On-Line Obstacle Avoidance for Redundant Manipulators Using Artificial Vision

This paper presents an on-line obstacle avoidance method for redundant robot manipulators with joint physical limits reflected as space state constraints. The proposed controller considers two parts. First a Cartesian PID is used to reach closed loop system stability. The second controller part considers an optimization problem divided in two sections. The first one considers the Cartesian position error minimization at the manipulator end-effector. The second section considers obstacles avoidance that due to controller nature they can be time-varying obstacles. In this part of the controller artificial potentials to generate repulsive forces and barrier functions to avoid to overcome the space state constraints are added to the optimization problem to solve it. The gradient flow approach is used to solve the optimization problem on-line. The proposed controller is validated using a three degree of freedom planar robot as an experimental platform and a CMOS camera to get the obstacle position. Experimental results are presented, which show the avoidance of the obstacle in manipulator operational space and the successful task performance in Cartesian space, keeping cartesian errors small and without exceeding manipulator space state constraints.