Policy Bifurcation in Safe Reinforcement Learning.
Wenjun ZouYao LyuJie LiYujie YangShengbo Eben LiJingliang DuanXianyuan ZhanJingjing LiuYaqin ZhangKeqiang LiPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- action selection
- markov decision process
- reinforcement learning algorithms
- partially observable environments
- reinforcement learning problems
- state space
- function approximators
- markov decision processes
- partially observable
- action space
- policy iteration
- function approximation
- actor critic
- approximate dynamic programming
- model free
- rl algorithms
- reward function
- policy gradient
- decision problems
- markov decision problems
- average reward
- control policy
- policy evaluation
- state and action spaces
- state dependent
- partially observable markov decision processes
- control policies
- agent learns
- state action
- temporal difference
- infinite horizon
- robotic control
- policy making
- supervised learning
- multi agent
- policy gradient methods
- learning algorithm
- approximate policy iteration
- partially observable domains
- control problems
- average cost
- long run
- piecewise linear
- partially observable markov decision process
- continuous state
- temporal difference learning
- monte carlo
- markov chain
- neural network
- model free reinforcement learning