10th International Conference on Recent Advances in Air and Space Technologies, RAST 2023, İstanbul, Turkey, 7 - 09 June 2023
A reinforcement learning (RL) comparative study is carried out for the vibration control of a mass-spring-damper system. First, an analysis of the implementation of two RL algorithms proximal policy optimization (PPO) and deep-Q network (DQN) for the vibration control of the system with a discrete action space is carried out. Thereafter, an investigation on the effect of defining an action space as discrete or continuous is performed. A custom RL environment is created in MATLAB, with two constructors, one defining a discrete action space and another a continuous action space. The DQN and PPO agent are then trained on the discrete environment. For the continuous action space environment, only a PPO agent is trained. The trained agents are then simulated for their respective environments and the results presented.