A fast RRT algorithm for motion planning of autonomous road vehicles

The Rapidly-exploring Random Tree (RRT) is a classical algorithm of motion planning based on incremental sampling, which is widely used to solve the planning problem of mobile robots. But it, due to the meandering path, the inaccurate terminal state and the slow exploration, is often inefficient in many applications such as autonomous road vehicles. To address these issues and considering the realistic context of autonomous road vehicles, this paper proposes a fast RRT algorithm that introduces an off-line template set based on the traffic scenes and an aggressive extension strategy of search tree. Both improvements can lead to a faster and more accurate RRT towards the goal. Meanwhile, our approach combines the closed-loop prediction approach using the model of vehicle, which can smooth the portion of off-line template and the portion of on-line tree generated, while a trajectory and control sequence for the vehicle would be obtained. Experimental results illustrate that our method is fast and efficient in solving planning queries of autonomous road vehicle in urban environments.