Integration of observer and neural network based sliding mode control and proportional integral derivative control for high performance electro hydrostatic actuator

Abstract Electro-hydrostatic actuator (EHA) is a self-contained electrically powered hydraulic actuator. This paper deals with the design of high-performance control schemes without demanding complete information about the EHA. The dynamics of EHA is nonlinear, and it is subjected to uncertainties and external disturbances. To deal with these problems, sliding mode control (SMC) is suitable. However, the drawback of SMC is chattering. To meet the high performance of EHA and reduce the chattering of SMC, proportional integral derivative (PID) control is proposed for the inner loop of EHA control. For the outer loop, extended state observer (ESO) and radial basis function neural network (RBFNN) based SMC is designed. ESO is used to estimate the states of EHA whereas RBFNN is used to get the approximate value of the unknown external disturbances, uncertainties and nonlinear dynamics of the EHA. Fixed centers and widths of RBFNN are used and the weights are updated based on an adaptive law derived using a Lyapunov stability analysis. An approximate nonlinear mathematical model of EHA is derived, and a virtual prototype of EHA is built in AMESim software. Then Matlab/Simulink and AMESim cosimulation is carried out. The cosimulation indicated the tracking performances of EHA subjected to different desired signals, parameter variations and external disturbances. Compared to using only PID control for both outer and inner loops, the integration of PID, ESO and RBNN based SMC of EHA indicates superior tracking performance and robustness. Chattering of SMC is also significantly reduced.