Applying Adaptive Wavelet Neural Network and Sliding Mode Control for Tracking Control of MEMS Gyroscope

Abstract

In this letter, an algorithm applying an adaptive wavelet neural network (AWNN) and sliding-mode control (SMC) is proposed, investigated and exploited for tracking control of micro-electromechanical system (MEMS) gyroscope. Such an AWNN model can be regarded as a special radius basis function neural network and utilizes Mexican hat function as activation function. Besides, Taylor expansion is used for analyzing activation radius, which is considered as an adaptive variable. The parameters of the MEMS gyroscope model are hard to obtain in engineering applications; thus, AWNN is designed to approximate the uncertain function of MEMS gyroscope and the unknown asymmetrical dead zone in the control scheme. The weights updating laws and the activation radius adaptive laws in AWNN are derived from the Lyapunov stability analysis, which results in the control error converging to the desired value and the weights and activation radius converging to its real value. To achieve the effect of error acceleration, a power function is used to design a sliding mode function. Computer simulation results substantiate the theoretical analysis and further demonstrate the efficacy of such an algorithm combined with AWNN and SMC for MEMS gyroscope control.