Design and Performance Comparison of Different Predictive Controllers for Magnetic Levitation System

This paper deals with design of model based predictive controllers for single axis magnetic levitation system and verifi es the performance. The objective is to track the position of a ferromagnetic ball levitated with the aid of an electromagnetic force. Inherent nonlinearity and open loop unstable dynamics add challenges to this control problem. Three different controllers are designed: model predictive controller (MPC), approximate neural network MPC and adaptive neuro-fuzzy inference system (ANFIS) MPC. Controllers are designed in MATLAB environment and their performances are compared. MPC offers the advantage of minimum control effort while the other two controllers are found to require less computational time at the cost of comparatively higher control energy.