Digital system for dynamic container loading with neural network-based memory exploiting hybrid genetic algorithm for carbon reduction

Abstract

Global logistics are increasingly important owing to shifting paradigm by strategically focusing on core competences, outsourcing manufacturing to pursue higher value proposition in the supply chain. Container loading has become a bottleneck in global logistics, while container shipping costs are increasing due to the tensions of regional conflicts. Indeed, loading pooled shipments into containers involves combinatorial complexity, while practice highly relied on the experienced workers. Furthermore, inefficient use of containers will increase the total carbon footprint. To address realistic issues, this study aims to develop a framework that integrates the placement heuristic, hybrid genetic algorithm, and deep learning model to enhance the effectiveness of searching the optimal solutions. Deep learning is employed to improve local search by guiding the search direction for genetic algorithm. Furthermore, a digital decision system is developed to enhance the efficiency for dynamic container loading. The comparison of the results has validated practical viability of the proposed approach to significantly improve the search efficiency and enhance the container utilization for carbon reduction. Indeed, the developed solution is employed in this case company.