Verifiable conditions for average optimality of continuous-time Markov decision processes.
Xiaolong ZouYonghui HuangPublished in: Oper. Res. Lett. (2016)
Keyphrases
- stationary policies
- markov decision processes
- average cost
- sufficient conditions
- state space
- action sets
- finite state
- optimal policy
- markov decision process
- dynamic programming
- average reward
- lot sizing
- reinforcement learning algorithms
- planning under uncertainty
- reinforcement learning
- policy iteration
- factored mdps
- optimal control
- finite horizon
- linear program
- partially observable
- action space
- infinite horizon
- risk sensitive
- decision theoretic planning
- transition matrices
- semi markov decision processes
- state and action spaces
- discounted reward
- reachability analysis
- reward function
- markov decision problems
- initial state