Verification and repair of control policies for safe reinforcement learning.
Shashank PathakLuca PulinaArmando TacchellaPublished in: Appl. Intell. (2018)
Keyphrases
- control policies
- reinforcement learning
- optimal policy
- continuous state
- control policy
- action space
- stochastic optimization problems
- function approximation
- model checking
- state space
- markov decision processes
- control strategies
- machine learning
- model free
- action selection
- dynamic programming
- bayesian networks
- learning algorithm
- control system
- multistage
- markov decision process
- finite horizon