An obstacle avoidance path planning method for intelligent buses based on genetic algorithm and extended RRT

Abstract

The invention discloses an intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT, belonging to the technical field of intelligent vehicle path planning. Firstly, a sampling flexible boundary is set according to a global path, the sampling space of an RRT algorithm is reduced, and a feasible path is guided to be close to the global path by heuristically expanding nodes. Then, a metric function is designed for a search tree node to select a parent node, a simplified bus model is established to detect collision, and then a path optimization area is constructed. The genetic algorithm is used to solve the optimal obstacle avoidance path, and finally a Bessel interpolation method is used to smooth the path. The invention can take into account both path quality and safety, quickly plan an obstacle avoidance local path, significantly improve the safety of the vehicle during driving, and effectively reduce the occurrence rate of vehicle accidents.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent vehicle path planning, and in particular relates to an intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT. Background Art

[0002] Vehicle obstacle avoidance planning is a key technology in the fields of autonomous driving, drone navigation, robot navigation, etc. The purpose is to enable the vehicle to avoid obstacles safely and smoothly and reach the target location. This problem usually involves multiple technical fields such as path planning, environmental perception, and control systems.

[0003] Vehicle path planning technology is one of the core technologies to solve the obstacle avoidance planning of autonomous driving. At present, path planning algorithms can be divided into two categories: the first is the traditional path planning algorithm, which is usually based on graph theory and mathematical optimization technology, such as A* algorithm, Dijkstra algorithm, RRT algorithm; the other is the intelligent algorithm, which is generally used to deal with complex, high-dimensional, uncertain and dynamic environments, mainly including genetic algorithms, ant colony algorithms, etc. Traditional path planning algorithms are suitable for static and structured environments and can find the global optimal solution, but their performance in complex, large-scale or dynamic environments is limited. Intelligent path planning algorithms can handle more complex scenarios through heuristic and swarm intelligence methods. Although they cannot guarantee the global optimal, they can quickly find feasible or approximately optimal paths. Therefore, in actual application scenarios, the combination of traditional algorithms and intelligent algorithms can often take into account both solution efficiency and quality, and achieve more ideal planning results. Summary of the invention

[0004] In order to solve at least one of the problems raised in the above background technology, the present invention provides an intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for intelligent bus obstacle avoidance path planning based on genetic algorithm and extended RRT, comprising the following steps:

[0006] Step 1: Set the flexible boundary of sampling: add sampling boundaries according to the global path, reduce the sampling space of the RRT algorithm, and improve the algorithm solution speed;

[0007] Step 2, heuristic expansion node: guided by the global path and path target point, control the generation position of sampling points, guide the feasible path to be close to the global path, and improve the generation speed of the feasible path;

[0008] Step 3, parent node measurement and reselection: design a measurement function for the search tree node based on the principle of being away from obstacles and close to the global path, and select the parent node for the new node with the best performance;

[0009] Step 4: Obstacle collision detection: Establish a simplified physical model of the bus, set a safe distance, and detect whether there is a danger of collision with obstacles during the driving process;

[0010] Step 5: Expand the feasible path area: construct the path optimization area centered on the feasible path searched by the RRT algorithm;

[0011] Step 6: Optimize the feasible path: Based on factors such as path length and path smoothness, a genetic algorithm is used to solve the feasible path in the optimization area to calculate the optimal bus obstacle avoidance path.

[0012] Step 7: Smooth optimization path: Use the Bessel interpolation method to smoothly optimize the path and obtain the final obstacle avoidance path.

[0013] Preferably, the step 1 of setting the flexible boundary of the sampling includes the following steps:

[0014] Step 11: Establish a rectangular coordinate system with the center point of the rear axle of the current vehicle as the origin O, the vehicle heading direction as the X-axis, and the Y-axis perpendicular to the driving direction;

[0015] Step 12: On the map, translate the global path to both sides along the Y axis Distance, the range passed by the global path during the translation process forms a planar closed area and is set as the sampling space. The sampling space is defined as:

[0016]

[0017] in is the global path fitting expression, and They are the x-coordinates of the lowest and highest points of the global path on the x-axis respectively.

[0018] Preferably, the step 2 of heuristically expanding the node includes the following steps:

[0019] Step 21: Sampling point selection in the global path; setting the target probability when the RRT algorithm performs random sampling , the probability of sampling Find the latest node as the node to be expanded, set the point closest to the node to be expanded in the global path as the nearest node, and then use the global path point with the nearest node Lc as the sampling point for expansion; Random sampling in the feasible area of ​​the sampling space;

[0020] Step 22: Design a global path gravity function to inspire and expand sampling points; when random sampling is selected within the feasible region of the sampling space, design a gravity function for the global path;

[0021] Step 23: Set the sampling point reselection strategy: Initialize the detection circle radius , when the sampling point is selected After that, the sampling point is the center of the circle, The detection circle is formed with a radius, and it is determined whether there is a node of the search tree within the detection circle. If so, it is considered that the sampling point The selection will cause repeated search and reselection of sampling points, otherwise the sampling points are considered feasible.

[0022] Preferably, the step 3 of measuring and reselecting the parent node includes the following steps: The nearest node to the sampling point Expand to get new nodes After that, it is a new node The nodes within a certain range are assigned a specified type of metric component, called metric nodes, which allows new nodes to It can judge the pros and cons of the existing node positions and select the best node for reconnection.

[0023] Preferably, the obstacle collision detection in step 4 includes the following steps:

[0024] Step 41: Set the safe distance , establish the bus physical model, make the minimum circumscribed rectangle of the bus physical model to replace the bus, and extend the length and width of the circumscribed rectangle respectively , forming an obstacle detection rectangle, taking the midpoint of the rear axle of the bus as the base point, and detecting the coordinates of the four vertices of the rectangle;

[0025] Step 42: Before the new node selects a parent node to try to connect, a detection rectangle is made with the new node and the parent node as the center of the rear axle of the vehicle and the connection direction as the vehicle heading direction, and the two detection rectangles are filled to form an obstacle detection frame containing the connection line. If there is an obstacle in the obstacle detection frame, it is considered that the connection has failed and returns to resampling; otherwise, it is determined whether the distance between the new node and the target point is less than the threshold If yes, the new node is considered as the target point, and a feasible path is obtained by returning to the parent node of the search tree, otherwise the next round of sampling connection is performed.

[0026] Preferably, the step 5 of expanding the feasible path area includes the following steps:

[0027] Step 51: Expand the starting point and end point of the feasible path searched by RRT, and set is the i-th node on the feasible path, and are the starting point and end point of the feasible path respectively. Orientation Direction of movement distance, and then Vertical sideways movement Distance gets two expansion nodes and , get the feasible path end point Two expansion nodes and ;

[0028] Step 52: Expand the intermediate nodes of the feasible path searched by RRT except the starting point and the end point, and set The size is , the node Along The bisector direction extends to both sides Length, get the extended node and ;

[0029] Step 53: Connect all the expansion nodes in sequence to form a polygonal expansion area that includes a feasible path.

[0030] Preferably, the step 6 of optimizing the feasible path includes the following steps:

[0031] Step 61: Build a grid map and divide the bounding rectangle of the expanded area into The grid map of the obstacle area and the grid outside the expansion area is recorded as 0, indicating that the vehicle cannot reach it, and the remaining grids are recorded as 1, indicating the route planning area. Indicates the number of grids in the horizontal direction. Indicates the number of vertical grids; initialization parameters, setting the maximum number of iterations , mutation probability , crossover probability ; Search in raster map A random path from the initial position to the final position constitutes the initial population;

[0032] Step 62: Design a fitness function to arrange the individuals in the population from low to high according to their fitness;

[0033] Step 63: Update and iterate the population. The individual paths in the population are used as parents to perform crossover and mutation operations. The generated offspring individuals form a new population. The roulette wheel selection method is used to screen the individuals in the new population. The fitness of the individuals in the new population is recalculated and sorted. Check whether the number of iterations has reached If yes, then output the optimized path with the highest fitness, otherwise return to step 63 to continue the loop iteration. Calculated by the following formula:

[0034] ;

[0035] in is the current individual path fitness, is the current number of individuals in the population.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. When the present invention performs path planning on a single fast random search tree, node generation lacks target guidance, and the search results have strong randomness, which leads to unstable paths and convergence speeds. In particular, random expansion will lead to unnecessary paths in driving conditions with fewer obstacles, thereby increasing the difficulty of vehicle driving and reducing the comfort of passengers. A smart bus obstacle avoidance path planning method based on genetic algorithm and extended RRT is proposed. By expanding the feasible domain of feasible paths, unnecessary calculation areas are reduced for genetic algorithms, and the effective node selection rate is increased. This method can limit the algorithm search range and improve the subsequent algorithm solution speed, thereby greatly improving the real-time performance of smart bus obstacle avoidance.

[0038] 2. The present invention introduces a global path flexible boundary to constrain the generation range of feasible paths, so that the generated results deviate as little as possible from the original driving route; compared with the results generated by the unconstrained fast random search tree algorithm, this constraint method reduces the randomness of path generation and has higher rationality and feasibility. This method can provide strong support for research and application in situations where actual collection is difficult or unsafe. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the overall flow of the algorithm described in the present invention;

[0040] Figure 2 It is a schematic diagram of the flexible boundary setting of the present invention;

[0041] Figure 3 A schematic diagram of the metric function design of the present invention;

[0042] Figure 4 This is a schematic diagram of an obstacle detection frame of an intelligent bus according to the present invention;

[0043] Figure 5 A schematic diagram of the feasible path area for expanding the present invention;

[0044] Figure 6 It is a schematic diagram of the genetic algorithm flow of the present invention;

[0045] Figure 7 This is a schematic diagram of the Bezier interpolation smoothing path of the present invention. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0047] like Figure 1 As shown in the figure, the intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT, the specific detailed steps are as follows:

[0048] Step 1: Set the flexible boundary of sampling: add sampling boundaries according to the global path, reduce the sampling space of the RRT algorithm, and improve the algorithm solution speed;

[0049] Step 11: Establish a rectangular coordinate system with the center point of the rear axle of the current vehicle as the origin O, the vehicle heading direction as the X-axis, and the axis perpendicular to the driving direction as the Y-axis;

[0050] Step 12: On the map, translate the global path to both sides along the Y axis distance, two flexible boundaries, the range of the global path during the translation process forms a planar closed area and is set as the sampling space, such as Figure 2 As shown in the figure, the global path consists of multiple path point sets. The solid line after translating it to both sides is the flexible boundary. The area within the flexible boundary is used as the sampling space of the fast random search tree. The sampling space is defined as:

[0051]

[0052] in is the global path fitting expression, and They are the x-coordinates of the lowest and highest points of the global path on the x-axis respectively.

[0053] Step 2: Heuristically expand nodes:

[0054] Guided by the global path and path target points, control the location of sampling point generation, guide the feasible path to be close to the global path, and improve the speed of feasible path generation;

[0055] Step 21: Sampling point selection in the global path; setting the target probability when the RRT algorithm performs random sampling , the probability of sampling Find the latest node as the node to be expanded; Figure 2 Selected sampling points shown , the point closest to the node to be expanded in the global path is set as the nearest node, and then the global path point with a distance of Lc (determined by the front overhang, axle length and safety distance) from the nearest node is used as the sampling point for expansion; Random sampling in the feasible area of ​​the sampling space; the goal-oriented and extended nodes are defined as follows:

[0056] ;

[0057] ;

[0058] in is the sampling point, is the global path target point, is a randomly generated point, is the node to be expanded. are the coordinates of the new node after expansion, The coordinates of the node to be expanded, To expand the step size, For each sampling time, A random number;

[0059] Step 22: Design a global path gravity function to inspire and expand sampling points; when random sampling is selected in the feasible area of ​​the sampling space, design a gravity function for the global path to increase the chance of sampling points falling near the global path, improve the degree of simulation between the feasible path and the global path, and expand the global path to the boundaries of the sampling space on both sides at the same time. The step length is translated, k is the number of grids in the sampling space, and the sub-area swept by the i-th translation is recorded as , the gravity function is designed as follows:

[0060] ;

[0061] Select the i-th sub-region The probability of taking a random sample is calculated as follows:

[0062] ;

[0063] Step 23: Set the sampling point reselection strategy: Initialize the detection circle radius , when the sampling point is selected After that, the sampling point is the center of the circle, The detection circle is formed with a radius, and it is determined whether there is a node of the search tree within the detection circle. If so, it is considered that the sampling point The selection will cause repeated search and reselection of sampling points, otherwise the sampling points are considered feasible.

[0064] Step 3, parent node measurement and reselection: design a measurement function for the search tree node based on the principle of being away from obstacles and close to the global path, and select the parent node for the new node with the best performance;

[0065] Step 31: Sample points by distance The nearest node to the sampling point Expand to get new nodes After that, it is a new node The nodes within a certain range are assigned a specified type of metric component, called metric nodes, which allows new nodes to Ability to judge the pros and cons of existing node positions and select the best node for reconnection:

[0066] Cumulative path length: Starting from the measurement node, trace back along the parent node until the distance traveled to the starting point of the search tree. The cumulative path length measurement formula is as follows:

[0067]

[0068] in is the number of search tree nodes passed during the backtracking process, is the length of the i-th reconnection;

[0069] Global path closeness: On the search tree branch where the measurement node is located, the average Euclidean distance between the interval path from the search tree starting point to the measurement node and the global path. The global path closeness measurement formula is defined as follows:

[0070] ;

[0071] in is the coordinate of the i-th node in the interval path from the starting point of the search tree to the metric node, is the coordinate of the point on the global path closest to the i-th node, The total number of search tree nodes on the interval path.

[0072] Obstacle rejection degree: The minimum value of the Euclidean distance between each node and the obstacle on the interval path determined by the measurement node and the starting point. The obstacle rejection degree measurement formula is calculated as follows:

[0073]

[0074] in is the coordinate of the point in the obstacle closest to the i-th node;

[0075] Turning angle: Connect the new node and the measurement node, and the measurement node and its parent node to obtain two connected lines. The angle formed by the two lines can reflect the difficulty of turning the vehicle. The turning angle measurement formula is calculated as follows:

[0076] ;

[0077] in Is a metric node The coordinates of Is the parent node of the metric node The coordinates of Is a new node The coordinates of

[0078] Step 32: The above four metric components are normalized to their maximum values. The maximum value normalization formula is defined as follows:

[0079]

[0080] in represents a certain metric, and Represents the new nodes respectively Within a certain range, the measurement component The minimum and maximum values ​​among all metric nodes;

[0081] Step 33: Set the total metric function as follows, and select the node with the largest total metric value in the search tree as the parent node; Figure 3 Reselect the parent node as shown. In a certain range Three nodes, calculate the total metric value of the three nodes, and select the one with the largest total metric value As the parent node, It is not considered because it is not within the scope;

[0082]

[0083] in are the weight coefficients corresponding to each metric function.

[0084] Step 4: Obstacle collision detection: Establish a simplified physical model of the bus, set a safe distance, and detect whether there is a danger of collision with obstacles during the driving process;

[0085] Step 41: Set the safe distance , establish the bus physical model, such as Figure 4 As shown in the figure, make the minimum circumscribed rectangle of the bus physical model to replace the bus, and at the same time extend the length and width of the circumscribed rectangle respectively. , forming an obstacle detection rectangle, with the midpoint of the rear axle of the bus as the base point, the coordinates of the four vertices A, B, C, and D of the detection rectangle are as follows:

[0086]

[0087] in , , , They are the coordinates of point A, point B, point C, and point D of the detected rectangle. is the coordinate of the center point of the vehicle's rear axle, is the vehicle wheelbase, is the vehicle width, and They are the front and rear overhangs of the vehicle. is the vehicle yaw angle;

[0088] Step 42: Before the new node selects a parent node to try to connect, a detection rectangle is made with the new node and the parent node as the center of the rear axle of the vehicle and the connection direction as the vehicle heading direction, and the two detection rectangles are filled to form an obstacle detection frame containing the connection line. If there is an obstacle in the obstacle detection frame, it is considered that the connection has failed and returns to resampling; otherwise, it is determined whether the distance between the new node and the target point is less than the threshold If yes, the new node is considered as the target point, and a feasible path is obtained by returning to the parent node of the search tree, otherwise the next round of sampling connection is performed.

[0089] Step 5: Expand the feasible path area: construct the path optimization area centered on the feasible path searched by the RRT algorithm;

[0090] Step 51: Expand the starting and ending points of the feasible paths searched by RRT, assuming is the i-th node on the feasible path, and are the starting point and end point of the feasible path respectively. Orientation Direction of movement distance, and then Vertical sideways movement Distance gets two expansion nodes and , and similarly we can get the endpoint of the feasible path Two expansion nodes and ,like Figure 5 Shown in Expand and ; The coordinates of its extended nodes are calculated as follows:

[0091] ;

[0092] ;

[0093] in , They are expansion nodes , The coordinate vector of is the coordinate of the i-th node of the feasible path;

[0094] Step 52: Expand the intermediate nodes of the feasible path searched by RRT except the starting point and the end point, assuming The size is , the node Along The bisector direction extends to both sides Length, get the extended node and ,like Figure 5 Shown in Expand and ; The coordinates of its extended nodes are calculated as follows:

[0095] ;

[0096] in It is an expansion node The coordinate vector of

[0097] Step 53: Figure 5 In the area shown, all the expansion nodes are connected in sequence to form a polygonal expansion area that contains a feasible path.

[0098] Step 6: Optimize the feasible path: Based on factors such as path length and path smoothness, a genetic algorithm is used to solve the feasible path in the optimization area to calculate the optimal bus obstacle avoidance path.

[0099] Step 61: Construct a grid map and divide the bounding rectangle of the expanded area into The grid map of the obstacle area and the grid outside the expansion area is recorded as 0, indicating that the vehicle cannot reach it, and the remaining grids are recorded as 1, indicating the route planning area. Indicates the number of grids in the horizontal direction. Indicates the number of vertical grids; initialization parameters, setting the maximum number of iterations , mutation probability , crossover probability ; Search in raster map A random path from the initial position to the final position constitutes the initial population;

[0100] Step 62: Design a fitness function and arrange the individuals in the population from low to high according to their fitness. The fitness function fit can be calculated by the following formula:

[0101] ;

[0102] Where a and b represent the weight coefficients of path length and path smoothness respectively, smothness represents the smoothness of individual paths, and length represents the length of individual paths:

[0103] ;

[0104] ;

[0105] in and are the vectors pointing from the i-th corner vertex to the adjacent corner vertex in the individual path, represents the number of diagonal nodes in the individual path, Represents the number of adjacent node pairs in an individual path;

[0106] Step 63: Genetic algorithm process is as follows Figure 6 As shown, the population is updated and iterated, and the individual paths in the population are used as parents for crossover and mutation operations. The generated offspring individuals form a new population. The roulette wheel selection method is used to screen the individuals in the new population, and the fitness of the individuals in the new population is recalculated and sorted. It is detected whether the number of iterations has reached If yes, then output the optimized path with the highest fitness, otherwise return to step 63 to continue the loop iteration. Calculated by the following formula:

[0107] ;

[0108] in is the current individual path fitness, is the current number of individuals in the population.

[0109] Step 7: Smooth optimization path: Use the Bessel interpolation method to smoothly optimize the path and obtain the final obstacle avoidance path.

[0110] Step 71: Set the turning points on the optimization path as nodes, and select n nodes with the same interval between every two nodes. Assume that the node set from the starting point to the end point in the optimization path is , path segment The midpoint of ;

[0111] Step 72: For any node except the starting point and the end point ,connect and Forming line segments ,exist Find a little on So that the following is true:

[0112]

[0113] Step 73: Along Direction shift to point With point When they overlap, the new line segment is recorded as ; Set the expansion factor , control point and Translate in the direction of its line segment to obtain a new control point and ,like Figure 7 As shown, To optimize the nodes on the path, is the new control point obtained, where the expansion coefficient The definition is as follows:

[0114]

[0115] Step 74: Set up in order Generate a Bezier curve for the control point to obtain the optimized path smoothing result, where the Bezier curve calculation formula is as follows:

[0116] ;

[0117] Step 75: Use the smoothed path as the obstacle avoidance path; Figure 7 As shown, the upper part is the result of the traditional Bezier smooth curve, and the lower part is the result of the Bezier interpolation smooth curve.

[0118] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. An intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT, characterized in that: The steps include: Step 1. Set the flexible boundary of sampling: add sampling boundary according to the global path; Step 2, heuristic expansion node: guided by the global path and path target point, control the location of sampling point generation, and guide the feasible path to be close to the global path; Step 3, parent node measurement and reselection: Design a measurement function for the search tree node based on the principle of being away from obstacles and close to the global path, and select the parent node for the new node with the best performance; Step 4: Obstacle collision detection: Establish a simplified physical model of the bus, set a safe distance, and detect whether there is a danger of collision with obstacles during the driving process; Step 5: Expand the feasible path area: construct the path optimization area centered on the feasible path searched by the RRT algorithm; Step 6: Optimize the feasible path: Combine the path length and path smoothness, use genetic algorithm to solve the feasible path in the optimization area, and calculate the best bus obstacle avoidance path; Step 7: Smooth optimization path: Use the Bessel interpolation method to smoothly optimize the path and obtain the final obstacle avoidance path.

2. According to claim 1, a method for intelligent bus obstacle avoidance path planning based on genetic algorithm and extended RRT is characterized in that: The step 1 of setting the flexible boundary of the sampling includes the following steps: Step 11: Establish a rectangular coordinate system with the center point of the rear axle of the current vehicle as the origin O, the vehicle heading direction as the X-axis, and the Y-axis perpendicular to the driving direction; Step 12: On the map, translate the global path to both sides along the Y axis Distance, the range passed by the global path during the translation process forms a planar closed area and is set as the sampling space. The sampling space is defined as: ； in is the global path fitting expression, and They are the x-coordinates of the lowest and highest points of the global path on the x-axis respectively.

3. The intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT according to claim 1 is characterized in that: The step 2 of heuristically expanding the node includes the following steps: Step 21: Sampling point selection in the global path; setting the target probability when the RRT algorithm performs random sampling , the probability of sampling Find the latest node as the node to be expanded, set the point closest to the node to be expanded in the global path as the nearest node, and then use the global path point with the nearest node Lc as the sampling point for expansion; Random sampling in the feasible area of ​​the sampling space; Step 22: Design a global path gravity function to inspire and expand sampling points; when random sampling is selected within the feasible region of the sampling space, design a gravity function for the global path; Step 23: Set the sampling point reselection strategy: Initialize the detection circle radius , when the sampling point is selected After that, the sampling point is the center of the circle, The detection circle is formed with a radius, and it is determined whether there is a node of the search tree within the detection circle. If so, it is considered that the sampling point The selection will cause repeated search and reselection of sampling points, otherwise the sampling points are considered feasible.

4. The intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT according to claim 1 is characterized by: In step 3, parent node measurement and reselection, The following steps are included: The nearest node to the sampling point Expand to get new nodes After that, it is a new node The nodes within a certain range are assigned a specified type of metric component, called metric nodes, which allows new nodes to It can judge the pros and cons of the existing node positions and select the best node for reconnection.

5. The intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT according to claim 1 is characterized in that: The obstacle collision detection in step 4 includes the following steps: Step 41: Set the safe distance , establish the bus physical model, make the minimum circumscribed rectangle of the bus physical model to replace the bus, and extend the length and width of the circumscribed rectangle respectively , forming an obstacle detection rectangle, taking the midpoint of the rear axle of the bus as the base point, and detecting the coordinates of the four vertices of the rectangle; Step 42: Before the new node selects a parent node to try to connect, a detection rectangle is made with the new node and the parent node as the center of the rear axle of the vehicle and the connection direction as the vehicle heading direction, and the two detection rectangles are filled to form an obstacle detection frame containing the connection line. If there is an obstacle in the obstacle detection frame, it is considered that the connection has failed and returns to resampling; otherwise, it is determined whether the distance between the new node and the target point is less than the threshold If yes, the new node is considered as the target point, and a feasible path is obtained by returning to the parent node of the search tree, otherwise the next round of sampling connection is performed.

6. The intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT according to claim 1 is characterized by: The step 5 of expanding the feasible path area includes the following steps: Step 51: Expand the starting point and end point of the feasible path searched by RRT, and set is the i-th node on the feasible path, and are the starting point and end point of the feasible path respectively. Orientation Direction of movement distance, and then Vertical sideways movement Distance gets two expansion nodes and , get the feasible path end point Two expansion nodes and ; Step 52: Expand the intermediate nodes of the feasible path searched by RRT except the starting point and the end point, and set The size is , the node Along The bisector direction extends to both sides Length, get the extended node and ; Step 53: Connect all the expansion nodes in sequence to form a polygonal expansion area that includes a feasible path.

7. The intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT according to claim 1 is characterized by: The step 6 of optimizing the feasible path includes the following steps: Step 61: Build a grid map and divide the bounding rectangle of the expanded area into The grid map of the obstacle area and the grid outside the expansion area is recorded as 0, indicating that the vehicle cannot reach it, and the remaining grids are recorded as 1, indicating the route planning area. Indicates the number of grids in the horizontal direction. Indicates the number of vertical grids; initialization parameters, setting the maximum number of iterations , mutation probability , crossover probability ; Search in raster map A random path from the initial position to the final position constitutes the initial population; Step 62: Design a fitness function to arrange the individuals in the population from low to high according to their fitness; Step 63: Update and iterate the population. The individual paths in the population are used as parents to perform crossover and mutation operations. The generated offspring individuals form a new population. The roulette wheel selection method is used to screen the individuals in the new population. The fitness of the individuals in the new population is recalculated and sorted. Check whether the number of iterations has reached If yes, then output the optimal path with the highest fitness, otherwise return to step 63 to continue the loop iteration; where the probability of each individual path being selected in the roulette selection method is Calculated by the following formula: ； in is the current individual path fitness, is the current number of individuals in the population.

8. The intelligent bus obstacle avoidance path planning method based on genetic algorithm and extended RRT according to claim 1 is characterized by: The step 7 smooth optimization path includes the following steps: Step 71: Set the turning points on the optimization path as nodes, and select n nodes with the same interval between every two nodes. Assume that the node set from the starting point to the end point in the optimization path is , path segment The midpoint of ; Step 72: For any node except the starting point and the end point ,connect and Forming line segments ,exist Find a little on So that the following is true: ； Step 73: Add the line segment Along Direction shift to point With point When they overlap, the new line segment is recorded as ; Set the expansion factor , control point and Translate in the direction of its line segment to get a new point and , where the expansion coefficient The definition is as follows: ； Step 74: Set in Order Generate Bezier curves for control points to obtain the optimized path smoothing result; Step 75: Use the smoothed path as the obstacle avoidance path.

Images