Sensor-less Obstacle Collision Detection for Robot Manipulator

Abstract

Accurate collision detection has been a challenging task for researchers for the past few years. Many studies include using sensors to detect a collision, but recent research has been done on sensor-less collision detection. Many researchers present different observer techniques to estimate collision, but the drawback of these techniques is that they are based on exact dynamic assumptions. In the real system, it is challenging to calculate the exact dynamic model of the system as the Degree of Freedom (DOF) in the system increases. This study proposed a Sliding Perturbation Observer (SPO) for obstacle collision detection, which is sensor-less and based on minimal dynamic (linear dynamics) information. The technique is implemented on 5 DOF robot manipulator. The external disturbance (which is collision) has been estimated by subtracting the perturbation (from SPO) from gravity (calculated graphically). A threshold has been set for the collision so that if the estimated disturbance exceeds the threshold, it can be concluded that the collision has occurred.