Contextual Action with Multiple Policies Inverse Reinforcement Learning for Behavior Simulation

• DOI: 10.5220/0007684908870894
• Publication Date: 2019-03-15T06:54:07Z
• Authors: Nahum Alvarez; Itsuki Noda
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Machine learning is a discipline with many simulator-driven applications oriented to learn behavior. However, behavior simulation it comes number of associated difficulties, like the lack clear reward function, actions that depend state actor and alternation different policies. We present method for called Contextual Action Multiple Policy Inverse Reinforcement Learning (CAMP-IRL) tackles those factors. Our allows extract multiple functions generates profiles from them. applied our large scale crowd simulator using intelligent agents imitate pedestrian behavior, making virtual pedestrians able switch between behaviors depending goal they have navigating efficiently across unknown environments.