Multi-Agent Communication for Dynamic Job-Shop Scheduling: A Robust Single-Machine Scheduling Model With Genetic Algorithm Optimization

Job-shop scheduling is crucial in manufacturing systems but struggles with uncertainties like random job arrivals and disruptions. Traditional scheduling methods often rely on centralized control and predefined rules, limiting flexibility in dynamic environments. This study has proposed a multi-agent communication based on the single-machine scheduling model for dynamic job-shop scheduling. In this model, each machine is considered an autonomous agent who can communicate with other agents to produce a robust local schedule against uncertain job arrivals, processing times, and disruptive events. When an agent becomes idle, and its queue length exceeds 1, it needs to decide which waiting job should be processed next. The proposed communication process allows the decision-making agent and other agents to exchange the intention of their local schedules. Accordingly, the decision-making agent can decide on the next processed job. Each agent schedules its local queue using the single-machine scheduling model, which aims to reduce local tardiness costs. Decisional entities accept extra costs if there is an inconsistency between the start and completion times of two consecutive operations. The genetic algorithm is employed to solve the single-machine scheduling models. For experimentation, a multi-agent job shop simulation program will be developed to implement the proposed communication process. The experimental results show that the proposed method outperforms the stand-alone single-machine scheduling model, achieving up to a 34% reduction in total tardiness cost and improving lead times and scheduling stability under uncertainties. Furthermore, the proposed method requires minimal computation time, making it potentially applicable to IoT-enabled manufacturing systems.