Decentralized learning multi-agent system for online machine shop scheduling problem

Customer profiles have rapidly changed over the past few years, with products being requested more customization and lower demand. In addition to advances in technologies owing Industry 4.0, manufacturers explore autonomous smart factories. This paper proposes a decentralized multi-agent system (MAS), including intelligent agents that can respond their environment autonomously through learning capabilities, cope an online machine shop scheduling problem. proposed system, participate auctions receive jobs process, learn how bid for correctly, decide when start processing job. The objective is minimize mean weighted tardiness of all jobs. contrast existing literature, MAS assessed on its producing novel insights concerning what relevant learning, re-learning needed, response dynamic events (such as rush jobs, increase time, unavailability). Computational experiments also reveal outperformance other systems by at least 25% common dispatching rules tardiness, well performance measures.