H₂/H_∞ Dynamic Output Synthesis Based on Regularized Reinforcement Learning

This study deals with a new method for designing a mixed ${H}_{2} / {H}_{\infty}$ dynamic output feedback controller based on reinforcement learning (RL). This paper aims to design an optimal machine learning-based controller that uses a robust and regularized reinforcement learning-based methodology to stabilize the system in the presence of disturbances. Furthermore, using the reinforcement learning algorithms, the dynamic output feedback controller is designed via policy evaluation and policy improvement steps. Eventually, the authors illustrate the advantages of the proposed methods through numerical simulations via error metrics.