Towards Safe Reinforcement Learning with a Safety Editor Policy.
Haonan YuWei XuHaichao ZhangPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- action selection
- markov decision process
- reinforcement learning problems
- markov decision processes
- reward function
- policy gradient
- actor critic
- function approximators
- reinforcement learning algorithms
- state and action spaces
- policy iteration
- policy evaluation
- function approximation
- partially observable environments
- control policy
- markov decision problems
- state space
- action space
- approximate dynamic programming
- partially observable
- model free
- average reward
- partially observable markov decision processes
- state dependent
- decision problems
- temporal difference
- multi agent
- infinite horizon
- dynamic programming
- rl algorithms
- control problems
- policy gradient methods
- inverse reinforcement learning
- partially observable domains
- supervised learning
- agent learns
- exploration exploitation tradeoff
- transition model
- approximate policy iteration
- model free reinforcement learning
- continuous state spaces
- control policies
- least squares
- learning process
- learning algorithm
- continuous state
- state action
- policy makers
- long run
- finite state
- monte carlo
- machine learning