A motion planner for car-like robots based on rapidly-exploring random trees

Abstract

We propose a motion planner for car-like robots based on the rapidly-exploring random tree (RRT) method. Our motion planner was designed especially for cars driving on roads. So, its goal is to build trajectories from the car’s initial state to the goal state in real time, which stay within the desired lane bounds and keep a safe distance from obstacles. For that, our motion planner combines several variants of the standard RRT algorithm. We evaluated the performance of our motion planner using an experimental robotic platform based on a Ford Escape Hybrid. Our experimental results showed that our motion planner is capable of planning trajectories in real time, which follow the lane and avoid collision with obstacles.