Abstract: Aiming at the global optimal path planning of mobile robots in complex environments， an improved RRT*FN path planning algorithm based on target bias extension and Bezier interpolation method is proposed. The improved algorithm adopts a certain probability to select the target bias of random points when the initial path is not found. After determining the initial path， the heuristic sampling method is used to make random sampling points iteratively select around the initial path， which improves the guidance of path planning. When the improved algorithm has not found the initial path， the nodes far from the target point and no child nodes in the tree are deleted. When the improved algorithm finds the initial path， the nodes far from the optimal path and without child nodes in the tree are deleted， and the high-performance nodes are reserved to improve the efficiency of algorithm convergence to the optimal path. The path is smoothed by Bezier interpolation. 2D and 3D comparative experiments were carried out on MATLAB simulation platform and ROS robot simulation platform respectively. The effectiveness and superiority of the proposed algorithm are verified.

Key words: rapid-exploration random trees; global path planning; target bias selection; Bezier interpolation; ROS robot simulation platform