FP-IRL: Fokker-Planck-based Inverse Reinforcement Learning - A Physics-Constrained Approach to Markov Decision Processes.
Chengyang HuangSiddhartha SrivastavaXun HuanKrishna C. GarikipatiPublished in: CoRR (2023)
Keyphrases
- inverse reinforcement learning
- markov decision processes
- reward function
- bayesian nonparametric
- partially observable environments
- partially observable
- reinforcement learning algorithms
- reinforcement learning
- optimal policy
- state space
- policy iteration
- finite state
- markov decision process
- dynamic programming
- finite horizon
- preference elicitation
- planning under uncertainty
- temporal difference
- infinite horizon
- action space
- average reward
- state action
- control policies