Robust local obstacle avoidance for mobile robot based on Dynamic Window approach

Abstract

To travel within the real world safely, an autonomous mobile robot has to perform obstacle avoidance. One of the most popular approaches to solve local obstacle avoidance is Dynamic Window approach. This approach is to search optimal control command of the robot in velocity space directly. The command which maximizes objective function is then selected as output command. In the original Dynamic Window approach, in order to avoid obstacles, it concerns only obstacles on robot's trajectory. Therefore, the robot might crash with some obstacles which are near the trajectory but not on the trajectory. Therefore, in this work, we propose field Dynamic Window approach (F-DWA) which objective function is modified to consider the obstacles near the trajectory as well. This can be done by the use of histogram grid representation of obstacles to estimate the crashing probability of the trajectory. The objective function of Dynamic Window approach also depends on the crashing probability as well as other traditional factors. To test the robustness of proposed algorithm, experiment in real world is conducted. The result is the robot can travel safer when it concerns near-trajectory obstacles.