Balance Reward and Safety Optimization for Safe Reinforcement Learning: A Perspective of Gradient Manipulation.
Shangding GuBilgehan SelYuhao DingLu WangQingwei LinMing JinAlois KnollPublished in: AAAI (2024)
Keyphrases
- reinforcement learning
- policy gradient
- function approximation
- viewpoint
- optimization algorithm
- state space
- learning algorithm
- eligibility traces
- optimization problems
- reinforcement learning algorithms
- markov decision processes
- learning process
- optimal policy
- global optimization
- action selection
- multi agent
- optimization method
- optimization process
- constrained optimization
- model free
- reward function
- machine learning
- reinforcement learning methods
- state action
- learning capabilities
- temporal difference
- dynamic programming