Reinforcement learning-enhanced adaptive sliding mode control for nonlinear systems

Abstract This paper presents a novel Adaptive Relaxed Sliding Mode Control (ARSMC) framework for nonlinear systems, enhanced by an actor-critic reinforcement learning algorithm. ARSMC addresses the limitations of traditional sliding mode control by providing a chattering-free control signal, minimizing steady-state error, and effectively managing unbounded uncertainties and disturbances. The algorithm’s stability is rigorously proven via the Lyapunov theorem. Simulations on an electronic throttle system demonstrate ARSMC’s superior performance over traditional and relaxed sliding mode controllers under various disturbances and uncertainties. Performance metrics, including integral absolute error (IAE) and integral squared error (ISE), highlight ARSMC’s improved control precision and robustness. The results also showcase enhancements in the rise time and maximum overshoot, further validating its effectiveness. The proposed ARSMC framework marks a significant advancement in nonlinear control, providing a robust solution for systems with complex dynamics and parameter uncertainties. It improves upon traditional sliding mode controllers by minimizing steady-state error and chattering while maintaining the ability to handle unbounded external disturbances.