Periodic Finite State Controllers for Efficient POMDP and DEC-POMDP Planning.
Joni PajarinenJaakko PeltonenPublished in: NIPS (2011)
Keyphrases
- partially observable markov decision processes
- partially observable markov decision process
- belief space
- partially observable
- reinforcement learning
- planning problems
- belief state
- decision theoretic
- finite state
- dynamical systems
- partial observability
- state space
- model free reinforcement learning
- planning under uncertainty
- point based value iteration
- predictive state representations
- infinite horizon
- markov decision problems
- markov decision process
- continuous state
- stochastic domains
- optimal policy
- knowledge base
- decision problems
- markov decision processes
- partially observable stochastic games
- dynamic programming
- multi agent
- bayesian networks
- dec pomdps
- partially observable stochastic domains
- ai planning
- hidden state
- machine learning
- dynamic environments
- decision making