Percentile performance criteria for limiting average Markov decision processes.
Jerzy A. FilarDmitry KrassKeith RossPublished in: IEEE Trans. Autom. Control. (1995)
Keyphrases
- markov decision processes
- average cost
- discounted reward
- optimal policy
- reinforcement learning
- policy iteration
- state space
- finite state
- transition matrices
- finite horizon
- dynamic programming
- partially observable
- infinite horizon
- reachability analysis
- risk sensitive
- decision theoretic planning
- reinforcement learning algorithms
- average reward
- state and action spaces
- stationary policies
- semi markov decision processes
- action space
- planning under uncertainty
- factored mdps
- learning algorithm
- model based reinforcement learning
- decision diagrams
- decision processes
- action sets
- total reward
- multistage
- finite number
- markov decision process
- long run