Reproduce “A Negotiation-Based Multi-Agent Reinforcement Learning Approach for Dynamic Scheduling of Reconfigurable Manufacturing Systems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Manonmani Sekar, Nasim Nezamoddini
A negotiation-based Multi-Agent Reinforcement Learning framework, integrating enhanced Deep Q-Networks, provides an intelligent solution for dynamic scheduling in Reconfigurable Manufacturing Systems. The approach demonstrably improves efficiency and robustness, notably reducing makespan by 15.5% and tardiness by 17.4% during machine breakdowns compared to reconfiguration-only strategies, while also cutting setup times by 74.8%.
Key insights
- The synergistic combination of a learning-based adaptive scheduling framework with an explicit negotiation mechanism significantly outperforms approaches relying solely on reconfiguration or negotiation in dynamic RMS environments.
- Advanced DQN enhancements, including attention mechanisms and prioritized experience replay, are crucial for achieving stable convergence and superior performance in complex, high-dimensional manufacturing scheduling problems.
- Implementing machine-centric agents, guided by a negotiation protocol, enables effective decentralized decision-making and resource allocation, which is particularly robust against unforeseen disruptions like machine breakdowns.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.