Generation of feedforward torque by reuse of ILC torque for three-joint robot arm in gravity

Abstract

The disadvantage that the iterative learning control (ILC) requires a learning process again for achieving another motion has been overcome by the basis-motion torque composition (BMTC). However, the increase of robot joints may cause difficulties in motion selections due to the algorithmic issue. We investigated the issue in motions of a three-joint robot arm in gravity and proposed a resolution method using redundant sets of motion torque. The calculation results demonstrated that the five sets of motions were enough for the torque composition though the motion sets which avoid algorithmic singularity were only 5 sets in 50,000 sets. The noise of generated torque was amplified to the signal/noise (S/N) ratio of 14.7[dB] in case of torque inputs of 40.0[dB]. However, the addition of one motion set improved to 29.4[dB].