7) If the current point is in closelist2, update the optimal path distance through the meeting points. If it is less than or equal to the optimal path through the non-meeting points, proceed to step 18); otherwise, proceed to step 8); if the current point is not in closelist2, proceed to step 8).
[0016] 8) Update the neighboring nodes of the current node, and update their heuristic function value, optimal path value, and parent node;
[0017] 9) If the current node is in openlist2, update the optimal path distance through the meeting point. If it is less than or equal to the optimal path distance through the non-meeting point, then execute step 18); otherwise, execute step 10).
[0018] 10) Set D = 1, then return to step 3);
[0019] 11) Traverse openlist2, find the minimum point, which represents the Euclidean distance from the point to the starting point; add it as the current point to closelist2 and delete it from openlist2; otherwise, there is no path.
[0020] 12) Check the index value of the current grid cell to determine if there is only one node in the grid cell with that index value. If so, skip this step. Otherwise, compare the heuristic function values of the nodes in the grid cell, take the node with the minimum value, and delete the node with the larger value.
[0021] 13) Determine if the current point is the starting point; if yes, execute 18); otherwise, execute 14).
[0022] 14) If the current point is in closelist1, update the optimal distance of the optimal path through the meeting point; if it is less than or equal to the optimal path distance through the non-meeting point, proceed to step 18); otherwise, proceed to step 17); if the current point is not in closelist2, proceed to step 15.
[0023] 15) Update the neighboring nodes of the current node, and update their heuristic function value, optimal path value, and parent node;
[0024] 16) If the current point is in openlist1, update the optimal path distance through the meeting point; if it is less than or equal to the optimal path distance through the non-meeting point, proceed to step 18); otherwise, proceed to step 17.
[0025] 17) Set D = 0, then return to step 7);
[0026] 18) Output the final path.
[0027] A further technical solution of the present invention is that the principle of the hybrid A* algorithm can be regarded as A*
[0028] The algorithm, generated by combining an actual vehicle dynamics model, searches for child nodes in all directions from the starting point, calculates the cost of each child node, and selects the point with the minimum cost as the next parent node, until the end point is reached. The hybrid A* algorithm search method mainly consists of two modes: forward and backward. Combined with the vehicle model, it divides the corners evenly and searches for child nodes within the corner range. It calculates whether the vehicle pose transformation from the current vehicle pose to the child node will collide with obstacles, and establishes an Openlist table to receive the child nodes searched within the corner range, and a Closelist table to receive the selected parent nodes and the child node with the minimum cost.
[0029] A further technical solution of the present invention is that,
[0030] The cost function for hybrid A* is:
[0031] f(n)=g(n)+h(n) (1)
[0032] Where f(n) is the cost of the current node n; g(n) represents the path cost from the starting point to the current point n; h(n) is the estimated cost from the current point to the target point. The hybrid A* algorithm uses the heuristic function of the A* algorithm, which includes: Manhattan distance, Euclidean distance, and Chebyshev distance. Here, Euclidean distance is selected to calculate the estimated cost h(n) of the node.
[0033] The Euclidean distance expression is:
[0034]
[0035] Where p1 is the current point, x1 is the x-coordinate of the current point, and y1 is the y-coordinate of the current point; p2 is the coordinate of the destination, x2 is the x-coordinate of the destination, and y2 is the y-coordinate of the destination.
[0036] A further technical solution of the present invention is to use a grid method to divide the map and to record the grid to facilitate the management of the relationship between the grid and the child nodes, thereby solving the problem that paths may not intersect. Using the divided grid, when the child nodes of the forward search and the reverse search are in the same grid, the search process is interrupted, the parent node of the reverse search is linked to the child node of the forward search, and finally the entire path is output, thus obtaining a final path composed of the forward search and the reverse search.
[0037] A further technical solution of the present invention is that,
[0038] When searching for a planned path, both forward and reverse searches are performed simultaneously, and the trajectory conforms to the vehicle structure, which can be directly output to provide control information.
[0039] By dividing the map into regions, and selecting only one usable node for each region, the path can be updated and the optimal path can be guaranteed.
[0040] This ensures that the final path can be generated when the child nodes of the forward and reverse searches only need to overlap within one region, and do not need to completely overlap.
[0041] The present invention, employing the above technical solution, offers the following advantages over existing technologies: Addressing the issue of excessively long search times in the hybrid A* algorithm during automatic parking path planning, this patent proposes a hybrid A* algorithm incorporating a bidirectional search algorithm. Furthermore, to address the difficulty of overlapping nodes in forward and reverse searches during bidirectional search, a grid-based optimization method is used to improve the bidirectional search algorithm, increasing the probability of successful node retrieval during forward and reverse searches. The node search process of the hybrid A* algorithm is also optimized, and the integration of the bidirectional search algorithm accelerates the search time in automatic parking path planning. This patent primarily optimizes the long search time of the hybrid A* algorithm in automatic parking by designing an improved hybrid A* algorithm that integrates a bidirectional search algorithm. It sets a start and end point as initial points, incorporates vehicle collision information, and sets a reasonable cost function to mutually search for optimal path information. This effectively reduces the numerous invalid nodes in the later stages of unidirectional search and decreases the number of searches. Simultaneously, the use of a grid-based method to divide the map effectively avoids the possibility of repeated searches within the same area. Furthermore, the combination of the grid method and the bidirectional search algorithm significantly increases the likelihood of intersections during forward and reverse searches. Attached Figure Description
[0042] To further illustrate the present invention, the following description is provided in conjunction with the accompanying drawings:
[0043] Figure 1 This is a schematic diagram illustrating the principle of the hybrid A* algorithm.
[0044] Figure 2 This is a schematic diagram illustrating the principle of the A* algorithm.
[0045] Figure 3 This is one of the diagrams illustrating bidirectional search optimization.
[0046] Figure 4 This is the second diagram illustrating bidirectional search optimization.
[0047] Figure 5 A schematic diagram of the reverse parking path generated by the fusion algorithm;
[0048] Figure 6 A schematic diagram showing the path generated for reversing into a parking space after the obstacle is changed in the fusion algorithm;
[0049] Figure 7 A schematic diagram illustrating the path generation for reverse parking using the traditional hybrid A* algorithm;
[0050] Figure 8 This is a schematic diagram of the path generation for lateral parking using the traditional hybrid A* algorithm.
[0051] Figure 9 This is a schematic diagram of the path generated for lateral parking using the fusion algorithm. Detailed Implementation
[0052] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.
[0053] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0054] This patent provides multiple parallel solutions; the different descriptions represent improved solutions or parallel solutions based on the basic solution. Each solution has its own unique characteristics. Furthermore, the technical features involved in the different embodiments of the invention described below can be combined with each other as long as they do not conflict with each other. Fixing methods not described herein can be any type of fixing, such as threaded fixing, bolt fixing, or adhesive bonding.
[0055] The present patent will be further described in conjunction with the specific embodiments described below.
[0056] In the global path planning stage of automatic parking, under the scheduling of the automatic parking system via vehicle sensors, the starting position, final parking position, and map information of the vehicle are known, and the path planning of the global path for automatic parking begins: initializing the grid and four lists; calculating the Euclidean distance from the starting point to the destination, and using this as the heuristic value for the hybrid A* algorithm; forward search using the current position of the vehicle as the starting point and the final parking position as the destination (reverse search is the opposite) to select the node with the minimum cost value as the parent node; using the hybrid A* algorithm to search for neighboring points and obtain the cost of this action; judging the cost of the parent node and the destination to obtain the estimated cost h, judging the cost of the parent node and the starting point to obtain the actual cost g, and judging whether the forward search and the reverse search have the same expansion point; repeating the operation until the forward and reverse searches have the same expansion point, merging the lists and outputting curve information.
[0057] The key steps to improve the bidirectional search hybrid A* algorithm are as follows:
[0058] 1) Create four lists: openlist1, openlist2, closelist1, and closelist2. openlist1 stores neighboring points expanded during the forward search, openlist2 stores neighboring points expanded during the reverse search, closelist1 stores expanded points during the forward search, and closelist2 stores expanded points during the reverse search. Set a conditional variable T to determine whether the search is forward or reverse.
[0059] 2) Put the starting point into closelist1 and the ending point into closelist2, and set D=0.
[0060] 3) If D = 0, proceed to step 4); if D = 1, proceed to step 11.
[0061] 4) Traverse openlist1, find the minimum point of the cost estimation function, which represents the Euclidean distance from the point to the destination, add it as the current point to closelist1, and delete it from openlist1; otherwise, there is no path.
[0062] 5) Check the index value of the current grid cell to determine if there is only one node in the grid cell with that index value. If so, skip this step. Otherwise, compare the heuristic function values of the nodes in the grid cell, take the node with the minimum value, and delete the node with the larger value.
[0063] 6) Determine if the current point is the destination; if yes, execute 18); otherwise, execute 7).
[0064] 7) If the current point is in closelist2, update the optimal path distance through the meeting points. If it is less than or equal to the optimal path distance through the unmet points, proceed to step 18); otherwise, proceed to step 8). If the current point is not in closelist2, proceed to step 8).
[0065] 8) Update the neighboring nodes of the current node, and update their heuristic function value, optimal path value, and parent node.
[0066] 9) If the current node is in openlist2, update the optimal path distance through the meeting point. If it is less than or equal to the optimal path distance through the non-meeting point, proceed to step 18); otherwise, proceed to step 10).
[0067] 10) Set D = 1 and return to step 3).
[0068] 11) Traverse openlist2, find the minimum point, which represents the Euclidean distance from the point to the starting point. Add it as the current point to closelist2 and delete it from openlist2; otherwise, no path exists.
[0069] 12) Check the index value of the current grid cell to determine if there is only one node in the grid cell with that index value. If so, skip this step. Otherwise, compare the heuristic function values of the nodes in the grid cell, take the node with the minimum value, and delete the node with the larger value.
[0070] 13) Determine if the current point is the starting point; if yes, execute 18); otherwise, execute 14).
[0071] 14) If the current point is in closelist1, update the optimal distance of the optimal path through the meeting points. If it is less than or equal to the optimal path distance through the unmet points, proceed to step 18); otherwise, proceed to step 17). If the current point is not in closelist2, proceed to step 15).
[0072] 15) Update the neighboring nodes of the current node, and update their heuristic function value, optimal path value and parent node.
[0073] 16) If the current point is in openlist1, update the optimal path distance through the meeting point. If it is less than or equal to the optimal path distance through the unmet point, proceed to step 18); otherwise, proceed to step 17).
[0074] 17) Set D = 0 and return to step 7).
[0075] 18) Output the final path.
[0076] Refer to and compare Figure 1 and Figure 2 1. The principle of the hybrid A* algorithm can be seen as an algorithm generated by combining the A* algorithm with an actual vehicle dynamics model. Its principle is similar to the A* algorithm: starting from the starting point, it searches for child nodes in all directions, calculates the cost of each child node, and selects the point with the minimum cost as the next parent node, continuing until the end point is reached. Unlike the A* algorithm, the hybrid A* algorithm's search method mainly consists of two modes: forward and backward. It combines this with the vehicle model, evenly dividing the corners and searching for child nodes within the corner range. It calculates whether the vehicle pose transformation from the current vehicle pose to the child node will collide with obstacles, and establishes an Openlist to receive child nodes searched within the corner range, and a Closelist to receive the already selected parent nodes and the child node with the minimum cost.
[0077] Figure 1 This is the search pattern for search points using the A* algorithm. Figure 2It is the search mode of the hybrid A* algorithm for searching neighboring points, which follows dynamic constraints.
[0078] The cost function for hybrid A* is:
[0079] f(n)=g(n)+h(n))(1
[0080] Where f(n) is the cost of the current node n; g(n) represents the path cost from the starting point to the current point n; and h(n) is the estimated cost from the current point to the target point. The hybrid A* algorithm uses the heuristic functions of the A* algorithm, mainly including Manhattan distance, Euclidean distance, and Chebyshev distance. This patent uses Euclidean distance to calculate the estimated cost h(n) of the node.
[0081] The Euclidean distance expression is:
[0082]
[0083] Where p1 is the current point, x1 is the x-coordinate of the current point, and y1 is the y-coordinate of the current point; p2 is the coordinate of the destination, x2 is the x-coordinate of the destination, and y2 is the y-coordinate of the destination.
[0084] refer to Figure 3 and Figure 4 2. The search process from the starting point to the ending point is inherently uncertain. The search path from the starting point to the ending point and the search path from the ending point to the starting point may not intersect completely. If they don't intersect, both paths will continue until the target point is reached, which is detrimental to optimizing the hybrid A* algorithm. This patent uses a grid system to divide the map and sequences the grids to facilitate managing the relationships between grids and child nodes. To address the issue of paths not necessarily intersecting, the patent uses the grid to interrupt the search process when the child nodes of the forward and reverse searches are in the same grid. It then links the parent node of the reverse search to the child node of the forward search, finally outputting the entire path. This results in a final path composed of the forward and reverse searches.
[0085] This experiment uses a simulation platform to draw a rectangle to represent a real vehicle model. The parking position is determined by randomly selecting a horizontal driving point and a vertical position from the destination. The requirement is that the front of the car should face outwards. This experiment only requires the length of the planned path, the frequency of the vehicle's forward and backward movement, and the curvature of the steering, without analyzing the vehicle's speed or the forces acting on it.
[0086] Vertical parking is a common scenario when parking, and the difficulty often lies in the situation where one parking space becomes available while the spaces on both sides are occupied. This can be challenging for inexperienced drivers. This paper simulates the path planning process using an improved hybrid A* algorithm in this scenario.
[0087] Figure 5 The red X direction in the center represents the rear of the vehicle. Red dots represent obstacles, and blue dots represent the path curve generated by the algorithm. The overall driving process is from... Figure 5 The system senses surrounding obstacles and uses algorithms to calculate the path that avoids obstacles and minimizes costs while adhering to dynamic constraints, ultimately reaching the location shown in the lower square in the diagram, such as a garage.
[0088] Depend on Figure 5 , Figure 6 It can be seen that the planned paths differ significantly with and without obstacles. When the penalty cost is the same, the average computation time for generating the path curve with obstacles is 31.34 seconds, and there is also reversing behavior. Furthermore, the path planning time fluctuates considerably when the obstacles change, making it difficult to intuitively compare the differences between the data. Therefore, this study only changes the starting pose, keeping the map obstacle information unchanged, and compares the search time, number of direction changes, and number of nodes for unidirectional and bidirectional path search using the improved hybrid A* algorithm. The results are shown in Table 1.
[0089] Table 1 Comparison of Improved Vertical Inbound Algorithms
[0090]
[0091] Another simulation experiment on parallel parking was added. Here, the improved hybrid A* algorithm was used to plan the path in this case, while changing the starting point. The search path time, number of direction changes, and number of nodes of the traditional hybrid A* algorithm and the improved hybrid A* algorithm were compared in the case of parallel parking, as shown in Table 2.
[0092] Table 2 Comparison of Improved Algorithms for Parallel Parking
[0093]
[0094]
[0095] In summary, for both reversing into a parking space and parallel parking, the improved hybrid A* algorithm reduced the search time by an average of 25.21% compared to the original hybrid A* algorithm.
[0096] The sources and definitions of the terms appearing in this patent document are as follows:
[0097] Openlist: A collection that stores nearby expansion points;
[0098] Closelist: A collection of points that have been expanded;
[0099] Cost: In heuristic search algorithms, it is a standard for priority judgment. The lower the cost, the higher the priority, and vice versa.
[0100] Parent node: In graph search algorithms, the parent node is the node that has been expanded to the next higher level of the current node.
[0101] Child node: In graph search algorithms, the next level node of the current node is the child node of the current node, and the current node is the child node of the parent node;
[0102] Heuristic function: Derived from heuristic search algorithms, it is a function that guides the search in a directional way and improves search efficiency;
[0103] Euclidean distance: a type of heuristic function, specifically representing the straight-line distance between two coordinate points;
[0104] Raster method: Dividing a map into rows and columns according to certain rules to form many regular grids;
[0105] Experimental coordinates [x, y, θ]: where x is the x-coordinate of the center point of the vehicle rectangle, y is the y-coordinate of the center point of the vehicle rectangle, and θ is the angle between the front of the vehicle rectangle (opposite to the red X) and the x-axis;
[0106] Graph search algorithms: Map data can often be represented using data structures such as graphs. The search algorithms used in graph structures are called graph search algorithms.
[0107] It should be noted that the multiple solutions provided in this patent include their own basic solutions, which are independent of each other and do not restrict each other. However, they can also be combined with each other without conflict to achieve multiple effects.
[0108] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claims.
Claims
1. A method for automatic parking path planning using a short-time bidirectional search algorithm, characterized in that, The process includes the following steps: a hybrid A* algorithm incorporating a bidirectional search algorithm; optimization of the bidirectional search algorithm using a grid method to address the problem that nodes in the forward and reverse searches are not easily overlapped in the bidirectional search; optimization of the node search process of the hybrid A* algorithm; and acceleration of the search time of the hybrid A* algorithm in the automatic parking path planning process by integrating the bidirectional search algorithm. The key steps to improve the bidirectional search hybrid A* algorithm are as follows: 1) Create four lists: openlist1, openlist2, closelist1, and closelist2. Openlist1 stores the nearest expanded points in the forward search, openlist2 stores the nearest expanded points in the reverse search, closelist1 stores the expanded points in the forward search, and closelist2 stores the expanded points in the reverse search. Set a judgment variable T to determine whether the search is forward or backward; 2) Put the starting point into closelist1 and the ending point into closelist2, and set D=0; 3) If D=0, proceed to step 4). If D=1, proceed to step 11). 4) Traverse openlist1, find the minimum point of the cost estimation function, which represents the Euclidean distance from the point to the destination, add it as the current point to closelist1, and delete it from openlist1; otherwise, there is no path. 5) Check the index value of the current grid cell to determine if there is only one node in the grid cell with that index value. If so, skip this step. Otherwise, compare the heuristic function values of the nodes in the grid cell, take the node with the minimum value, and delete the node with the larger value. 6) Determine if the current point is the destination; if so, execute step 18); otherwise, execute step 7). 7) If the current point is in closelist2, update the optimal path distance through the meeting points. If it is less than or equal to the optimal path distance through the unmet points, proceed to step 18); otherwise, proceed to step 8); if the current point is not in closelist2, proceed to step 8). 8) Update the neighboring nodes of the current node, and update their heuristic function value, optimal path value, and parent node; 9) If the current node is in openlist2, update the optimal path distance through the meeting point. If it is less than or equal to the optimal path distance through the non-meeting point, proceed to step 18); otherwise, proceed to step 10). 10) Set D=1, then return to step 3). 11) Traverse openlist2, find the minimum point, which represents the Euclidean distance from the point to the starting point; add it as the current point to closelist2 and delete it from openlist2; otherwise, there is no path. 12) Check the index value of the current grid cell to determine if there is only one node in the grid cell with that index value. If so, skip this step. Otherwise, compare the heuristic function values of the nodes in the grid cell, take the node with the minimum value, and delete the node with the larger value. 13) Determine if the current point is the starting point; if so, execute 18); otherwise, execute 14). 14) If the current point is in closelist1, update the optimal distance of the optimal path through the meeting points; if it is less than or equal to the optimal path distance through the non-meeting points, proceed to step 18); otherwise, proceed to step 17). If the current point is not in closelist2, proceed to step 15). 15) Update the neighboring nodes of the current node, and update their heuristic function value, optimal path value, and parent node; 16) If the current point is in openlist1, update the optimal path distance through the meeting point; if it is less than or equal to the optimal path distance through the non-meeting point, proceed to step 18); otherwise, proceed to step 17). 17) Set D=0 and return to step 7). 18) Output the final path.
2. The method for automatic parking path planning using a short-time bidirectional search algorithm as described in claim 1, characterized in that, In the global path planning stage of automatic parking, under the scheduling of the automatic parking system via vehicle sensors, the starting position, final parking position, and map information of the vehicle are known, and path planning for the global automatic parking path begins: initializing the grid and four lists; calculating the Euclidean distance from the starting point to the destination, and using this as the heuristic value for the hybrid A* algorithm; forward search using the vehicle's current position as the starting point and the final parking position as the destination to select the node with the minimum cost value as the parent node; using the hybrid A* algorithm to search for neighboring points and obtain the cost of this action; determining the cost of the parent node and the destination to obtain the estimated cost h, determining the cost of the parent node and the starting point to obtain the actual cost g, and determining whether the forward search and the reverse search have the same expansion point; repeating the operation until the forward and reverse searches have the same expansion point, merging the lists, and outputting curve information.
3. The method for automatic parking path planning using a short-time bidirectional search algorithm as described in claim 1, characterized in that, The Hybrid A* algorithm is viewed as an algorithm generated by combining the A* algorithm with an actual vehicle dynamics model. It searches for child nodes in all directions from the starting point, calculates the cost of each child node, and selects the point with the minimum cost as the next parent node, until the end point is reached. The Hybrid A* algorithm has two search modes: forward and backward. It combines the vehicle model, divides the corners evenly, and searches for child nodes within the corner range. It calculates whether the vehicle pose transformation from the current vehicle pose to the child node will collide with obstacles. It establishes an Openlist to receive child nodes searched within the corner range and a Closelist to receive the selected parent nodes and the child node with the minimum cost.
4. The method for automatic parking path planning using a short-time bidirectional search algorithm as described in claim 3, characterized in that, The cost function for hybrid A* is: ; Where f(n) is the cost of the current node n; g(n) represents the path cost from the starting point to the current point n; h(n) is the estimated cost from the current point to the target point. The hybrid A* algorithm uses the heuristic function of the A* algorithm, which includes: Manhattan distance, Euclidean distance, and Chebyshev distance. Here, Euclidean distance is selected to calculate the estimated cost h(n) of the node. The Euclidean distance expression is: ; Where p1 is the current point, x1 is the x-coordinate of the current point, and y1 is the y-coordinate of the current point; p2 is the coordinate of the destination, x2 is the x-coordinate of the destination, and y2 is the y-coordinate of the destination.
5. The method for automatic parking path planning using a short-time bidirectional search algorithm as described in claim 4, characterized in that, The map is divided using a grid method, and the grids are ordered to facilitate the management of the relationship between grids and child nodes, thus solving the problem that paths may not intersect. Using the divided grids, when the child nodes of the forward search and the reverse search are in the same grid, the search process is interrupted, the parent node of the reverse search is linked to the child node of the forward search, and finally the entire path is output, thus obtaining a final path composed of the forward search and the reverse search.
6. The method for automatic parking path planning using a short-time bidirectional search algorithm as described in any one of claims 1-5, characterized in that, When searching for a planned path, both forward and reverse searches are performed simultaneously, and the trajectory conforms to the vehicle structure, which can be directly output to provide control information. By dividing the map into regions, and selecting only one usable node for each region, the path can be updated and the optimal path can be guaranteed. This ensures that the final path can be generated when the child nodes of the forward and reverse searches only need to overlap within one region, and do not need to completely overlap.