Online kinematic Jacobian uncertainty compensation for robot manipulators using neural network

Abstract

For the Cartesian position controlled robot it is required to have an accurate mapping from the Cartesian space to the joint space in order to command the desired joint trajectories to achieve desired movements in the Cartesian space. That requires the correct kinematic Jacobian information. Since the actual mapping from the Cartesian space to the joint space is obtained at the joint coordinate not at the actuator coordinate, uncertainty in the Jacobian can be present. In the paper two feasible neural network schemes are proposed to compensate for the kinematic Jacobian uncertainty. Uncertainty in the Jacobian can be compensated by identifying either the actuator Jacobian matrix off-line or the inverse of that in online fashion. The case study of a stencilling robot is examined.