Using Rewards for Belief State Updates in Partially Observable Markov Decision Processes.
Masoumeh T. IzadiDoina PrecupPublished in: ECML (2005)
Keyphrases
- belief state
- partially observable markov decision processes
- fully observable
- reinforcement learning
- state space
- partial observability
- belief space
- markov decision processes
- planning under uncertainty
- belief revision
- partially observable
- continuous state
- finite state
- partially observable markov decision process
- optimal policy
- stochastic domains
- reward function
- planning problems
- dynamic programming
- dynamical systems
- state variables
- heuristic search
- video sequences
- objective function