Finite-state approximation of Markov decision processes with unbounded costs and Borel spaces.
Naci SaldiSerdar YükselTamás LinderPublished in: CDC (2015)
Keyphrases
- markov decision processes
- finite state
- average cost
- policy iteration algorithm
- stationary policies
- optimal policy
- state space
- reinforcement learning
- dynamic programming
- policy iteration
- transition matrices
- reinforcement learning algorithms
- average reward
- partially observable markov decision processes
- total cost
- risk sensitive
- decision theoretic planning
- action sets
- partially observable
- finite horizon
- planning under uncertainty
- continuous state
- action space
- approximation methods
- vector quantizer
- markov chain
- decentralized control
- graphical models
- expected cost
- policy evaluation
- markov decision process