Reinforcement Learning, Adaptive Control, MPC for Vibration Control: A Survey

This paper reviews reinforcement learning (RL), adaptive control, model predictive control (MPC), and modal control strategies. It synthesizes progress across structural, aerospace, mechanical, and civil engineering applications, addressing key challenges e.g., model uncertainty, nonlinear dynamics, sensing and actuation constraints, and real-world deployment considerations. The review highlights several developments: actor-critic and off-policy RL methods extending optimal control to data-rich environments; adaptive control techniques ensuring robustness against unmodeled dynamics; MPC facilitating constraint handling and preview capabilities; and modal control retaining its essential role in managing lightly damped flexible structures. A comparative analysis of algorithms is provided, along with a summary of experimental validations. The article concludes by identifying open research challenges in safety assurance, sample efficiency, integration with physical principles, and the establishment of trustworthy benchmarking practices.