Using Neural Network Model Predictive Control for Controlling Shape Memory Alloy-Based Manipulator

Abstract

This paper presents a new setup and investigates neural model predictive and variable structure controllers designed to control the single-degree-of-freedom rotary manipulator actuated by shape memory alloy (SMA). SMAs are a special group of metallic materials and have been widely used in the robotic field because of their particular mechanical and electrical characteristics. SMA-actuated manipulators exhibit severe hysteresis, so the controllers should confront this problem and make the manipulator track the desired angle. In this paper, first, a mathematical model of the SMA-actuated robot manipulator is proposed and simulated. The controllers are then designed. The results set out the high performance of the proposed controllers. Finally, stability analysis for the closed-loop system is derived based on the dissipativity theory.