Robot path planning method based on improved RRT algorithm

Abstract

The invention discloses a robot path planning method based on an improved RRT algorithm, and the method comprises the steps: limiting a sampling region through a maximum sampling radius and a newly-added node at a previous moment, and enabling an obtained sampling point to be deviated to a target point with a large probability; meanwhile, by limiting the sampling points, the method can be close to the target point to the maximum extent under the condition that the number of obstacles is small, and can be close to the target point to the minimum extent under the condition that the number of obstacles is large, so that the sampling frequency of the sampling points when the newly added nodes of the random tree are obtained is reduced, and the path generation efficiency is improved. In addition, redundant nodes in the path are removed after the path is generated, and pruning operation is performed on the nodes in the path, so that optimization of the path nodes is realized, the number of corners of the generated path is reduced, the path planning performance is improved, and compared with RRTstar and Astar algorithms, the path length can be effectively shortened, and the search efficiency is improved.

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 recent years, mobile robots have been widely used in military operations, emergency rescue, warehouse management, logistics and transportation. Path planning is an indispensable and important part of mobile robots. It usually refers to finding a collision-free and optimal path from the starting point to the end point. At present, path planning algorithms can be roughly divided into three categories: search-based algorithms, biological heuristic algorithms and sampling-based algorithms. Search-based path planning algorithms mainly include A* algorithm, Dijkstra algorithm and Theta* algorithm. This type of algorithm divides the map into grids and then uses search algorithms to obtain the path from the starting point to the end point. Although such methods are simple to calculate, the time they take to find the path and the length of the path depend on the fineness of the grid division. Biologically inspired path planning algorithms mainly include ant colony algorithm (ACA), genetic algorithm (GA) and particle swarm optimization (PSO) algorithm. This type of algorithm has strong adaptability and parallelism, but the calculation time required is relatively long and it is easy to fall into the local optimal solution. The sampling-based path planning algorithm obtains collision-free paths from the starting point to the end point by randomly obtaining sampling points on the map and then connecting these points. It mainly includes the rapidly expanding random tree (RRT) algorithm, the rapidly expanding random tree star (RRT*) algorithm, and the probabilistic roadmap (PRM) algorithm.

[0003] The randomness and blindness of the RRT algorithm are too high. The RRT* algorithm reduces the path cost by finding parent nodes and rewiring, but increases the search time and is prone to falling into local optimal solutions. Summary of the invention

[0004] In view of the fact that the randomness and blindness of the above-mentioned existing RRT algorithm are too high, and the RRT* algorithm cannot solve the problems of path length and search time at the same time, the purpose of the present invention is to provide an improved fast-expanding random tree path planning algorithm suitable for complex environments with high-dimensional spaces and non-convex obstacles.

[0005] A robot path planning method based on an improved RRT algorithm proposed in the present invention comprises the following steps:

[0006] Step 1: Get the environment information around the robot and build a coordinate system. The environment information includes the initial point Q start , obstacles and target point Q goal The location information of the initial point Q start Add to the random tree as the root node of the random tree.

[0007] Step 2: Get the sampling point Q in the constructed coordinate system sample ; Sampling point Q sample Azimuth and sampling radius r limit The expression is:

[0008]

[0009] r limit =a 3 r max +a 4 r new +a 5 eps

[0010] Among them, a 1 is the goal-oriented coefficient, a 1 >k max -k min θ initial The initial point Q start and the target point Q goal The angle between the line connecting the two and the x-axis; k is a random number with an expected value of zero, k∈[k min , k max ]; k min and k max are the minimum and maximum values ​​of the preset random number k; a 2 is the radian global coefficient; a 3 is the expansion coefficient; r max is the maximum sampling radius; a 4 is the global coefficient of step length; r new The last newly added node Q′ new With the initial point Q start The Euclidean distance of 5 is the random coefficient of the step size related to the sampled collision coefficient col, a 5 =rand-col; rand is a random number in the preset range; col is the sampling collision coefficient; Q is the node on the random tree; esp is the maximum growth step length, and its expression is:

[0011]

[0012] Among them, s is the preset minimum growth step size; x step is the step size coefficient.

[0013] If there is only a root node on the random tree, the root node is used as the last newly added node Q′ new According to the sampling point Q sample Get a random point Q rand .

[0014] Step 3: Traverse all nodes on the random tree and calculate the relationship between each node and the random point Q rand Euclidean distance, get the distance random point Q rand The nearest node is used as the feature node Q near . According to the random point Q rand and feature node Q near Get the newly added node Q new .

[0015] Step 4: If a new node Q is added new With characteristic node Q near If the line collides with an obstacle, repeat steps 2 and 3 until the newly added node Q is obtained. new With characteristic node Q near There are no obstacles on the connection line, and the newly added node Q new Add to the random tree. Update the step coefficient x according to the number of collisions num step and the sampled collision coefficient col.

[0016] Step 5: Update the newly added node Q new The parent node Q parent .

[0017] Step 6: Traverse the existing nodes on the random tree and calculate the relationship between each node and the target point Q goal The Euclidean distance between them; if there is a node and the target point Q goal If the Euclidean distance between them is less than the minimum growth step s, then this node is used as the target point Q goal The parent node, and from the target point Q goal Start selecting the parent node Q that connects each node in turn parent Until the initial point Q start , as the initial point Q start Move to the target point Q goal If the path is not found, the iteration process ends; otherwise, steps 2 to 5 are repeated to continue expanding the random tree.

[0018] Preferably, in step 2, if the robot can move in three-dimensional space, a sampling point Q is also constructed. sample The polar angle θ of is expressed as:

[0019] θ=b 1 (θ initial +k)+2b 2 π

[0020] Among them, b 1 is the goal-oriented coefficient, b 1 >k max -k min ; b 2 is the radian global coefficient.

[0021] As a preference, in step 5, after obtaining the initial point Q start Move to the target point Q goal After the path is obtained, the redundant nodes in the path are removed. The method for removing redundant nodes is as follows:

[0022] Take the target point Q goal As the end point, the initial point Q start As the starting point, connect the end point and the starting point; if the line connecting the starting point and the end point collides with an obstacle, the child node Q of the starting point is used. child Update the starting point until the line connecting the starting point and the end point does not collide with any obstacles. Use the updated starting point as the end point, and the initial point Q start As the starting point, repeat the above process until the end point is equal to the initial point Q start The connecting line does not collide with the obstacle.

[0023] As a preference, in step 5, after removing redundant nodes, each node in the path is selected in turn as the mobile node Q choose ; First move the mobile node Q choose Move towards its parent node, with a translation vector d for each move 1 for:

[0024]

[0025] Until the mobile node Q choose The line connecting its child nodes collides with an obstacle or moves the node Q choose After reaching its parent node, stop moving towards its parent node; then move the mobile node Q choose Move to its child node, with the translation vector d for each move 2 for:

[0026]

[0027] Until the mobile node Q choose The line connecting to its parent node collides with an obstacle or moves the node Q choose After reaching its child node, stop moving towards its child node. After all nodes in the path are moved, get the updated path as the initial point Q start Move to the target point Q goal The final path.

[0028] Preferably, in the step 2, if there is no potential field U in the process of obtaining the sampling point req , with sampling point Q sample As a random point Q rand ; On the contrary, random point Q rand The expression is:

[0029]

[0030] Among them, Q req is the center point of the potential field; F req is the sampling point Q sample The repulsive force received.

[0031] Preferably, the sampling point Q sample The repulsive force F req The expression is:

[0032]

[0033] Among them, k req is the potential field gain coefficient; r is the potential field action radius; n is the adjustment factor coefficient.

[0034] As a preference, the potential field U req The confirmation method is as follows:

[0035] If the neighboring node set Q neighbor If the number of elements in [] exceeds the preset threshold, a new node Q is added. new As the center point of the potential field Q req , generating only affects the next random point Q rand The potential field U req , and reassign the maximum sampling radius r max and distance r new , the maximum sampling radius r′ after assignment max and distance r′ new The expression is:

[0036] r′ max =r max -1.1r

[0037] r′ new =r new -1.1r.

[0038] As a preferred embodiment, in step 5, the newly added node Q is updated new The parent node Q parent The method is as follows:

[0039] Add a new node Q with the center of the circle as the center new , all nodes within the circle with a radius of potential field action radius r are regarded as the neighboring node set Q neighbor []. Get the neighboring node set Q neighbor [] From each node to the newly added node Q new The total cost of cost:

[0040] cost=cost neighbor +dist(Qneighbor -Q new )

[0041] Among them, Q neighbor is the set of neighboring nodes Q neighbor Neighboring nodes in []; cost neighbor The initial point Q start Arrival at neighboring node Q neighbor The cost.

[0042] Take the neighboring node set Q neighbor [] From each node to the newly added node Q new The node corresponding to the minimum cost is the minimum cost node Q min ; If the minimum cost node Q min With the newly added node Q new If there is no collision between nodes, the node Q with the minimum cost min As a new node Q new Parent node.

[0043] As a preferred embodiment, in the step 3, the newly added node Q is obtained. new The method is as follows:

[0044]

[0045] Among them, dist(Q rand ,Q near ) is a random point Q rand With characteristic node Q near The Euclidean distance between .

[0046] Preferably, in step 4, the step length coefficient x step The update method is: in the original step coefficient x step Add a small step size Δx to the step , small step length Δx step The update method of the sampling collision coefficient col is as follows: add a small coefficient Δcol to the original sampling collision coefficient col, the initial value of the small coefficient Δcol is 0, and after two consecutive collisions, the small coefficient Δcol increases by a fixed value each time the number of collisions increases until the small coefficient Δcol reaches the preset maximum value.

[0047] The present invention has the following beneficial effects:

[0048] 1. The present invention limits the sampling area by the maximum sampling radius and the newly added nodes at the previous moment, so that the acquired sampling points can be biased towards the target point with a greater probability; at the same time, by limiting the sampling points, the present invention can be close to the target point to the greatest extent when there are fewer obstacles, and can be close to the target point to a lesser extent when there are more obstacles, thereby reducing the number of sampling points when acquiring new nodes in the random tree and improving the efficiency of path generation.

[0049] 2. After the path is generated, the present invention removes redundant nodes in the path and performs pruning operations on the nodes in the path, thereby optimizing the path nodes, reducing the number of corners in the generated path, and improving the path planning performance. Compared with the RRTstar and Astar algorithms, the present invention can effectively shorten the path length and improve the search efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is the overall flow chart of the present invention.

[0051] Figure 2 Schematic diagram of the azimuth angle of random points in the present invention.

[0052] Figure 3 Schematic diagram of the sampling radius of random points in the present invention.

[0053] Figure 4 Schematic diagram of random tree growth in the present invention.

[0054] Figure 5 This is a schematic diagram of reselecting a parent node in the present invention.

[0055] Figure 6 This is a schematic diagram of generating the escape potential field in the present invention.

[0056] Figure 7 It is a schematic diagram of removing redundant nodes in the present invention.

[0057] Figure 8 It is a schematic diagram of pruning in the present invention.

[0058] Fig. 9 This is a simulation diagram of the present invention and the RRT* algorithm performing path planning in a complex map.

[0059] Fig.10 This is a simulation diagram of the present invention and the RRT* algorithm performing path planning in a single-pass narrow road map. DETAILED DESCRIPTION

[0060] The present invention will be further described below in conjunction with the accompanying drawings.

[0061] like Figure 1 As shown, a robot path planning method based on the improved RRT algorithm includes the following steps:

[0062] Step 1: Obtain the environment information around the robot and build a coordinate system; if the robot moves in a two-dimensional plane, establish a plane coordinate system; if the robot moves in a three-dimensional space, establish a spatial rectangular coordinate system. The environment information includes the initial point Q start , obstacles and target point Q goal The location information of the initial point Q start Add to the random tree as the root node of the random tree.

[0063] Step 2: If Figure 2 , 3 As shown in 4, the sampling point Q is obtained in the constructed coordinate system sample ; If the robot moves in a two-dimensional plane, the sampling point Q sample Azimuth (the angle formed with the x-axis) and the sampling radius r limit The expression is:

[0064]

[0065] r limit =a 3 r max +a 4 r new +a 5 eps

[0066] Among them, a 1 is the goal-oriented coefficient, a 1 >k max -k min θ initial The initial point Q start and the target point Q goal The angle between the line connecting the two and the x-axis; k is a random number with an expected value of zero, k∈[k min , k max ]; k min and k max are the minimum and maximum values ​​of the preset random number k; a 2 is the radian global coefficient; a 3 is the expansion coefficient; r max is the maximum sampling radius, r max =max(dist(Q,Q start ));a 4 is the global coefficient of step length; r new The last newly added node Q′ new With the initial point Q start The Euclidean distance of 5 is the random coefficient of the step size related to the sampled collision coefficient col, a 5=rand-col; rand is a random number in the preset range; col is the sampling collision coefficient; Q is a node on the random tree; max(·) is the maximum value; dist(Q,Q start ) is the random tree node to the initial point Q start Euclidean distance; esp is the maximum growth step, and its expression is:

[0067]

[0068] Among them, s is the preset minimum growth step size; x step is the step size coefficient, the initial value is 0.

[0069] If the robot moves in three-dimensional space, add sampling points Q based on the two-dimensional space sample The polar angle θ (the angle formed with the z-axis) is expressed as:

[0070] θ=b 1 (θ initial +k)+2b 2 π

[0071] Among them, b 1 is the goal-oriented coefficient, b 1 >k max -k min ; b 2 is the radian global coefficient.

[0072] In this embodiment, the target orientation coefficient a 1 and b 1 The value of is 0.3; the minimum and maximum values ​​of the random number k are -0.1 and 0.1 respectively; the radian global coefficient a 2 and b 2 The value of is 0.0583rand; the random number rand takes values ​​in [0,3].

[0073] If there is only a root node on the random tree, the root node is used as the last newly added node Q′ new .

[0074] Step 3: According to the sampling point Q sample Get a random point Q rand ; If there is no potential field U in the process of obtaining sampling points req , with sampling point Q sample As a random point Q rand ; On the contrary, random point Q rand The expression is:

[0075]

[0076] Among them, Q reqis the center point of the potential field; F req is the sampling point Q sample The repulsive force is expressed as:

[0077]

[0078] Among them, ▽U req is the repulsive potential field U rep The gradient of k req is the potential field gain coefficient, its value is 5; r is the potential field action radius, r = 3×esp.

[0079] Potential field U req The expression is:

[0080]

[0081] Among them, n is the adjustment factor coefficient, and its value is 2.

[0082] Step 4: If Figure 4 As shown, traverse all nodes on the random tree and calculate the distance between each node and the random point Q rand Euclidean distance, get the distance random point Q rand The nearest node is used as the feature node Q near . According to the random point Q rand and feature node Q near Get the newly added node Q new , add node Q new The expression is:

[0083]

[0084] Among them, dist(Q rand ,Q near ) is a random point Q rand With characteristic node Q near The Euclidean distance between .

[0085] Step 5: If a new node Q is added new With characteristic node Q near If the line collides with an obstacle, repeat steps 2 and 3 until the newly added node Q is obtained. new With characteristic node Q near There are no obstacles on the connection line, and the newly added node Q new Add to the random tree. To obtain the newly added node Q added to the random tree new During the process, the number of random points generated is taken as the number of collisions num, and the step coefficient x is updated according to the number of collisions num step and the sampled collision coefficient col.

[0086] Step size factor x stepThe update method is: in the original step coefficient x step Add a small step size Δx to the step , small step length Δx step It decreases with the increase of the number of collisions num. The updating method of the sampling collision coefficient col is: adding a small coefficient Δcol to the original sampling collision coefficient col, the initial value of the small coefficient Δcol is 0, after two consecutive collisions, each time the number of collisions increases, Δcol+0.5, until the small coefficient Δcol reaches the preset maximum value 2.

[0087] Step 6: If Figure 5 As shown, the center of the circle is the newly added node Q new , all nodes within the circle with a radius of potential field action radius r are regarded as the neighboring node set Q neighbor []. Get the neighboring node set Q neighbor [] From each node to the newly added node Q new The total cost of cost:

[0088] cost=cost neighbor +dist(Q neighbor -Q new )

[0089] Among them, Q neighbor is the set of neighboring nodes Q neighbor Neighboring nodes in []; cost neighbor The initial point Q start Arrival at neighboring node Q neighbor The cost.

[0090] Take the neighboring node set Q neighbor [] From each node to the newly added node Q new The node corresponding to the minimum cost is the minimum cost node Q min ; If the minimum cost node Q min With the newly added node Q new If there is no collision between nodes, the node Q with the minimum cost min As a new node Q new Parent node. Figure 5 As shown, the total cost from point a to point c is less than the total cost from point b to point c, so point a is reselected as the parent node of point c.

[0091] Step 7: If Figure 6 As shown, in order to escape the local optimal solution, if the neighboring node set Q neighbor If the number of elements in [] exceeds the preset threshold, a new node Q is added. new As the center point of the potential field Q req , generating only affects the next random point Q rand The potential field U req, and reassign the maximum sampling radius r max and distance r new , the maximum sampling radius r′ after assignment max and distance r′ new The expression is:

[0092] r′ max =r max -1.1r

[0093] r′ new =r new -1.1r

[0094] The maximum sampling radius r' after assignment max and distance r′ new The maximum sampling radius r of the next sampling node max and distance r new .

[0095] Step 8: Traverse the existing nodes on the random tree and calculate the relationship between each node and the target point Q goal The Euclidean distance between them; if there is a node and the target point Q goal If the Euclidean distance between them is less than the minimum growth step s, then this node is used as the target point Q goal The parent node, and from the target point Q goal Start selecting the parent node Q that connects each node in turn parent Until the initial point Q start , as the initial point Q start Move to the target point Q goal If the path is not found, the iteration process ends; otherwise, steps 2 to 6 are repeated to continue expanding the random tree.

[0096] Step 9: If Figure 7 , 8 As shown in Figure 1, the path is clipped, redundant nodes are removed, and the random tree is pruned. goal As the end point, the initial point Q start As the starting point, connect the end point and the starting point; if the line connecting the starting point and the end point collides with an obstacle, the child node Q of the starting point is used. child Update the starting point until the line connecting the starting point and the end point does not collide with any obstacles. Use the updated starting point as the end point, and the initial point Q start As the starting point, repeat the above process until the end point is equal to the initial point Q start The connecting line does not collide with the obstacle.

[0097] Select each node in the path as the mobile node Q choose ; First move the mobile node Q choose Move towards its parent node, with a translation vector d for each move 1for:

[0098]

[0099] Until the mobile node Q choose The line connecting its child nodes collides with an obstacle or moves the node Q choose After reaching its parent node, stop moving towards its parent node; then move the mobile node Q choose Move to its child node, with the translation vector d for each move 2 for:

[0100]

[0101] Until the mobile node Q choose The line connecting to its parent node collides with an obstacle or moves the node Q choose After reaching its child node, stop moving towards its child node. After all nodes in the path are moved, get the updated path as the initial point Q start Move to the target point Q goal The final path.

[0102] Step 10: Fig. 9 , Fig.10 As shown, the present invention and the RRT* algorithm are used to plan paths for complex maps and single-pass narrow road maps, respectively. It can be seen from the figure that the path finally generated by the present invention has fewer corners and smaller corner curvature than the path generated by the RRT* algorithm. The average time and average path length of 10 path plannings of the present invention, the RRT*, and the clipped RRT* algorithm on complex maps are shown in Table 1.

[0103] Table 1 Final path comparison

[0104]

[0105]

[0106] It can be seen from Table 1 that compared with the time and path length of the final path generated by the traditional RRT* algorithm, the final path generated by the present invention is better.

Claims

1. A robot path planning method based on an improved RRT algorithm, characterized in that: The following steps are involved: Step 1: Obtain the environment information around the robot and build a coordinate system; the environment information includes the initial point Q start , obstacles and target point Q goal location information; The initial point Q start Join the random tree as the root node of the random tree; Step 2: Get the sampling point Q in the constructed coordinate system sample ; Sampling point Q sample Azimuth and sampling radius r limit The expression is: r limit =a3r max +a4r new +a5eps Among them, a1 is the goal-oriented coefficient, a1>k max -k min θ initial The initial point Q start and the target point Q goal The angle between the line connecting the two and the x-axis; k is a random number with an expected value of zero, k∈[k min , k max ]; k min and k max are the minimum and maximum values ​​of the preset random number k; a2 is the radian global coefficient; a3 is the expansion coefficient; r max is the maximum sampling radius; a4 is the global coefficient of step length; r new For the last newly added node Q n ' ew With the initial point Q start Euclidean distance; a5 is the random coefficient of the step size related to the sampling collision coefficient col, a5 = rand-col; rand is a random number in a preset range; col is the sampling collision coefficient; Q is a node on the random tree; esp is the maximum growth step size, and its expression is: Among them, s is the preset minimum growth step size; x step is the step length coefficient; If there is only a root node on the random tree, the root node is used as the last newly added node Q n ' ew ; According to the sampling point Q sample Get a random point Q rand ; Step 3: Traverse all nodes on the random tree and calculate the relationship between each node and the random point Q rand Euclidean distance, get the distance random point Q rand The nearest node is used as the feature node Q near ; According to the random point Q rand and feature node Q near Get the newly added node Q new ; Step 4: If a new node Q is added new With characteristic node Q near If the line collides with an obstacle, repeat steps 2 and 3 until the newly added node Q is obtained. new With characteristic node Q near There are no obstacles on the connection line, and the newly added node Q new Add to the random tree; update the step coefficient x according to the number of collisions num step and the sampled collision coefficient col; Step 5: Update the newly added node Q new The parent node Q parent ; Step 6: Traverse the existing nodes on the random tree and calculate the relationship between each node and the target point Q goal The Euclidean distance between them; if there is a node and the target point Q goal If the Euclidean distance between them is less than the minimum growth step s, then this node is used as the target point Q goal The parent node of start Move to the target point Q goal If the path is not found, the iteration process ends; otherwise, steps 2 to 5 are repeated to continue expanding the random tree.

2. The robot path planning method based on the improved RRT algorithm according to claim 1, characterized in that: In step 2, if the robot can move in three-dimensional space, a sampling point Q is also constructed. sample The polar angle θ of , the expression of polar angle θ is: θ=b1(θ initial +k)+2b2π Where b1 is the goal-oriented coefficient, b1>k max -k min ; b2 is the global coefficient of radians.

3. The robot path planning method based on the improved RRT algorithm according to claim 1, characterized in that: In step 5, after obtaining the initial point Q start Move to the target point Q goal After the path is obtained, the redundant nodes in the path are removed. The method for removing redundant nodes is as follows: Take the target point Q goal As the end point, the initial point Q start As the starting point, connect the end point and the starting point; if the line connecting the starting point and the end point collides with an obstacle, the child node Q of the starting point is used. child Update the starting point until the line connecting the starting point and the end point does not collide with obstacles; use the updated starting point as the end point, and the initial point Q start As the starting point, repeat the above process until the end point is equal to the initial point Q start The connecting line does not collide with the obstacle.

4. The robot path planning method based on the improved RRT algorithm according to claim 3, characterized in that: In step 5, after removing redundant nodes, each node in the path is selected as the mobile node Q in turn. choose ; First move the mobile node Q choose Move to its parent node, and the translation vector d1 of each movement is: Until the mobile node Q choose The line connecting its child nodes collides with an obstacle or moves the node Q choose After reaching its parent node, stop moving towards its parent node; then move the mobile node Q choose Move to its child node, and the translation vector d2 of each movement is: Until the mobile node Q choose The line connecting to its parent node collides with an obstacle or moves the node Q choose After reaching its child node, stop moving towards its child node; after all nodes in the path are moved, obtain the updated path as the initial point Q start Move to the target point Q goal The final path.

5. The robot path planning method based on the improved RRT algorithm according to claim 1, characterized in that: In the step 2, if there is no potential field U in the process of obtaining the sampling point req , with sampling point Q sample As a random point Q rand ; On the contrary, random point Q rand The expression is: Among them, Q req is the center point of the potential field; F req is the sampling point Q sample The repulsive force received.

6. The robot path planning method based on the improved RRT algorithm according to claim 5, characterized in that: The sampling point Q sample The repulsive force F req The expression is: Among them, k req is the potential field gain coefficient; r is the potential field action radius; n is the adjustment factor coefficient.

7. The robot path planning method based on the improved RRT algorithm according to claim 5, characterized in that: The potential field U req The confirmation method is as follows: If the neighboring node set Q neighbor If the number of elements in [] exceeds the preset threshold, a new node Q is added. new As the center point of the potential field Q req , generating only affects the next random point Q rand The potential field U req , and reassign the maximum sampling radius r max and distance r new , the maximum sampling radius r after assignment m ' ax and distance r n ' ew The expression is: r m ′ ax ＝r max -1.1r r n ′ ew ＝r new -1.1r The maximum sampling radius r after assignment m ' ax and distance r n ' ew The maximum sampling radius r of the next sampling node max and distance r new .

8. The robot path planning method based on the improved RRT algorithm according to claim 1, characterized in that: In step 5, the newly added node Q is updated new The parent node Q parent The method is as follows: Take the center of the circle as the new node Q new , all nodes within the circle with a radius of potential field action radius r are regarded as the neighboring node set Q neighbor []; Get the neighboring node set Q neighbor [] From each node to the newly added node Q new The total cost of cost: cost=cost neighbor +dist(Q neighbor -Q new ) Among them, Q neighbor is the set of neighboring nodes Q neighbor Neighboring nodes in []; cost neighbor The initial point Q start Arrival at neighboring node Q neighbor the cost; Take the neighboring node set Q neighbor [] From each node to the newly added node Q new The node corresponding to the minimum cost is the minimum cost node Q min ; If the minimum cost node Q min With the newly added node Q new If there is no collision between nodes, the node Q with the minimum cost min As a new node Q new Parent node.

9. The robot path planning method based on the improved RRT algorithm according to claim 1, characterized in that: In step 3, the newly added node Q is obtained new The method is as follows: Among them, dist(Q rand ,Q near ) is a random point Q rand With characteristic node Q near The Euclidean distance between .

10. The robot path planning method based on the improved RRT algorithm according to claim 1, characterized in that: In step 4, the step length coefficient x step The update method is: in the original step coefficient x step Add a small step size Δx to the step , small step length Δx step It decreases with the increase of the number of collisions num; the updating method of the sampling collision coefficient col is: adding a small coefficient Δcol to the original sampling collision coefficient col, the initial value of the small coefficient Δcol is 0, and after two consecutive collisions, the small coefficient Δcol increases by a fixed value each time the number of collisions increases until the small coefficient Δcol reaches the preset maximum value.

Images