Escaping local optima in POMDP planning as inference.
Pascal PoupartTobias LangMarc ToussaintPublished in: AAMAS (2011)
Keyphrases
- planning problems
- partially observable markov decision process
- decision theoretic
- belief space
- partially observable markov decision processes
- partially observable
- stochastic domains
- planning under uncertainty
- motion planning
- reinforcement learning
- partial observability
- bayesian networks
- belief state
- optimal policy
- dynamical systems
- heuristic search
- ai planning
- bayesian inference
- finite state
- optimal solution
- markov decision processes
- evolutionary algorithm
- state space
- optimization algorithm
- sequential decision making problems
- point based value iteration
- hidden state
- dynamic bayesian networks
- domain independent
- probabilistic inference
- inference process
- markov decision problems
- search space
- planning domains
- predictive state representations
- belief networks
- partially observable stochastic domains