Abstract: Learning feedforward control based on the available dynamic/kinematic system model and sensor information is generally effective for reducing the tracking error for a learned trajectory. For new trajectories, however, the ...