Safe Reinforcement Learning through Phasic Safety-Oriented Policy Optimization.
Sumanta DeyPallab DasguptaSoumyajit DeyPublished in: SafeAI@AAAI (2023)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- action selection
- optimization algorithm
- machine learning
- markov decision process
- optimization problems
- partially observable environments
- markov decision processes
- optimization method
- global optimization
- state space
- state dependent
- multi objective
- learning algorithm
- actor critic
- policy gradient
- state action
- control policy
- policy iteration
- control problems
- optimization model
- partially observable markov decision processes
- sufficient conditions
- partially observable
- action space
- reward function
- average reward
- function approximation
- reinforcement learning methods
- markov decision problems
- evaluation function
- supply chain
- continuous state spaces
- learning process
- least squares