Game-based Abstraction for Markov Decision Processes.
Marta Z. KwiatkowskaGethin NormanDavid ParkerPublished in: QEST (2006)
Keyphrases
- markov decision processes
- decision theoretic planning
- state abstraction
- finite state
- transition matrices
- optimal policy
- policy iteration
- state space
- reinforcement learning
- finite horizon
- dynamic programming
- average cost
- partially observable
- reinforcement learning algorithms
- action space
- decision processes
- reward function
- factored mdps
- markov decision process
- planning under uncertainty
- infinite horizon
- reachability analysis
- model based reinforcement learning
- average reward
- macro actions
- risk sensitive