Saute RL: Almost Surely Safe Reinforcement Learning Using State Augmentation.
Aivar SootlaAlexander I. Cowen-RiversTaher JafferjeeZiyan WangDavid Henry MguniJun WangHaitham AmmarPublished in: ICML (2022)
Keyphrases
- reinforcement learning
- state space
- function approximation
- action space
- optimal policy
- model free
- reinforcement learning algorithms
- partially observable
- learning algorithm
- hidden state
- markov decision processes
- continuous state
- supervised learning
- control problems
- policy iteration
- learning process
- control policy
- state action
- state abstraction
- real valued
- learning agent
- state transitions
- state and action spaces