Abstract
We study the use of inverse reinforcement learning (IRL) as a tool for recognition of agents on the basis of observation of their sequential decision behavior. We model the problem faced by the agents as a Markov decision process (MDP) and model the observed behavior of an agent in terms of forward planning for the MDP. The reality of the agent’s decision problem and process may not be expressed by the MDP and its policy, but we interpret the observation as optimal actions in the MDP. We use IRL to learn reward functions for the MDP and then use these reward functions as the basis for clustering or classification models. Experimental studies with GridWorld, a navigation problem, and the secretary problem, an optimal stopping problem, show algorithms’ performance in different learning scenarios for agent recognition where the agents’ underlying decision strategy may be expressed by the MDP policy or not. Empirical comparisons of our method with several existing IRL algorithms and with direct methods that use feature statistics observed in state-action space suggest it may be superior for agent recognition problems, particularly when the state space is large but the length of the observed decision trajectory is small.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.