Reactive planning on a collaborative robot for industrial applications

A challenge for roboticists consists in promoting collaborative robotics for industrial applications, i.e. allowing robots to be used close to humans, without barriers. Safety becomes the key issue. For manipulation tasks, a part of the problem is solved using new arms like the KUKA LWR, which is able to physically detect a collision from torque measurements on each joint. Nevertheless it is better to avoid collisions, especially if the obstacle is the arm or the head of an operator. This paper describes how obstacles could be detected and avoided, using a single Kinect sensor for the monitoring of the workspace and the reactive planner of the KineoWorks^TM software library for the real-time selection of an avoidance trajectory. Experimental results are provided as a proof that this dynamic obstacle avoidance strategy works properly.