Model-based least squares compensation for computation delays in leader-follower coordination of two manipulators

A scheme for the coordinated control of two manipulators carrying a common rigid object using discrete position feedback has been developed and experimentally tested using two PUMA robot arms and a hierarchical multiprocessor architecture. A leader-follower coordination scheme was adopted because of its computational efficiency. The computation of desired follower arm joint positions consumes one sampling interval. Another sampling interval is lost due to the fact that a desired follower arm position has to be computed before it can be assumed. The author emphasizes the design of a class of prediction equations to compensate for these timing delays. These equations are developed using only limited knowledge about the path of the leader arm end-effector. All other trajectory parameters are estimated using only this limited knowledge and current and past leader arm positions