The risk probability criterion for discounted continuous-time Markov decision processes.
Haifeng HuoXiaolong ZouXianping GuoPublished in: Discret. Event Dyn. Syst. (2017)
Keyphrases
- markov decision processes
- risk sensitive
- state space
- optimality criterion
- average reward
- stationary policies
- optimal policy
- finite state
- dynamic programming
- infinite horizon
- reinforcement learning
- finite horizon
- policy iteration
- expected reward
- average cost
- probability distribution
- partially observable
- transition matrices
- markov chain
- discount factor
- dynamical systems
- planning under uncertainty
- reinforcement learning algorithms
- model based reinforcement learning
- semi markov decision processes
- reachability analysis
- action sets
- factored mdps
- action space
- decision making
- markov decision process
- discounted reward
- decision theoretic planning
- state and action spaces
- optimal control
- search space
- multistage
- total reward
- initial state