Abstract: Aiming at local optimal path planning of mobile robots in complex environments， a bidirectional rapidly exploring random tree star(Bi-RRT^*)path planning algorithm based on target offset expansion and Cantmull-Rom spline interpolation is proposed. In the Bi-RRT^* algorithm， two search trees are created at the same time， and opposite searches are alternately performed with the target bias of random points selected at a certain probability， in order to improve the overall convergence efficiency of the algorithm. By reselecting the parent node of the current node and rewiring the nodes， the sensitivity to the environment of path planning is enhanced. To ensure the path safe and feasible， the obstacles in environment are expanded and collision detection is performed on the initial path. Redundant nodes are trimmed so that the feasible path length is shortened. Then， the path is smoothed by the Cantmull-Rom spline interpolation. The 2D and 3D comparative experiments were conducted for the improved Bi-RRT^* algorithm based on Matlab and ROS robot simulation platform， respectively， and the experiment results verified the effectiveness and superiority of the improved Bi-RRT^* algorithm.

Key words: rapidly exploring random tree; Cantmull-Rom spline interpolation; ROS robot simulation platform; improved Bi-RRT^*