Integral Reinforcement Learning Control of an Uncertain 2-DOF Helicopter System With Input Quantization and State Constraints

This study presents an integral reinforcement learning (RL) control strategy for an uncertain two-degree-of-freedom (2-DOF) helicopter system by considering the input quantization and state constraints. Initially, a hysteresis quantizer is utilized to reduce oscillations when transmitting input signals. Within the control framework, the model uncertainty is estimated by an action neural network (NN), while the long-term cost function is approximated by a critic NN. Moreover, integral barrier Lyapunov functions (IBLFs) are employed to directly restrict the states, thereby guaranteeing the convergence of the helicopter system. Lyapunov stability theory demonstrates that all signals within the closed-loop system are semiglobal uniform ultimate bounded. Ultimately, the efficacy of the proposed control strategy is validated via both simulations and experiments.