An on-line cartesian space obstacle avoidance scheme for robot arms

This paper presents an on-line obstacle avoidance scheme for robot arms. This method is directly formulated in the Cartesian space and uses complete information about the workspace. The strategy is to detect the collision and to plan the motion of each link successively, starting from the base link. Each robot link in 3D Cartesian space will be mapped into two 2D planes such that the collision detection and motion planning can be executed in 2D spaces. Depending on its spatial shape, the end-effector can be approached as further robot links. Thus, the obstacle avoidance is more effectively including the consideration of end-effector. The proposed algorithm is implemented on a transputer network which simulates the control system of a six-link robot PUMA 560.