UMBRELLA: Uncertainty-Aware Model-Based Offline Reinforcement Learning Leveraging Planning.
Christopher DiehlTimo SievernichMartin KrügerFrank HoffmannTorsten BertramPublished in: CoRR (2021)
Keyphrases
- reinforcement learning
- partial observability
- model free
- sequential decision problems
- action selection
- sequential decision making
- planning problems
- partially observable
- function approximation
- reinforcement learning algorithms
- state space
- reinforcement learning problems
- belief space
- real time
- sensing actions
- learning algorithm
- decision theory
- uncertain data
- partially observable domains
- stochastic domains
- external events
- complex domains
- robust optimization
- planning process
- decision theoretic
- ai planning
- plan recognition
- belief functions
- motion planning
- transfer learning
- heuristic search
- optimal policy
- supervised learning
- dynamic programming
- learning process
- search space
- search algorithm
- multi agent
- decision making
- machine learning