Non-collision checking RRT* algorithm for mobile robot motion planning

The rapidly-exploring random tree* (RRT*) is an asymptotic optimal sampling-based motion planning method, which has advantages of exploring high-dimensional space. However, it needs to implement lots of collision checking during each iteration of the RRT*. This results in the slow convergence speed and low efficiency in complex environments. To address these issues, this paper proposes a non-collision checking RRT* motion planning method. The collision checking operation in RRT* extension can be eliminated, and the collision risk assessment function in the cost function is added. The parent node connecting selection of each node is affected. When a node or edge collides with obstacles, the collision risk assessment function will increase significantly. In this way, the cost function value is increased synchronously. Meanwhile, the upper bound of the path cost is designed to ensure collision avoidance ability of motion planning. Finally, simulation results show that the proposed method has faster convergence speed, which is about 40% higher than RRT*. More advantages can be found in complex environments. The real-time planning ability of the proposed method is evaluated by the experiment of the mobile robot in the real environment. © 2020, Science Press. All right reserved.