Neural network and extended state observer based sliding mode control of electro-hydrostatic actuators

Electro-hydrostatic actuators (EHA) are integrated, electrically powered, hydraulic actuators that are used to drive aircraft control surfaces or other moving parts that need hydraulic power. Some of the advantages of EHA are improved reliability, efficiency and reduction in the overall weight of the actuation system. However, design of a high performance EHA controller is challenging because of the variations of its parameters, nonlinear actuator friction, leakages and model uncertainty. To achieve the desired performance of an EHA, this paper proposes a hybrid control algorithm that combines the merits of radial basis function neural network (RBFNN) and sliding mode control (SMC). An RBFNN is used to approximate the uncertainties of EHA and the weights of its output layers are updated based on Lyapunov stability analysis. Besides, implementation of this control method demands full state availability of EHA and an extended state observer is designed. Furthermore, the mathematical model of an EHA involves derivative of a friction force and it is obtained by using a continuous approximation of a LuGre friction model. The performance of the proposed controller is compared with a PID controller. Simulation results illustrated the chattering elimination, superior tracking performance and robustness of the RBFNN based SMC.