Robot Path Planning Research Based On Rapidly-exploring Random Trees

Path planning,as a fundamental part of mobile robot work,has now become a key and difficult direction of development in robotics.Existing path planning algorithms guide robots to work in simple environments,but in practice robots need to complete their work in complex environments,which puts higher requirements on robot movement.With the scholars’ in-depth research on path planning,path planning has now shown the development trend from simple environment to complex environment.Rapidly-exploring Random Trees Star(RRT*)is a random sampling-based path planning algorithm that searches for feasible paths in the robot’s configuration space with random equal probability,which is characterized by high speed,high success rate and wide range of applications,and is highly advantageous in the field of path planning for mobile robots in complex environments.Q-RRT*(Quick Rapidly-exploring Random Trees Star)algorithm is a global path planning algorithm,which adds two local path optimizations to the RRT algorithm.As a global path planning algorithm,the ant colony algorithm is a colony intelligence algorithm that simulates ants foraging,with high environmental adaptability and features such as fast feedback,parallel computation of multiple ant colonies,and easy integration with Q-RRT*.The research done in this paper is as follows.(1)For path planning by RRT* algorithm in complex environments,the blindness of searching path nodes leads to the generation of a large number of invalid nodes,making the invalid search account for the vast majority of the entire path planning.In this paper,the idea of potential field method is introduced to drive the expansion of nodes closer to the end point.The experimental results prove that the method in this paper can meet the robot path planning requirements while ensuring the real-time and stability of the algorithm.(2)To address the problem that the RRT* algorithm is biased away from the target point and close to the obstacle when searching the neighboring nodes of the sampling point,and the path planning time is long.A heuristic-based neighboring node finding algorithm is proposed to make the neighboring nodes more biased to the target point.Firstly,node sampling is performed using the target probability,greedy idea is introduced to find the neighboring points of the sampled points,then the idea of potential field is used to expand new nodes,and finally the algorithm of quadratic parent node replacement is designed for path optimization.(3)To address the problems of slow convergence and low node utilization in path planning of the Q-RRT* algorithm,this thesis incorporates a novel ant colony algorithm to improve it.The algorithm introduces pheromone to reduce the sampling range,improves the finding of neighboring points and new nodes,and increases the utilization rate of nodes and the global iteration speed.(4)To address the problem that the Q-RRT* algorithm is prone to fall into local optimality in complex environments and the algorithm search comes to a halt.In this thesis,a node correction mechanism is designed.The robot will check the deviation size of itself from the straight line of the current node and the target node after reaching the next node,and make a primary correction according to the angle function.After the primary correction,the robot proposes a jump point search and makes a secondary node correction according to the degree of influence of the obstacles on itself.