A Hybrid Genetic Algorithm and Multi-Agent Reinforcement Learning Framework for Optimized Supply Chain Management Using Edge-Enabled Digital Twins

In order to cope with the growing complexity and dynamism of contemporary supply chain management (SCM), this paper suggests a hybrid optimization model that combines Genetic Algorithm (GA) and Multi-Agent Reinforcement Learning (MARL) in an Edge-Enabled Digital Twin (EEDT) setting. The framework uses GA to optimize strategic decisions globally, and MARL to adapt to changes at supply chain nodes in a decentralized manner, with the help of constant feedback provided by digital twins deployed at the edge. Experimental analysis with real-world and synthetic SCM data shows that the performance is greatly improved compared to the baseline algorithms, such as standalone GA, MARL, PSO, and ACO. In particular, the proposed model enhances fill rate by about 91%-96.5%, inventory turnover ratio by about 20% (improved by about 4.7 to 5.78), and stockout rates by as much as 4.3% (compared to higher baseline levels of more than 5.5%). Also, the transportation costs are decreased by approximately 10-15%, and the average delivery time is decreased in high traffic conditions (e.g., 78 minutes versus 83-85 minutes with the baseline methods). The framework also converges more quickly and more steadily, with a convergence rate of up to 99.8% and a better prediction accuracy with MAPE of 3.5% and a value of R2 of nearly 0.98. These findings indicate that the suggested GA-MARL-EEDT framework offers a scalable, adaptive, and cost-effective solution to next-generation intelligent supply chain optimization.