Intermittently Proving Dynamic Programming to Solve Infinite MDPs on GPUs.
Tsutomu InamotoYoshinobu HigamiShin-ya KobayashiPublished in: CANDAR (2013)
Keyphrases
- dynamic programming
- markov decision processes
- markov decision problems
- state space
- optimal policy
- reinforcement learning
- greedy algorithm
- decision theoretic planning
- optimal control
- stereo matching
- coarse to fine
- parallel algorithm
- decision theoretic
- infinite horizon
- partially observable
- average cost
- markov decision process
- finite horizon
- planning under uncertainty
- general purpose
- dec pomdps
- learning algorithm