Real-Time Obstacle Avoidance Algorithm for Dynamic Environment on Probabilistic Road Map

This paper deals with the motion planning problem for a robot operating in presence of dynamic obstacles. In our approach, we used the Probabilistic Road Map technique for sampling the configuration space in offline mode followed by finding an optimal path with the help of A-star. Whenever a dynamic obstacle appears in the path of the robot, we locally resample the map for obstacle avoidance. The algorithm samples the free configuration space in the vicinity of the obstacles but, with a fewer number of samples, then again uses an $A^{\ast} $ based algorithm to find the path around the dynamic obstacles and finally superimpose the two paths. Motion planning in presence of dynamic obstacles can be memory and computation-intensive. Therefore, classical motion planning techniques fail to perform reliably when implemented in online mode. The approach we propose here is memory efficient and versatile. The algorithm can be efficient from the viewpoint of computation time but, can also produce an optimal path depending upon the number of nodes used in the re-planning phase and tuning of certain parameters. The algorithm is capable of handling both polygonal and circular obstacles. We demonstrate these capabilities with the help of simulations.