Constrained Variational Policy Optimization for Safe Reinforcement Learning.
Zuxin LiuZhepeng CenVladislav IsenbaevWei LiuSteven WuBo LiDing ZhaoPublished in: ICML (2022)
Keyphrases
- reinforcement learning
- optimal policy
- concave convex procedure
- policy search
- action selection
- global optimization
- image segmentation
- optimization algorithm
- control policy
- state space
- state and action spaces
- optimization process
- markov decision problems
- partially observable
- action space
- function approximation
- control policies
- optimization problems
- optimization method
- actor critic
- function approximators
- constrained optimization
- markov decision processes
- markov decision process
- learning algorithm
- long run
- policy gradient
- policy evaluation
- approximate dynamic programming
- dynamic programming
- partially observable domains
- sufficient conditions