Abstract
In collision avoidance in a double-arm robot is focused on. One robot arm (Robot2) is taken as a moving obstacle and another arm (Robot1) should move from a start position to the goal one while avoiding the collision with Robot2. A comparatively short path can be searched efficiently in configuration space (C-space), i.e., joint angle space, by using Rapidly-exploring Random Tree (RRT) method. In this study, to cope with unknown obstacle's trajectory, real-time RRT method is proposed. This method assumes that the robot can obtain the information of unknown object movement at every interval. The robot predicts future obstacle's trajectory based on the information obtained before and now, and generates the trajectory avoiding the collision with the predicted trajectory based on the RRT search within the interval. Both in the simulation and the experiment, the proposed real-time RRT method effectively worked for collision avoidance.
