Robotic juggling by iterative learning control using optimization

Abstract

This paper proposes the control strategy for throwing as one of the dynamic manipulation. In dynamic manipulation, the robot's motion is dynamic and quick and there is no constant contact state. The dynamic manipulation has strong nonlinearity between control input and output, which often makes the manipulation unstable. And modeling errors affect the success of its manipulation task seriously. The dynamic manipulation requires more powerful actuator than the static manipulation. We propose the control method using the learning control that can consider the modeling error problem, the performance limits of the actuator and the stability of the learning control.

Imitation Learning for Nonprehensile Manipulation Through Self-Supervised Learning Considering Motion Speed

“…In particular, it is difficult for model-based methods to adapt to change in the state of contact between the object and environment. Iterative learning control, which performs repetitive motions and corrects modeling errors, has also been reported [28]. However, [28] assumes that the desired trajectory is predetermined, which makes it difficult to adapt to changes in the environment.…”

“…Iterative learning control, which performs repetitive motions and corrects modeling errors, has also been reported [28]. However, [28] assumes that the desired trajectory is predetermined, which makes it difficult to adapt to changes in the environment.…”