Continuous Time Markov Decision Processes with Expected Discounted Total Rewards.
Qiying HuJianyong LiuWuyi YuePublished in: International Conference on Computational Science (2003)
Keyphrases
- markov decision processes
- state space
- reinforcement learning
- finite state
- dynamic programming
- optimal policy
- markov chain
- transition matrices
- reinforcement learning algorithms
- stationary policies
- policy iteration
- partially observable
- reward function
- decision theoretic planning
- risk sensitive
- dynamical systems
- planning under uncertainty
- model based reinforcement learning
- finite horizon
- reachability analysis
- sequential decision making under uncertainty
- average cost
- decision processes
- factored mdps
- total reward
- markov decision process
- average reward
- discounted reward
- real time dynamic programming
- state abstraction
- decentralized control
- action space
- action sets
- long run
- planning problems