An adaptive scheme for robot joint trajectory generation

It has been shown recently that robot inverse kinematic transformation can be easily carried out by using a nonlinear dynamic system called “joint space command generator.” In this paper, an adaptive version of the command generator is proposed. The main feature of this command generator is that the inverse Jacobian is estimated on-line (using least-squares algorithms), thus the explicit form of the inverse Jacobian is not required. Since the Jacobian is dependent on the end-effector orientation representation, the adaptive command generator is then inherently flexible in accommodating any choice of orientation representation during Cartesian trajectory planning. The only knowledge required is the forward kinematics of the robot. Thus inverse kinematic transformation of robot trajectories can be carried out with a minimum amount of kinematic information. Extensive simulation studies of the proposed scheme have been carried out which showed that the adaptive command generator is computationally feasible and highly accurate. Different orientation representations have been tested with equal success. Some typical simulation results are presented to illustrate the performance of the adaptive command generator.