Using an agent-based neural-network computational model to improve product routing in a logistics facility

This study tests whether a simplified neural-network computational model can make routing decisions in a logistics facility more efficiently than five ׳intelligent׳ routing heuristics from the logistics literature. The experiment uses a real-world simulation scenario based on the Hamburg Harbor Car Terminal, a logistic site faced with managing approximately 46,500 car-routing decisions on a yearly basis. The simulation environment has been built based on a data set provided by the Terminal operator to reflect a real-world case. The simulation results show that the percent-improvement of the neural-net model׳s performance is 48% better than that of the best routing heuristic tested in previous studies. To test the applicability of the method with more complex logistic scenarios, we relaxed the sequence constraints for routing in a subsequent simulations study. If logistic complexity in terms of more freedom in decision-making is increased, the neural net model׳s percent-improvement performance of routing decisions is around three times better than the best-performing heuristic.