Nonapproximability Results for Partially Observable Markov Decision Processes
Christopher LusenaJudy GoldsmithMartin MundhenkPublished in: Universität Trier, Mathematik/Informatik, Forschungsbericht (2000)
Keyphrases
- partially observable markov decision processes
- finite state
- reinforcement learning
- planning under uncertainty
- dynamical systems
- decision problems
- dynamic programming
- belief state
- optimal policy
- continuous state
- planning problems
- markov decision processes
- multi agent
- state space
- belief space
- sequential decision making problems
- partially observable domains
- partial observability
- markov chain
- partially observable stochastic games
- partially observable markov
- partially observable
- dec pomdps
- stochastic domains
- point based value iteration
- approximate solutions
- initial state
- infinite horizon