An Improved Rapidly-exploring Random Tree Star Algorithm for Manipulator Path Planning

The rapidly-exploring random tree star (RRT*), which is a variant of the rapidly-exploring random tree (RRT), has been used in the path planning of robots and gained good practicability during the last decade. However, because the algorithm parameters are fixed in the whole process of path planning, the RRT* is inefficient for the complex environment. For addressing this problem, this paper proposes two schemes to improve the efficiency of robot path planning by dynamically adjusting the parameters of RRT*. First, the dynamic incremental distance scheme (DID) is proposed to promote the flexibility to avoid obstacles by adjusting the incremental distance of RRT* according to the obstacle density of environment. Then, in order to guarantee the RRT* is not trapped in the local minima and covers the unexplored areas quickly, the subregion sampling scheme based on Gaussian distribution (SSG) is presented. Based on the proposed schemes, the improved RRT* can rapidly explore a valid path by reducing the number of iterations. Finally, a series of experiments based on the various complex 2D environments and a 7-DOF manipulator in complex environments are adopted to demonstrate the effectiveness of the improved RRT*. The results show that the improved RRT* can significantly improve the efficiency of manipulator path planning without increasing the path distance.