BetaZero: Belief-State Planning for Long-Horizon POMDPs using Learned Approximations.
Robert J. MossAnthony CorsoJef CaersMykel J. KochenderferPublished in: CoRR (2023)
Keyphrases
- belief state
- approximation methods
- partially observable markov decision processes
- belief space
- partially observable
- state space
- point based value iteration
- partial observability
- belief revision
- partial knowledge
- stochastic domains
- partially observable markov decision process
- dynamic bayesian networks
- reactive planning
- reinforcement learning
- machine learning
- dynamic environments
- semi supervised
- search space
- initial state
- optimal solution
- markov decision processes
- planning under uncertainty
- planning problems
- continuous state