Dynamic Manipulation of Flexible Objects with Torque Sequence Using a Deep Neural Network

TL;DR

An acquisition method of a flexible object motion equation model using a deep neural network and a control method to realize a target state by calculating an optimized time-series joint torque command is proposed.

Abstract

For dynamic manipulation of flexible objects, we propose an acquisition method of a flexible object motion equation model using a deep neural network and a control method to realize a target state by calculating an optimized time-series joint torque command. By using the proposed method, any physics model of a target object is not needed, and the object can be controlled as intended. We applied this method to manipulations of a rigid object, a flexible object with and without environmental contact, and a cloth, and verified its effectiveness.

Figures (13)

• Figure 1: Our goal in this study.
• Figure 2: System overview of this study.
• Figure 3: Network structure of Dynamics-Net and calculation of optimized time-series torque command.
• Figure 4: The method to generate initial time-series torque commands before the optimization.
• Figure 5: Details of experimental setup A (rigid object), B (flexible object), and C (flexible object on the floor).