GreNet based Industrial Line Resource Scheduling for Flexible Manufacturing

With the development of the Industrial Internet of Things (IIoT), effective resource scheduling is crucial for enhancing production efficiency and competitiveness. However, existing scheduling methods face challenges in flexibility, production efficiency, and environmental impact. To this end, this study proposes a novel method called GreNet that combines graph attention networks (GAT) and deep reinforcement learning (DRL) to comprehensively learn scheduling strategies for green flexible manufacturing lines. The GreNet minimizes maximum makespan, cumulative energy consumption, and costs by allocating tasks to appropriate equipment and automated guided vehicles (AGVs). It utilizes heterogeneous graph representations to capture the dynamics and complexity of the environment. A three-stage heterogeneous graph neural network is designed to extract features. Proximal policy optimization (PPO) is employed to balance stability and high returns. Experimental results demonstrate that the proposed method outperforms traditional scheduling rules and exhibits good generalization ability.