A Multi-Agent System for Operating Hyperconnected Freight Transportation in the Physical Internet

In the emerging Physical Internet, hyperconnected freight transportation allows using multiple short-haul drivers through hyperconnected meshed hub networks to deliver long-haul freight in multiparty multimodal relay mode. One of its core challenges lies in dynamic operational planning associated with the flows of freight, containers, carriers, and drivers to ensure agile, efficient, resilient, and sustainable performances given demand knowledge and state monitoring. In this paper, we focus on large-scale hyperconnected freight transportation, with testbed as the road-based tractor-hauler delivery of vehicles in the Southeastern USA. We propose an autonomous operating system for dynamic operational decision making from sources through hubs to destinations, including freight shipment consolidation, assignment of freight to haulers and haulers to tractors, multi-hop routing, coordination and scheduling of drivers, tractors and haulers, as well as driver-task matching based on individual prefrences. The system is based on a multi-agent network architecture, where each agent is responsible for different key operational decisions by running heuristic optimization algorithms and interacts with one another through information exchange. We provide preliminary results from simulation-based experiments encompassing vehicle delivery of a major automotive manufacturer from multiple sources, including factories, ports, and railheads, to 500+ dealers across the Southeastern United States. The analysis of results proves that each of agents can effectively and efficiently cope with the large-scale decision making and the overall proposed system is capable of operating hyperconnected freight transportation with reliable order-to-delivery performance, high hauler filling rates, and desirable environemnt friendliness, while improving quality of life by enabling truckers to be back home daily and alleviate truck driver shortage issues.