Exploiting Fully Observable and Deterministic Structures in Goal POMDPs.
Håkan WarnquistJonas KvarnströmPatrick DohertyPublished in: ICAPS (2013)
Keyphrases
- machine learning
- fully observable
- partially observable markov decision processes
- partial observability
- partially observable
- reinforcement learning
- belief space
- markov decision problems
- planning problems
- hidden state
- markov decision processes
- goal state
- belief state
- state space
- dynamical systems
- finite state
- planning under uncertainty
- decision problems
- dynamic programming
- decentralized control
- optimal policy
- initial state
- multi agent
- higher order
- infinite horizon
- planning domains
- policy iteration
- dynamic environments
- knowledge base
- approximate solutions
- linear programming
- sufficient conditions
- belief revision
- heuristic search
- orders of magnitude