Robot path planning method based on improved RRT algorithm

Abstract

The present invention belongs to the field of path planning technology, and specifically relates to a robot path planning method based on an improved RRT algorithm, wherein the steps include obtaining environmental space information that requires path planning, defining parameters, obtaining the minimum distance between coordinate points and obstacles; constructing a bidirectional random tree and initializing it, obtaining sampling points through a sampling function; finding the nearest neighbor point, and selecting a new node to make a safe distance judgment, and resampling if it is not satisfied; generating a discretized path; setting a cost function when a local node is expanded; looping and iterating to generate a discrete point cost, taking the minimum cost value, finding the optimal node and inserting it into the current tree; obtaining an overall path Path after local optimization; discretizing the entire Path, setting the cost function of the node during global planning, selecting the node with the minimum cost for reconnection, and finally obtaining a smooth and effective optimal path. The present invention effectively improves the optimality and smoothness of the path.

Description

Technical Field

[0001] The invention belongs to the technical field of path planning, and in particular relates to a robot path planning method based on an improved RRT algorithm. Background Art

[0002] In today's robotics field, path planning, as a key technology in robot autonomous navigation, mainly solves how to find an optimal or feasible route to the target location from the starting position in a specific scenario. The RRT algorithm is a representative search algorithm. Its basic principle is to start from a tree containing a starting point, and through an iterative process, randomly select a point as a sample in each iteration, find the node closest to the sample point in the tree, and then expand a certain distance in the direction from the node to the sample point to generate a new node. If the new node is not in the obstacle, it is regarded as a feasible node and added to the tree. This process is repeated until the termination condition of the target point is reached. There are three common RRT improvement methods.

[0003] (1) RRT* algorithm. The RRT* algorithm introduces a path optimization mechanism to ensure that the path found is asymptotically optimal. However, the RRT* algorithm requires a large amount of sampling and calculation to find the optimal path, especially in complex environments where computing resources are limited.

[0004] (2) Bidirectional Extended RRT Algorithm. Bidirectional Extended RRT Algorithm is a relatively effective path planning method. It speeds up the path search by simultaneously extending two trees from the starting point and the target point. Compared with the unidirectional RRT algorithm, it can search for paths more efficiently. However, the path generated by this algorithm is not as smooth as the RRT* algorithm, especially where the two trees merge, there may be a large turning angle, and additional processing steps are required to smooth the path.

[0005] (3) Informed RRT algorithm. The core idea of ​​the Informed RRT algorithm is to introduce a heuristic function based on RRT to guide the sampling process towards the target direction. By constructing an elliptical sampling area between the starting point and the target point, invalid sampling can be reduced. However, the performance of this algorithm depends largely on the setting of the ellipse parameters. Inappropriate parameters lead to a decrease in search efficiency compared to the basic RRT algorithm.

[0006] In summary, it can be seen that the existing RRT improvements still have some defects and shortcomings, which affect its application in robot autonomous navigation. Summary of the invention

[0007] In view of the above deficiencies in the prior art, the purpose of the present invention is to provide a robot path planning method based on the RRT improved algorithm, which effectively improves the optimality and smoothness of the path, and also improves the sampling efficiency, enhances the adaptability of the environment and the controllability of the safe path.

[0008] To achieve the above purpose, the technical solution adopted by the present invention is a robot path planning method based on an improved RRT algorithm, comprising the following steps:

[0009] S1. Obtain the information of the environment space map that needs to be planned, including the map size, the shape and position of the obstacles, define the starting point source and the target point goal, set the termination path exploration threshold Threshold, the expansion step size Step_Size, the discrete step size step_size and the node safety distance Constraint;

[0010] S2. Use the bwdist function to obtain the minimum distance between each coordinate point in the map and the obstacle to form a distance field.

[0011] S3. Build and initialize a bidirectional random tree. Add the starting point source and the target point goal as root nodes to the two trees T1 and T2 respectively. The goal is the target point of T1 and the source is the target point of T2. The two trees expand from the root nodes respectively and obtain the sampling points through the sampling function. ;

[0012] S4. Calculation The Euclidean distance to each child node in the current tree is compared to find the nearest neighbor. , and select a new node newPoint to judge the safety distance. If the safety distance is met, execute S5. If not, return to S3 and resample;

[0013] S5. and newPoint, generating The discretized path to newPoint;

[0014] S6. In the stepwise expansion of the path planning, setting the cost function when the local node is expanded;

[0015] S7, based on the cost function when the local node is expanded, generate discrete point costs in a loop iteration, take the minimum cost value, find the optimal node and insert it into the current tree, determine whether to resample based on Threshold, and continue to iterate if resampling is not necessary;

[0016] S8. After multiple cycles of iteration, an overall path Path is finally obtained after local optimization;

[0017] S9. Discretize the entire Path and enter the global planning part. Set the cost function of the node during global planning. Iteratively calculate the cost of each node to each forward node, select the node with the minimum cost for reconnection, and finally obtain a smooth and effective optimization path.

[0018] As a preferred solution of the present invention, in the S1, Threshold is used to determine whether to terminate the path exploration. When the node distance between the two generated trees is less than or equal to the Threshold, it is considered that a feasible path has been found and further exploration is stopped; Step_Size determines the maximum distance between the new node and the nearest node in each iteration; step_size determines the discrete point distance between the discretized nodes in the path in each iteration; Constraint is the minimum limit distance between the node and the obstacle.

[0019] As a preferred solution of the present invention, in S2, the bwdist function is a function used in MATLAB to calculate the distance transform of a binary image, and the distance field distancefield is expressed as:

[0020] (1);

[0021] In the formula, is the minimum distance between the point (x, y) and the obstacle boundary; is the set of boundary points of obstacles, is one of the boundary points.

[0022] As a preferred solution of the present invention, in S3, the process of obtaining the sampling points by the sampling function is:

[0023] S3.1. Randomly select a coordinate point in the map as a sampling point , denoted as ;

[0024] S3.2. Set a random point sampling function, expressed as:

[0025] (2);

[0026] (3);

[0027] (4);

[0028] In the formula, is a random sampling function; represents the target point, which is recorded as ; α, β, γ are three dynamic coefficients that control the sampling point to be close to the target point; , is the transfer function; That is the point Minimum distance to obstacle boundary.

[0029] As a preferred embodiment of the present invention, in the above S4, the method for determining the safe distance is as follows: As a starting point, On the line connecting the end point, select a new node with a length of Step_Size, that is, newPoint, and detect the difference between newPoint and Whether there is an obstacle on the line, and detect Whether the safety distance is met, That is, the minimum distance between newPoint and the obstacle boundary.

[0030] As a preferred solution of the present invention, in said S5, the step of generating a discretized path is:

[0031] S5.1. Calculate the number of discretization steps numSteps and generate the discrete point scale factor waypoints for use in Generate a series of evenly distributed path points between and newPoint and save them to path; the calculation formula for numSteps is:

[0032] (5);

[0033] In the formula, for Euclidean distance to newPoint;

[0034] S5.2, generate The discrete point to newPoint is expressed as:

[0035] (6);

[0036] In the formula, represents the kth discrete point; Indicates the scale factor corresponding to the kth discrete point.

[0037] As a preferred solution of the present invention, in S6, the cost function is expressed as:

[0038] (7);

[0039] In the formula, Represents the cost function corresponding to the expansion of the current tree to the kth discrete point; express and The distance cost; express and The smoothness cost of for The previous node of That is Minimum distance to obstacle boundary; , , is the corresponding weight parameter.

[0040] As a preferred solution of the present invention, in S7, the minimum cost value is expressed as During the iteration process, when the current tree is expanded to the kth discrete point, the corresponding minimum cost value is It is expressed as:

[0041] (8);

[0042] Where k takes values ​​from 1 to numSteps;

[0043] Check whether the minimum distance between the optimal node of the current tree and the node of another tree is less than or equal to Threshold. If the condition is met, continue to iterate, otherwise re-sample.

[0044] As a preferred solution of the present invention, in S9, the cost function of the node during global planning includes a smoothness cost function and a safety cost function, which are respectively expressed as:

[0045] (9);

[0046] (10);

[0047] In the formula, , They represent the smoothness cost function and security cost function from the current node i to the forward node j respectively; , They represent the smoothness cost and security cost from the current node i to the forward node j respectively; , is the corresponding weight parameter;

[0048] The cost function of the node during global planning is obtained by combining the smoothness cost function and the safety cost function, which is expressed as:

[0049] (11);

[0050] In the formula, Represents the cost function from the current node i to the forward node j during global planning.

[0051] As a preferred solution of the present invention, the Pareto decision theory is introduced into the cost function of the node during global planning. Based on the three objectives of distance, smoothness and safety, the Pareto decision theory is used to achieve the optimal balance between different objectives. The Pareto decision is performed by calculating the cost vector from each node to each forward node, and the nodes of the Pareto optimal solution set are selected for reconnection, so as to finally obtain a smooth and effective optimization path.

[0052] The algorithm involved in the present invention can be executed by an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the above algorithm calculation is implemented by executing the software by the processor.

[0053] The beneficial effects of the present invention are:

[0054] The present invention combines the bidirectional RRT algorithm with the optimality principle of discrete dynamic programming, aiming to make up for the shortcomings of the traditional RRT path planning method, and at the same time solve the defects of the current RRT improved algorithm to a certain extent. The DP-RRT path planning method of the present invention consists of two parts: local discrete dynamic programming in the process of generating a path and global discrete dynamic programming after generating a feasible path. By introducing operations such as node cost function, obstacle distance field, and path discretization, it not only effectively improves the optimality and smoothness of the path, but also improves sampling efficiency, enhances environmental adaptability, and controllability of safe paths.

[0055] The discrete dynamic programming adopted by the present invention can not only optimize the local path during the expansion process, but also define the cost function more flexibly, provide more accurate heuristic information, and guide the expansion direction of RRT; introduce Pareto decision to comprehensively consider multiple goals, avoiding the limitation of optimizing only a single goal in traditional path planning. Compared with the traditional RRT improved algorithm based on geometry guidance, dynamic programming provides more flexibility and global optimization capabilities. The idea of ​​the DP-RRT path planning method of the present invention can also be extended to three-dimensional environment planning to better cope with complex and changing environmental challenges. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the process of the present invention;

[0057] Figure 2 It is the traditional RRT algorithm path planning trajectory in the verification process of the present invention Figure 1 ;

[0058] Figure 3It is the traditional RRT algorithm path planning trajectory in the verification process of the present invention Figure 2 ;

[0059] Figure 4 It is the traditional RRT algorithm path planning trajectory in the verification process of the present invention Figure 3 ;

[0060] Figure 5 It is the traditional RRT algorithm path planning trajectory in the verification process of the present invention Figure 4 ;

[0061] Figure 6 It is the improved RRT algorithm path planning trajectory in the verification process of the present invention Figure 1 ;

[0062] Figure 7 It is the improved RRT algorithm path planning trajectory in the verification process of the present invention Figure 2 ;

[0063] Figure 8 It is the improved RRT algorithm path planning trajectory in the verification process of the present invention Figure 3 ;

[0064] Fig. 9 It is the improved RRT algorithm path planning trajectory in the verification process of the present invention Figure 4 . DETAILED DESCRIPTION

[0065] The embodiments of the present invention are further described below in conjunction with the accompanying drawings:

[0066] like Figure 1 As shown, the robot path planning method based on the RRT improved algorithm includes the following steps:

[0067] S1. Obtain the information of the environment space map that needs to be planned, including the map size, the shape and position of the obstacles, define the starting point source and the target point goal, set the termination path exploration threshold Threshold, the expansion step size Step_Size, the discrete step size step_size and the node safety distance Constraint;

[0068] S2. Use the bwdist function to obtain the minimum distance between each coordinate point in the map and the obstacle to form a distance field.

[0069] S3. Build and initialize a bidirectional random tree. Add the starting point source and the target point goal as root nodes to the two trees T1 and T2 respectively. The goal is the target point of T1 and the source is the target point of T2. The two trees expand from the root nodes respectively (i.e., T1 is planned from source to goal, and T2 is planned from goal to source). The sampling points are obtained through the sampling function. ;

[0070] S4. Calculation The Euclidean distance to each child node in the current tree is compared to find the nearest neighbor. , and select a new node newPoint to judge the safety distance. If the safety distance is met, execute S5. If not, return to S3 and resample;

[0071] S5. and newPoint, generating The discretized path to newPoint;

[0072] S6. In the stepwise expansion of the path planning, setting the cost function when the local node is expanded;

[0073] S7, based on the cost function when the local node is expanded, generate discrete point costs in a loop iteration, take the minimum cost value, find the optimal node and insert it into the current tree, determine whether to resample based on Threshold, and continue to iterate if resampling is not necessary;

[0074] S8. After multiple cycles of iteration, an overall path Path is finally obtained after local optimization;

[0075] S9. Discretize the entire Path and enter the global planning part. Set the cost function of the node during global planning. Iteratively calculate the cost of each node to each forward node, select the node with the minimum cost for reconnection, and finally obtain a smooth and effective optimization path.

[0076] In S1, Threshold is used to determine whether to terminate the path exploration. When the node distance between the two generated trees is less than or equal to Threshold, it is considered that a feasible path has been found and further exploration is stopped; Step_Size determines the maximum distance between the new node and the nearest node in each iteration; step_size determines the discrete point distance between the discretized nodes in the path in each iteration; Constraint is the minimum limit distance between the node and the obstacle.

[0077] In S2, the bwdist function is a function used in MATLAB to calculate the distance transform of a binary image, and the distance field distancefield is expressed as:

[0078] (1);

[0079] In the formula, is the minimum distance between the point (x, y) and the obstacle boundary; is the set of boundary points of obstacles, is one of the boundary points.

[0080] In S3, the process of obtaining sampling points through the sampling function is:

[0081] S3.1. Randomly select a coordinate point in the map as a sampling point , denoted as ;

[0082] S3.2. Set a random point sampling function, expressed as:

[0083] (2);

[0084] (3);

[0085] (4);

[0086] In the formula, is a random sampling function; represents the target point, which is recorded as ; α, β, γ are three dynamic coefficients that control the sampling point to be close to the target point; , is the transfer function; That is the point Minimum distance to obstacle boundary.

[0087] In S4, the method for determining the safety distance is as follows: As a starting point, On the line connecting the end point, select a new node with a length of Step_Size, that is, newPoint, and detect whether newPoint is Whether there is an obstacle on the line, and detect Whether the safety distance is met, That is, the minimum distance between newPoint and the obstacle boundary.

[0088] In S5, the steps for generating the discretized path are:

[0089] S5.1. Calculate the number of discretization steps numSteps and generate the discrete point scale factor waypoints for use in Generate a series of evenly distributed path points between and newPoint and save them to path; the calculation formula for numSteps is:

[0090] (5);

[0091] In the formula, for Euclidean distance to newPoint;

[0092] S5.2, generate The discrete point to newPoint is expressed as:

[0093] (6);

[0094] In the formula, represents the kth discrete point; Indicates the scale factor corresponding to the kth discrete point.

[0095] Regarding the generation of discrete point scaling factors waypoints, for example, if numSteps is 5, then linspace(0, 1, numSteps) will generate a vector [0, 0.25, 0.5, 0.75, 1]. These values ​​can be used as weights or scaling factors for interpolation or generating path points between two points. The function of the linspace function is to generate a series of values ​​evenly distributed between the two specified endpoints.

[0096] In S6, the cost function is expressed as:

[0097] (7);

[0098] In the formula, Represents the cost function corresponding to the expansion of the current tree to the kth discrete point; express and The distance cost; express and The smoothness cost of for The previous node of That is Minimum distance to obstacle boundary; , , are the corresponding weight parameters, which are , , The weight parameters can be set according to the requirements. For example, you can set =0.5, =9.8, =0.1.

[0099] In S7, the minimum cost is recorded as During the iteration process, when the current tree is expanded to the kth discrete point, the corresponding minimum cost value is It is expressed as:

[0100] (8);

[0101] Where k takes values ​​from 1 to numSteps;

[0102] Check whether the minimum distance between the optimal node of the current tree and the node of another tree is less than or equal to Threshold. If the condition is met, continue to iterate, otherwise re-sample.

[0103] In S9, the cost function of the node during global planning includes the smoothness cost function and the safety cost function, which are expressed as:

[0104] (9);

[0105] (10);

[0106] In the formula, , They represent the smoothness cost function and security cost function from the current node i to the forward node j respectively; , They represent the smoothness cost and security cost from the current node i to the forward node j respectively; , are the corresponding weight parameters, which are , The weight parameters can be set according to the requirements. For example, you can set =0.01, =1.8.

[0107] The cost function of the node during global planning is obtained by combining the smoothness cost function and the safety cost function, which is expressed as:

[0108] (11);

[0109] In the formula, Represents the cost function from the current node i to the forward node j during global planning.

[0110] Distance cost refers to the total length of the path from the starting point to the target point. In path planning, distance cost is usually used to minimize the total length of the path, thereby reducing travel time or energy consumption. Smoothness cost focuses on the smoothness of the path, that is, the smoothness of the path turns. In robot path planning, a smooth path can reduce mechanical wear and improve driving comfort and safety. Safety cost involves safety considerations in path planning, such as avoiding collisions and maintaining a safe distance. Safety is crucial in path planning for robots or autonomous vehicles.

[0111] In global planning, the Pareto decision theory is introduced into the cost function of the node. Based on the three goals of distance, smoothness and safety, the Pareto decision theory (Pareto optimality) is used to achieve the best balance between different goals. The Pareto decision is made by calculating the cost vector from each node to each node in the forward direction, and the nodes in the Pareto optimal solution set are selected for reconnection, and finally a smooth and effective optimization path is obtained. For example, define the vector , Dominant vector , v is a vector containing three elements, each element corresponds to the cost of a target, w is another vector containing three elements, which is used to compare with vector v to determine whether v is dominated by w; if and only if and When, it can be expressed as v <w。

[0112] Among them, the vectors in v and w are given by , , express. means that for all t, it belongs to the set {1, 2, 3}, which means that for every element in v and w (i.e., for the three objectives: distance, smoothness, and safety), the tth element in v is not greater than the tth element in w; It means that there is at least one l belonging to the set {1, 2, 3}, which means that among the three elements of v and w, there is at least one element (that is, at least on one target) where the lth element of v is strictly less than the lth element of w.

[0113] When the Pareto optimal solution is introduced to optimize the path generated by RRT, multiple objectives can be considered simultaneously and a balance can be found between these objectives, avoiding the defects that may be caused by single-objective optimization.

[0114] The verification process is:

[0115] Based on four environment space maps, the path planning effects of the improved RRT (DP-RRT) of this embodiment and the traditional RRT are compared. Figure 2-Figure 9 shown. Figure 2-Figure 5The green route in the figure is the path planned by the traditional RRT method. Figure 6-Figure 9 The light blue line in the figure is the overall path obtained by the improved RRT method based on S8 after local optimization, and the black dotted line is the smooth and effective optimization path finally obtained by the improved RRT method. Figure 2-Figure 9 The remaining branches in are exploration paths, with the starting point source at the upper left and the target point goal at the lower right. As can be seen from the figure, the path generated by the traditional RRT algorithm is relatively rough. The improved RRT method can generate smoother and shorter paths, reducing unnecessary turns or path lengths. It is more uniform when exploring space, effectively avoiding blind spots or inefficient exploration of space, especially in complex environments. At the same time, it avoids the distance between the path and the obstacle being too close, enhancing the adaptability of the environment and the controllability of the safe path. The improved RRT algorithm can flexibly adjust the algorithm parameters and structure when facing different application scenarios and complexities.

[0116] Table 1 Path parameters

[0117]

[0118] The path parameters are shown in Table 1. By analyzing the performance data of the traditional RRT method and the DP-RRT planning algorithm in path planning, the path optimized by the DP-RRT planning algorithm is compared with the traditional RRT algorithm. The average path distance is shortened by 30%, redundant points are reduced, the sampling efficiency is greatly improved, and the average path curvature is reduced by 59%, which effectively improves the optimality and smoothness of the path. In summary, the DP-RRT (RRT improved algorithm) path planning method proposed in this invention provides a new solution for robot path planning.

Claims

1. The robot path planning method based on the improved RRT algorithm is characterized by The following steps are involved: S1. Obtain the information of the environment space map that needs to be planned, including the map size, the shape and position of the obstacles, define the starting point source and the target point goal, set the termination path exploration threshold Threshold, the expansion step size Step_Size, the discrete step size step_size and the node safety distance Constraint; S2. Use the bwdist function to obtain the minimum distance between each coordinate point in the map and the obstacle to form a distance field. S3. Build and initialize a bidirectional random tree. Add the starting point source and the target point goal as root nodes to the two trees T1 and T2 respectively. The goal is the target point of T1 and the source is the target point of T2. The two trees expand from the root nodes respectively and obtain the sampling points through the sampling function. ; S4. Calculation The Euclidean distance to each child node in the current tree is compared to find the nearest neighbor. , and select a new node newPoint to judge the safety distance. If the safety distance is met, execute S5. If not, return to S3 and resample; S5. and newPoint, generating The discretized path to newPoint; S6. In the stepwise expansion of the path planning, setting the cost function when the local node is expanded; S7, based on the cost function when the local node is expanded, generate discrete point costs in a loop iteration, take the minimum cost value, find the optimal node and insert it into the current tree, determine whether to resample based on Threshold, and continue to iterate if resampling is not necessary; S8. After multiple cycles of iteration, an overall path Path is finally obtained after local optimization; S9. Discretize the entire Path and enter the global planning part. Set the cost function of the node during global planning. Iteratively calculate the cost of each node to each forward node, select the node with the minimum cost for reconnection, and finally obtain a smooth and effective optimization path.

2. The robot path planning method based on the RRT improved algorithm according to claim 1, characterized in that: In S1, Threshold is used to determine whether to terminate the path exploration. When the node distance between the two generated trees is less than or equal to Threshold, it is considered that a feasible path has been found and further exploration is stopped; Step_Size determines the maximum distance between the new node and the nearest node in each iteration; step_size determines the discrete point distance between the discretized nodes in the path in each iteration; Constraint is the minimum limit distance between the node and the obstacle.

3. The robot path planning method based on the RRT improved algorithm according to claim 1, characterized in that: In S2, the bwdist function is a function used in MATLAB to calculate the distance transform of a binary image, and the distance field distancefield is expressed as: （1）； In the formula, is the minimum distance between the point (x, y) and the obstacle boundary; is the set of boundary points of obstacles, is one of the boundary points.

4. The robot path planning method based on the RRT improved algorithm according to claim 3 is characterized in that: In the above S3, the process of obtaining the sampling points through the sampling function is: S3.

1. Randomly select a coordinate point in the map as a sampling point , denoted as ; S3.

2. Set a random point sampling function, expressed as: （2）； （3）； （4）； In the formula, is a random sampling function; represents the target point, which is recorded as ; α, β, γ are three dynamic coefficients that control the sampling point to be close to the target point; , is the transfer function; That is the point Minimum distance to obstacle boundary.

5. The robot path planning method based on the RRT improved algorithm according to claim 4 is characterized in that: In the above S4, the method for determining the safety distance is as follows: As a starting point, On the line connecting the end point, select a new node with a length of Step_Size, that is, newPoint, and detect the difference between newPoint and Whether there is an obstacle on the line, and detect Whether the safety distance is met, That is, the minimum distance between newPoint and the obstacle boundary.

6. The robot path planning method based on the RRT improved algorithm according to claim 5, characterized in that: In the above S5, the step of generating a discretized path is: S5.

1. Calculate the number of discretization steps numSteps and generate the discrete point scale factor waypoints for use in Generate a series of evenly distributed path points between and newPoint and save them to path; the calculation formula for numSteps is: （5）； In the formula, for Euclidean distance to newPoint; S5.2, generate The discrete point to newPoint is expressed as: （6）； In the formula, represents the kth discrete point; Indicates the scale factor corresponding to the kth discrete point, where k takes a value from 1 to numSteps.

7. The robot path planning method based on the RRT improved algorithm according to claim 6, characterized in that: In S6, the cost function is expressed as: （7）； In the formula, Represents the cost function corresponding to the expansion of the current tree to the kth discrete point; express and The distance cost; express and The smoothness cost of for The previous node of That is Minimum distance to obstacle boundary; , , is the corresponding weight parameter.

8. The robot path planning method based on the RRT improved algorithm according to claim 7, characterized in that: In S7, the minimum cost is recorded as During the iteration process, when the current tree is expanded to the kth discrete point, the corresponding minimum cost value is It is expressed as: （8）； Check whether the minimum distance between the optimal node of the current tree and the node of another tree is less than or equal to Threshold. If the condition is met, continue to iterate, otherwise re-sample.

9. The robot path planning method based on the RRT improved algorithm according to claim 8, characterized in that: In S9, the cost function of the node during global planning includes a smoothness cost function and a safety cost function, which are respectively expressed as: （9）； （10）； In the formula, , They represent the smoothness cost function and security cost function from the current node i to the forward node j respectively; , They represent the smoothness cost and security cost from the current node i to the forward node j respectively; , is the corresponding weight parameter; The cost function of the node during global planning is obtained by combining the smoothness cost function and the safety cost function, which is expressed as: （11）； In the formula, Represents the cost function from the current node i to the forward node j during global planning.

10. The robot path planning method based on the RRT improved algorithm according to claim 9, characterized in that: The Pareto decision theory is introduced into the cost function of the node during global planning. Based on the three goals of distance, smoothness and safety, the Pareto decision theory is used to achieve the best balance between different goals. The Pareto decision is made by calculating the cost vector from each node to each forward node, and the nodes of the Pareto optimal solution set are selected for reconnection, finally obtaining a smooth and effective optimization path.

Images