Bisimulation Metrics for Continuous Markov Decision Processes.
Norm FernsPrakash PanangadenDoina PrecupPublished in: SIAM J. Comput. (2011)
Keyphrases
- markov decision processes
- action space
- state space
- reinforcement learning
- continuous state spaces
- optimal policy
- finite state
- transition matrices
- policy iteration
- reachability analysis
- planning under uncertainty
- reinforcement learning algorithms
- decision theoretic planning
- dynamic programming
- decision processes
- markov decision process
- risk sensitive
- partially observable
- reward function
- factored mdps
- state and action spaces
- action sets
- model based reinforcement learning
- average cost
- infinite horizon
- discounted reward
- objective function
- stochastic shortest path
- continuous state
- markov decision problems
- finite horizon
- long run
- linear programming