Reinforcement learning with distance-based incentive/penalty (DIP) updates for highly constrained industrial control systems.
Hyungjun ParkDaiki MinJong-hyun RyuDong Gu ChoiPublished in: CoRR (2020)
Keyphrases
- highly constrained
- reinforcement learning
- control system
- industrial automation
- profit sharing
- state space
- industrial applications
- function approximation
- constrained problems
- temporal difference
- multi agent
- markov decision processes
- reinforcement learning algorithms
- outlier detection
- penalty function
- learning algorithm
- critical infrastructure
- distance measure
- optimal policy
- closed loop
- model free
- quality improvement
- control policies
- dynamic programming
- industrial processes
- objective function
- constraint satisfaction
- game theory
- optimal control
- machine learning