Sliding Mode Control with Gaussian Process Regression for Underactuated Mechanical Systems

This work introduces a new control scheme for uncertain underactuated mechanical systems. The proposed approach is mainly based on sliding mode control, but a Gaussian process regressor is also embedded in the control law for uncertainty estimation and compensation. The convergence properties of the closed-loop signals are analytically proved by means of the Lyapunov stability theory. Numerical simulations with an inverted pendulum on a cart are presented to confirm the improved performance of the proposed control scheme.