Obstacle avoidance using limit cycles

摘要

Recent years have shown an increased interest in the use of autonomous vehicles. These vehicles must find a path towards a desired location while avoiding obstacles. Since finding the optimal solution is computationally expensive, approximation techniques are often used. In the last decades several path planning techniques have been proposed. One novel technique considers mobile robots and uses the limit cycle strategy to avoid static obstacles (D. Kim and J. Kim, 2003). The main advantage of this method is the low computational cost, which makes this method especially useful for small robots with limited computational power. In short, the method relies on creating a circular limit cycle that a robot should follow in the proximity of an obstacle to avoid collision. In this thesis a new algorithm incorporating ellipsoidal limit cycles is presented for avoiding static 3D obstacles by using 3D limit cycles. The ellipsoidal limit cycles represent a safety zone around an obstacle. By using ellipsoidal limit cycles, the shape of the obstacles can be better represented. Ellipsoidal limit cycles are also useful when there is a preferred direction in which the obstacle should be avoided, e.g., to prevent passing in front of the obstacle. Further, the limit cycle method is combined with the velocity obstacle approach in order to avoid moving obstacles in 2D and 3D. To ensure that the robot chooses the optimal rotation direction with multiple and/or moving obstacles, a tree search is performed yielding the globally optimal rotation direction around each obstacle.