A new approach for robot motion planning using rapidly-exploring Randomized Trees

In the last few years, car-like robots became increasingly important. Thus, motion planning algorithms for this kind of problem are needed more than ever. Unfortunately, this problem is computational difficult and so probabilistic approaches like Probabilistic Roadmaps or Rapidly-exploring Randomized Trees are often used in this context. This paper introduces a new concept for robot motion planning especially for car-like robots based on Rapidly-exploring Randomized Trees. In contrast to the conventional method, the presented approach uses a pre-computed auxiliary path to improve the distribution of random states. The main contribution of this approach is the significantly increased quality of the computed path. A proof-of-concept implementation evaluates the quality and performance of the proposed concept.