Reinforcement Learning Integrated Nonlinear Controller for Guaranteed Stability

This paper considers the stabilization of a class of uncertain upper-triangular systems. Reinforcement learning has demonstrated the ability to handle complex control tasks for its model-free property and adaptability. However, the stability cannot be assured. To guarantee the stability of the uncertain system, a reinforcement learning integrated nonlinear controller is introduced. It shows that the proposed controllers provide the adaptability of reinforcement learning and guaranteed stability for the uncertain upper-triangular system.