Reproducibility analysis and experimental validation of motion-copying system

Abstract

This paper studies motion reproducibility of a motion-copying system, which is capable of storing and reproduc ing motion of human operators. In the paper, the motion repro ducibility is expressed as transfer function from a force response of the storage phase to a force response of the reproduction phase. According to analysis of the derived transfer function, this paper confirms that the motion reproducibility is varied depending on variations of force gain, position gain, bandwidth of disturbance observer, environmental stiffness, and viscosity frictions. By the experiment, validity of the analysis method of the motion reproducibility is verified.

Time Series Motion Generation Considering Long Short-Term Motion

“…In consideration of this, it is assumed that offline trajectory generation is better than online trajectory generation. In contrast, offline trajectory generation cannot handle sudden changes as much as motion copying can [10]. Therefore, it is necessary to have a learning method with both the long-term estimation ability of offline trajectory generation and the immediate response capability of online generation.…”

“…This is because action and reaction forces applied at the same place cannot be separated, hence making a force controller is difficult. In contrast, Yokokura et al demonstrated that the force information needed to reproduce motion can be obtained using bilateral control [10], which is a remote-control technology using a master robot and a slave robot [11] [12]. The master robot senses an action force from an operator, and the slave robot senses the reaction force from the environments.…”

“…The master robot senses an action force from an operator, and the slave robot senses the reaction force from the environments. However, this method [10] is intended to reproduce only the same behavior. Thus, the task cannot be achieved if the environment changes even slightly.…”