GA-pattern matching based manipulator control for real-time visual tracking

A control scheme for robot manipulators is presented, which makes use of visual information to direct the hand of a manipulator towards an object which position and orientation are not known, in its task space. The vision system employs genetic algorithm (GA) and a pattern matching technique, to search for a rectangular contour of a block in the robot task space image. The information derived from the vision system are continuously input into the robot controller so that a motion control is operated and the manipulator follows a path planned by GA, to reach the unknown target. Simulation results performed on a two-link robot manipulator, show that the manipulator hand reached the object as fast as the searching for an unknown target is completed, confirming the effectiveness of the controller. In fact this is a visual control strategy to the path planning problem for robot manipulators. This system may be used for real-time control of a manipulator to pick up an object which position and orientation are not known previously.