Control and Obstacle Collision Avoidance Method applied to Human-Robot Interaction

In this work, we present a control and obstacle collision avoidance method for redundant robot manipulators operating in partially structured environments in the presence of humans. The control algorithm is based on the concept of artificial potential fields and it uses the pseudo-inverse of the Jacobian matrix with a weighting factor for the mechanical joint limits, taking advantage of the robot redundancy for the purpose of obstacle avoidance and control goal achievement. The detection algorithm uses a depth sensor based on the structured light to obtain a 2-1/2-D description of the surroundings from a point cloud. Repulsive fields are created around the detected obstacles, allowing for the robot to perform the task of interest without collisions. A filtering methodology based on geometric elements is presented to filter the RGB-D scene captured by the depth sensor, eliminating the robot body and the obstacles located outside its workspace. Experimental results, obtained with a Motoman DIA10 robot and a Microsoft Kinect(TM), illustrate the feasibility of the proposed scheme.