Improved RRT pension service robot path planning method

Abstract

The invention relates to a robot path planning method, in particular to an improved RRT old-age service robot path planning method, which introduces a variable step size thought, can reduce the number of iterations by increasing the step size under the condition of fewer obstacles existing in an environment map, indirectly improves the problem of excessive redundant nodes, and improves the path planning efficiency of the old-age service robot. And the generation rate of the path is accelerated. And for the condition that many obstacles exist in the environment map, the step length can be reduced, so that the robot can find a feasible path in a complex environment, and the problem of path planning failure in the complex environment is solved. Meanwhile, an optimized target bias strategy is introduced, after the obstacle is passed, the target bias probability is increased, random nodes have the larger probability to directly grow around target points, the length of the generated path is shortened, and the path is smoother. The improved path planning method can adapt to environments with different complexities. A simulation experiment is carried out on the method provided by the invention, the feasibility of the method is verified, and the method has very high use value.

Description

Technical Field

[0001] The present invention belongs to the field of path planning for mobile robots, and particularly relates to a path planning method for an elderly care service robot that improves the RRT algorithm. Background Art

[0002] The change in the population structure has led to a sharp increase in the demand for elderly care service robots. In complex environments such as elderly care institutions, in order to ensure that service robots can efficiently and accurately perform various tasks, achieving fast path planning, precise obstacle avoidance, and smooth routes has become a basic requirement.

[0003] RRT is a rapidly exploring random tree algorithm. Its algorithm principle is simple and easy to understand. It does not require rasterization of the environmental map and has strong applicability to most environments. It does not require environmental gradient information. Even if the known map changes, the RRT algorithm can plan an effective path. Due to its random expansion characteristics, it has problems such as weak directivity, long search time, tortuous generated paths, and excessive redundant nodes in the process of connecting the tree, as well as problems such as excessive redundant nodes, over-dispersion, and weak directivity. Summary of the Invention

[0004] The purpose of the present invention is to provide a path planning method for an elderly care service robot that improves the RRT algorithm. This method introduces the idea of variable step size. When there are fewer obstacles in the environmental map, it can increase the step size to reduce the number of iterations, indirectly improving the problem of excessive redundant nodes and accelerating the path generation rate. It solves the problem of path planning failure in complex environments. The improved path planning method can adapt to environments with different complexities.

[0005] The technical solution of the present invention is as follows:

[0006] A path planning method for an elderly care service robot that improves the RRT algorithm, the method comprising the following steps:

[0007] Step 1: Initialize the environmental parameters to provide basic environmental information for path planning;

[0008] Step 2: Generate a random sampling point. First, generate a new random sampling point according to the initial target bias probability;

[0009] Step 3: Find the nearest node in the random tree to the new sampling point;

[0010] Step 4: Determine whether there is an obstacle between the random sampling point and the nearest node;

[0011] Step 5: Determine whether there is an obstacle between the sampling points randomly generated multiple times based on the same parent node;

[0012] Step 6: Increase or decrease the step size according to the specific situation;

[0013] Step 7: Dynamically adjust the target bias probability according to the number of obstacles passed;

[0014] Step 8: Derive backward from the target point to the root node to find a path from the starting point to the target point.

[0015] In the improved RRT path planning method for an elderly care service robot, the initialization of environmental parameters in Step 1 is specifically as follows: Determine the information of the initial point Xstart and the target point Xgoal, the total area of the environmental map and the total area of obstacles, and set the step size base step and the initial target bias probability P0 according to the environmental information.

[0016] In the improved RRT path planning method for an elderly care service robot, Step 2 is specifically as follows: Generate a random sampling point Xrand, and according to the initial target bias strategy, point to the target point Xgoal with a probability of P0, and perform random sampling in the environmental space.

[0017] In the improved RRT path planning method for an elderly care service robot, in Steps 3 and 4, if there is no obstacle between Xrand and the nearest node and the distance between them is less than the preset step size step0, then directly use Xrand as the new node of the tree, denoted as Xnew. If the distance between the two exceeds the step size, extend a distance of one step size along the direction of Xrand on Xnear to generate a new node as Xnew.

[0018] In the improved RRT path planning method for an elderly care service robot, in Steps 5 and 6, if the random sampling point passes through an obstacle for the first time, this new node will not be directly discarded, but the step size will be changed to the minimum step size Smin when generating this new node. The formula for Smin is as follows:

[0019]

[0020] If two consecutive random sampling points of the same parent node do not pass through an obstacle, it is defaulted that there are fewer obstacles at this position, and the step size of the second newly generated node will be changed to twice the original. If the random sampling point still does not pass through an obstacle subsequently, the subsequent step sizes will continue to increase according to the formula and cannot exceed the maximum step size Smax, as shown in the formula:

[0021]

[0022] The value of Smax is related to the ratio of the area of the environmental map to the total area of the obstacles. And for the selection of the subsequent step size, the priority of the parent node is higher than that of the child node. If the subsequent randomly sampled node passes through an obstacle, the step size will be reduced according to the formula until it is reduced to the minimum step size Smin. The formula for Smax is as follows:

[0023] step=2(i - 1)step0 i≥2

[0024] In the improved RRT path planning method for elderly care service robots, each time the random tree passes through an obstacle in step 7, the size of the target bias probability p0 is increased, and the growth formula is as follows:

[0025] P=P0 + 0.005*N q P≤0.7

[0026] The advantages and positive effects of the present invention are:

[0027] The path planning method of the present invention introduces the idea of variable step size. When there are fewer obstacles in the environmental map, it can increase the step size to reduce the number of iterations, indirectly improving the problem of too many redundant nodes and accelerating the path generation rate. When there are more obstacles in the environmental map, it can reduce the step size so that the robot can find a feasible path in a complex environment, solving the problem of path planning failure in a complex environment. At the same time, the present invention introduces an optimized target bias strategy. After passing through an obstacle, the target bias probability will be increased, making it more likely for random nodes to grow directly around the target point, shortening the length of the generated path and making the path smoother. The improved path planning method can adapt to environments with different complexities. The simulation experiment of the method proposed in the present invention verifies the feasibility of the method, which has strong application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is the overall flowchart of the improved RRT algorithm;

[0029] Figure 2 is the schematic diagram of the RRT algorithm;

[0030] Figure 3 is the schematic diagram of the random tree expansion with reduced step size of the present invention;

[0031] Figure 4 is the schematic diagram of the random tree expansion with increased step size of the present invention;

[0032] Figure 5 is the simulation experiment result diagram of the improved RRT algorithm;

[0033] Figure 6 is the path comparison diagram generated by the traditional RRT algorithm and the improved RRT algorithm of the present invention. Detailed implementation mode

[0034] The present invention will be described in detail below in conjunction with the embodiments in the accompanying drawings.

[0035] An improved RRT path planning method for a pension service robot, and the steps adopted by this method are as follows:

[0036] Step 1: Determine the environmental map, obstacle information of the working space where the service robot is located, the information of the initial point Xstart and the target point Xgoal, set the initial step size step, the initial bias probability P0, the target threshold, etc. according to the environmental information.

[0037] Step 2: The RRT algorithm starts from the starting point to create a random tree, and takes this starting point as the root node Xstart of the whole tree, and performs random expansion according to the set initial target bias probability. Randomly generate a new sampling point Xrand in the environmental space.

[0038] Step 3: Judge whether there is an obstacle between the random sampling point Xrand and the root node Xstart. If there is no obstacle between Xrand and Xnear, and the distance between them is less than the preset step size step, then directly take Xrand as the new node of the tree, denoted as Xnew.

[0039] Step 3: If the distance between the two exceeds the step size, extend a distance of one step size along the direction of Xrand on Xnear to generate a new node as Xnew.

[0040] Step 4: If there is an obstacle between the same parent node and the sampling points Xrand randomly generated multiple times later, the original step size will be reduced to Smin.

[0041] Step 5: If the two consecutive random sampling points of the same parent node do not pass through an obstacle, the original step size will be increased according to the formula, and the maximum cannot exceed Smax.

[0042] Step 6: For each randomly generated sampling point, there is a target bias probability P to intervene. A random number P1 will be generated within the range of 0 to 1 before each sampling. If P1 is greater than P, the algorithm will randomly select a point in the configuration space as the sampling point; otherwise, if P1 is less than or equal to P, the target point will be directly selected as the sampling point.

[0043] Step 7: The new target bias strategy will be increased according to the formula after the random tree passes through an obstacle, and the maximum does not exceed 0.7.

[0044] Step 8: Judge whether the distance between the new node Xnew and the target point is less than the target threshold. If it is satisfied, it is considered that a feasible path is found; otherwise, return to Step 2 to continue searching for the path.

[0045] Compared with the prior art, the main improvements of this method are as follows:

[0046] In steps 3 and 4, if a randomly sampled point passes by an obstacle for the first time, instead of directly discarding this new node, the step size will be changed to the minimum step size Smin when generating this new node. This method can effectively improve the problem that the traditional RRT fails in path planning due to excessive discarded nodes when there are many obstacles. At the same time, it will also improve the quality of path generation. The formula for Smin is as follows:

[0047]

[0048] In step 5, for the case of fewer obstacles, if two consecutive randomly sampled points from the same parent node do not pass by an obstacle, it is defaulted that there are fewer obstacles at this position, and the step size for generating the new node for the second time will be changed to twice the original. If the randomly sampled points still do not pass by an obstacle subsequently, the subsequent step sizes will continue to increase according to the formula, and cannot exceed the maximum step size Smax as shown in the formula:

[0049]

[0050] The step size growth formula is as follows:

[0051] step=2(i - 1)step0 i≥2

[0052] In the formula: i is the number of times the randomly sampled node has not passed by an obstacle continuously. S is the area of the environmental map, and S_0 is the total area of the obstacles.

[0053] In steps 6 and 7, a brand - new target - biasing strategy is used to replace the original strategy. The original fixed target - biasing probability is changed to increase according to the formula. When the random tree passes by an obstacle during random expansion, it is defaulted that the possibility of an obstacle existing between the new node and the target point decreases. At this time, the value of the target - biasing probability P is appropriately increased, and then compared with P1. The functional relationship formula for the change of the P value is as follows, and its maximum value is limited to 0.7.

[0054] P=P0 + 0.005*N q P≤0.7

[0055] Embodiment

[0056] As shown in the attached Figure 1 The technical solution for the present invention to solve the above - mentioned technical problem is to design an improved RRT path - planning method for a pension service robot, which is applied to a pension institution. The steps adopted by this method are as follows:

[0057] Step 1: Initialize.

[0058] This step determines the environmental map information, sets the map (-19, -19) as the starting point Xstart for path planning, (19, 19) as the target point Xgoal, sets multiple obstacles, sets the initial step size step to 1, the initial target bias probability P to 0.1, and sets the target threshold to 1.

[0059] Step 2: Generate a random sampling point Xrand.

[0060] Compare the initial target bias probability P with the randomly generated probability P1. If P1 is greater than P, the algorithm will randomly select a point in the configuration space as the sampling point; otherwise, if P1 is less than or equal to P, the target point will be directly selected as the sampling point.

[0061] Step 3: Find Xnear

[0062] Find the parent node Xnear closest to the random sampling point Xrand.

[0063] Step 4: Change the step size step

[0064] Judge whether the same parent node and two or more consecutive random sampling points pass through obstacles. If they pass through obstacles, reduce the step size, and the minimum reduction is to Smin. The formula is as follows:

[0065]

[0066] If they do not pass through obstacles, increase the step size, and it cannot exceed the maximum step size Smax. The formula is as follows:

[0067]

[0068] The step size growth formula is as follows:

[0069] step = 2(i - 1)step0 i ≥ 2

[0070] In the formula: i is the number of consecutive times the random node does not pass through obstacles. S is the area of the environmental map, and S_0 is the total area of the obstacles.

[0071] Step 5: Change the target bias probability P

[0072] According to the number of obstacles passed by the generated path, appropriately increase the initial target bias probability P, and then compare it with P1 to make the new random sampling point more likely to fall near the target point. The functional relationship formula for the change of the P value is as follows, and its maximum value is limited to 0.7.

[0073] P = P0 + 0.005 * N q P ≤ 0.7

[0074] Step 6: Judge whether the end point is reached

[0075] Determine whether the distance between Xnew and the target point Xgoal is within the threshold range 1. If it is, it is considered that the path planning has reached the target point; otherwise, return to step 2 to continue searching for the path.

[0076] The above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An improved RRT path planning method for elderly care service robots, characterized in that, The method includes the following steps: Step 1: Initialize the environmental parameters to provide basic environmental information for path planning; Step 2: Generate random sampling points. First, generate new random sampling points according to the initial target bias probability; Step 3: Find the nearest node in the random tree to the new sampling point; Step 4: Determine whether there are obstacles between the random sampling point and the nearest node; Step 5: Determine whether there are obstacles between the sampling points randomly generated multiple times based on the same parent node; Step 6: Increase or decrease the step size according to the specific situation; Step 7: Dynamically adjust the target bias probability according to the number of obstacles passed; Step 8: Deduce backward from the target point to the root node to find a path from the starting point to the target point.

2. An improved RRT-based path planning method for elderly care service robots according to claim 1, characterized in that, In the specific implementation of Step 1 for initializing the environmental parameters: Determine the information of the starting point Xstart and the target point Xgoal, the total area of the environmental map and the total area of the obstacles, and set the step size base step and the initial target bias probability P0 according to the environmental information.

3. An improved RRT-based path planning method for elderly care service robots according to claim 1, characterized in that, Specifically, Step 2 is: Generate a random sampling point Xrand, which is directed towards the target point Xgoal with a probability P0 according to the initial target bias strategy, and perform random sampling in the environmental space.

4. An improved RRT-based path planning method for elderly care service robots according to claim 1, characterized in that, In Steps 3 and 4, if there is no obstacle between Xrand and the nearest node, and the distance between them is less than the preset step size step0, then directly take Xrand as the new node of the tree, denoted as Xnew; if the distance between the two exceeds the step size, then extend a distance of one step size along the direction of Xrand on Xnear to generate a new node as Xnew.

5. An improved RRT-based path planning method for elderly care service robots according to claim 1, characterized in that, In Steps 5 and 6, if the random sampling point passes through an obstacle for the first time, this new node will not be directly discarded, but the step size will be changed to the minimum step size Smin when generating this new node; The formula for Smin is as follows: If the consecutive two random sampling points of the same parent node do not pass through an obstacle, it is defaulted that there are fewer obstacles at this position, and the step size for generating the new node for the second time is changed to twice the original. If the random sampling point still does not pass through an obstacle subsequently, the subsequent step sizes will continue to increase according to the formula, and cannot exceed the maximum step size Smax, as shown in the formula: It is set that the value of Smax is related to the ratio of the area of the environmental map to the total area of the obstacles; and for the selection of subsequent step sizes, the priority of the parent node is higher than that of the child node; if the subsequent random sampling node passes through an obstacle, the step size will be reduced according to the formula until it is reduced to the minimum step size Smin; The formula for Smax is as follows: step = 2(i - 1)step0, i ≥ 2.

6. An improved RRT-based path planning method for elderly care service robots according to claim 1, characterized in that, In Step 7, each time the random tree passes through an obstacle, the size of the target bias probability p0 is increased, and the growth formula is as follows: P = P0 + 0.005 * N q P ≤ 0.7

Images