VISION-BASED MOVING OBSTACLE AVOIDANCE FOR CABLE-DRIVEN PARALLEL ROBOTS USING IMPROVED RAPIDLY EXPLORING RANDOM TREE

In this study, the improved vision-based rapidly exploring random tree (RRT) algorithm is proposed to address moving obstacle avoidance for cable-driven parallel robots (CDPRs). The improved RRT algorithm is goal-biased with dynamic stepsize makes it possible to implement in dynamic environments. For the implementation of algorithm on CDPRs, the improved RRT considers various collisions caused by the cable. The axis-aligned bounding box (AABB) algorithm is used for the fast re-planning during the RRT process. Additionally, the improved RRT algorithm premeditates the complex constrains include force feasible workspace (FFW) and the segment-to-segment angle. The related simulation is given in order to illustrate the algorithm. An experimental setup is also presented using the drone as a moving obstacle and the Faster-RCNN vision algorithm to obtain the coordinate of the drone. The experiment result shows that the proposed algorithm can apply in CDPRs with the dynamic environment validly.