An Alternative Formulation of Dynamic-Programming Updates for POMDPs.
Weihong ZhangNevin Lianwen ZhangPublished in: AI&M (2002)
Keyphrases
- dynamic programming
- partially observable markov decision processes
- dec pomdps
- markov decision processes
- markov decision problems
- reinforcement learning
- stereo matching
- optimal policy
- infinite horizon
- state space
- linear programming
- optimal control
- finite state
- dp matching
- policy search
- dynamic programming algorithms
- planning under uncertainty
- coarse to fine
- greedy algorithm
- point based value iteration
- theoretical justification
- continuous state
- single machine
- distributed constraint optimization
- frequent updates
- search algorithm