An adaptive neural network backstepping control method for reticle stage of lithography

Abstract

This paper proposes an adaptive neural network backstepping tracking control method for the long-stroke reticle stage of lithography. The main feature of this method is that it can real-time online estimate the unknown model parameters of long-stroke reticle stage actuator under some nonlinear effects including load disturbance, force ripple and slot effect by radial basis function(RBF) neural network. First, in the study according to the actual system a nominal model of long-stroke reticle stage is established. Then, with this nominal model the derivation of the control method, the design of the adaptive law and the Lyapunov stability proof are made. From the results of the analysis above, it can be seen that this control method can sufficiently guarantee the convergence of the system's position and velocity errors. At last, this control method is verified by a simulation experiment. From the result of fifth-order S-curve tracking test, the tracking accuracy of the position and velocity can meet the design requirements well. Because it doesn't need to model the actual mechanical system and external uncertainties precisely, this method is very suitable for application in the field of ultra-precision manufacturing.