Direct Adaptive Control of Nonaffine Magnetic Levitation System Using Radial Basis Function Neural Network

Magnetic levitation system (MLS) is seldom modeled as a nonlinear system. This hinders the incorporation of nonlinearity due to its control input into the model of the system. Literature survey reveals that it can also be modeled as a nonaffine nonlinear system. The controller design of non affine nonlinear system is performed with the help of artificial neural network (NN). But, NNs have disadvantages like local minima trapping and high training time. These issues are resolved to a great extent by resorting to the use of radial basis function NN (RBFNN) for the controller design of nonaffine nonlinear system. In this paper, we have proposed the use of RBFNN to design an adaptive controller for a magnetic ball position system based on MLS. The control objective is for the system output, ball position, to track a desired trajectory. The proposed controller fulfills the control objective better than an adaptive controller based on sigmoid activation function NN.