WCSAC: Worst-Case Soft Actor Critic for Safety-Constrained Reinforcement Learning.
Qisong YangThiago D. SimãoSimon H. TindemansMatthijs T. J. SpaanPublished in: AAAI (2021)
Keyphrases
- actor critic
- reinforcement learning
- worst case
- temporal difference
- policy gradient
- optimal control
- approximate dynamic programming
- reinforcement learning algorithms
- neuro fuzzy
- gradient method
- policy iteration
- function approximation
- state space
- upper bound
- learning algorithm
- model free
- lower bound
- rl algorithms
- control problems
- average reward
- np hard
- markov decision processes
- optimal policy
- adaptive control
- action selection
- dynamic programming
- control policy
- supervised learning
- reinforcement learning methods
- recursive least squares
- neural network
- transfer learning
- linear program