Modeling and movement control of Mobile SMA-Net

Abstract

This paper discusses the modeling of Mobile SMA-Net, and movement control. Mobile SMA-Net is a robot which combines the spring of a shape memory alloy in the shape of a lattice, and has very redundant flexibility. Therefore, the physical characteristic is modeled based on a real robot's measurement data, a simulator is created, and movement is controlled. Since the robot has very large flexibility, in order that the robot may be controlled, he ingratiates himself in learning. However, in order that the robot may change his own form in the action process, learning and the mechanisms which can be re-learned are needed. In this paper, robust control is realized using reinforcement learning.