QD-RRT: A motion planning method for manipulators based on RRT and Q-distance function
Hao Wu, Ye Ding, Xiangyang Zhu · Scientia Sinica Technologica · 2021
This paper introduces a goal-biased growing strategy and an obstacle avoidance method based on the Q-distance (QD) function to the rapidly-exploring random trees (RRT) algorithm. In the process of random tree generation, the goal-biased strategy guides the random tree to grow toward the target point with a certain probability. The penetration depth between a robot and environmental obstacles is calculated using the QD function when the robot collides with environment obstacles under some configurations, which are represented by the newly generated path nodes. The gradient of the QD function with respect to each joint angle of the robot is calculated using the differentiability of the QD function. The algorithm, QD-RRT, reduces the blindness and randomness of the RRT algorithm. Verified by MATLAB and CoppeliaSim, QD-RRT has faster computational speed and could generate a shorter path than RRT. The advantages of the improved algorithm are more obvious when the obstacles in the environment are more complicated or the path should pass through a long and narrow channel. This algorithm can be applied to some variants of RRT, such as RRT*.