Modeling and levitation control of a novel M-DOF actuator based on neural network

Abstract

Modeling and torque characteristics analysis are important for actuators or motors with complicated structures. Based on the presented configuration design, the magnetic field and torque calculation for a multiple degrees-of-freedom (M-DOF) spherical actuator are performed using 3D FEM. To reduce the computation workload, a simplified torque calculation model and non-linear system dynamic model have been proposed, and magnetic levitation mechanism is developed for alleviating the friction and resistance to improve the dynamic performance of this M-DOF actuator. A linear mathematical model of the levitation force is deduced and corresponding control scheme with power drive unit is discussed. Then the diagonal recurrent neural network (DRNN) based control scheme is presented to improve the performance of levitation by adjusting parameters online as identifiers. Simulation and experiment results show that the actuator can effectively implement 3-DOF motion combined with levitation mode, and that the proposed approach can be applied to M-DOF actuators of the same kind effectively.