Path Planning for Robotic Manipulator in Complex Multi-Obstacle Environment Based on Improved_RRT

Abstract

To plan a successful and practically executable path for a manipulator in a complex obstacle environment, the Improved_RRT method is proposed. In this article, we develop the collision detection model between the manipulator and obstacle based on the cylinder and sphere bounding box, and the shortest distance between the link and obstacle at each moment is obtained. Afterward, hybrid constrained sampling is used instead of original random sampling. The distribution of nodes obtained by sampling is close to the passages between obstacles. During the sampling process, the forward search process is accelerated. In the expansion, we propose a novel approach of RRT_Connect in collaboration with the artificial potential field method. Meanwhile, the adaptive gravitational field, dynamic step size, and new node constraint strategy are adopted. In the process of expansion, the algorithm uses the information of the environment and nodes to expand to the target area in a short time, which reduces the excessive exploration and collision area expansion. Considering the low success rate of algorithm planning in complex, multiobstacle and dynamic environments, local minimum processing and a dynamic path planning method are proposed, which generate double trees in the Cartesian and configuration spaces, respectively, for local path replanning. The improved algorithm is more robust to the singular problems of the manipulator, and the path is smoother after pruning optimization. Simulation results show that, compared with other algorithms, the proposed method can effectively plan a safe and optimal obstacle avoidance path with the best comprehensive performance. Experiments on the actual robotic manipulator platform further prove the efficiency of our method.