Non-randomized policies for constrained Markov decision processes.
Richard C. ChenEugene A. FeinbergPublished in: Math. Methods Oper. Res. (2007)
Keyphrases
- markov decision processes
- optimal policy
- markov decision process
- reward function
- decision processes
- average cost
- reinforcement learning
- finite state
- discounted reward
- state space
- policy iteration
- stationary policies
- macro actions
- transition matrices
- total reward
- dynamic programming
- decentralized control
- finite horizon
- decision problems
- expected reward
- policy iteration algorithm
- infinite horizon
- long run
- partially observable markov decision processes
- decision theoretic planning
- control policies
- markov decision problems
- partially observable
- action sets
- risk sensitive
- sufficient conditions
- reinforcement learning algorithms
- average reward
- multistage
- reachability analysis
- planning under uncertainty
- initial state
- model based reinforcement learning
- semi markov decision processes
- factored mdps
- state abstraction
- learning algorithm