First Passage Optimality for Continuous-Time Markov Decision Processes With Varying Discount Factors and History-Dependent Policies.
Xianping GuoXinYuan SongYi ZhangPublished in: IEEE Trans. Autom. Control. (2014)
Keyphrases
- markov decision processes
- stationary policies
- average cost
- optimal policy
- state space
- markov decision process
- finite state
- action sets
- average reward
- total reward
- reinforcement learning
- dynamic programming
- reward function
- policy iteration algorithm
- decision processes
- policy iteration
- decision theoretic planning
- infinite horizon
- reachability analysis
- reinforcement learning algorithms
- planning under uncertainty
- linear program
- optimal control
- lot sizing
- transition matrices
- finite horizon
- decentralized control
- factored mdps
- state and action spaces
- model based reinforcement learning
- discounted reward
- multistage
- partially observable
- macro actions
- state abstraction
- action space
- risk sensitive
- control policy
- initial state
- real time dynamic programming
- markov chain