Real-time collision avoidance for a redundant manipulator in an unstructured environment

摘要

The problem of redundant manipulator collision avoidance in an unstructured environment is addressed in this paper based on the concept of modified impedance control. Instead of using a limited degree of redundancy (as in conventional methods) to find a collision-free trajectory, in the proposed approach, a robot's commanded joint torques are augmented by an "artificial" joint torque to provide correction for collision avoidance. "Artificial" collision forces are generated online according to the robot's posture and environment information (through knowledge of the robot kinematics and environment or proximity sensors on the robot). The corresponding artificial collision forces are converted to equivalent joint torques that would accomplish the collision avoidance manoeuvre. Then, the commanded joint torques are augmented so that a collision-free joint torque profile is achieved. Robot-to-environment collisions, robot self-collisions and robot constraints such as joint limit and singularity avoidance can be achieved using this method.