Inverse Reinforcement Learning for Optimal Control Systems
摘要
To overcome the need forOptimal control manual specification of cost functions in an agent, inverse reinforcement learningReinforcement Learning (RL) (inverse RL) (Ng and Russell 2000) has been introduced to infer the hidden cost functions from demonstrated behaviors. Inverse RL is commonly employed in apprenticeship learning scenarios (Abbeel and Ng 2004; Chu et al. 2020; Lin et al. 2019; Self et al. 2020; Song et al. 2018; Syed and Schapire 2007), where a learner leverages observations of an expert’s behavior to uncover the unknown expert cost functions and replicate the expert’s behavior.