Safe Reinforcement Learning via Shielding under Partial Observability.
Steven CarrNils JansenSebastian JungesUfuk TopcuPublished in: AAAI (2023)
Keyphrases
- partial observability
- reinforcement learning
- partially observable
- symbolic model checking
- belief state
- state space
- planning problems
- markov decision process
- fully observable
- belief space
- function approximation
- learning agent
- partial information
- model free
- transfer learning
- machine learning
- markov decision processes
- planning under partial observability
- partially observable markov decision processes
- action selection
- planning domains
- hidden state
- dynamic environments
- deterministic domains