An Agent for Collaborative Optimization of Route Planning and Three-Dimensional Loading in Land Logistics
TL;DR
This study sets out to build an intelligent agent that plans vehicle routes and handles three-dimensional cargo loading at the same time and suggests that tying routing and loading together in a closed-loop setup makes plans easier to carry out.
Abstract
Cross-border trade is growing fast. The same is true for urban distribution, e-commerce fulfillment, port transfer, and cold-chain transportation. Because of that, land-logistics decisions need to be workable in practice and flexible enough to adjust, especially in heavy-traffic places like Singapore. This study sets out to build an intelligent agent that plans vehicle routes and handles three-dimensional cargo loading at the same time. The agent treats routing as a capacitated vehicle routing problem with time windows. For small and medium cases, it turns the subproblems into quadratic unconstrained binary optimization so simulated-annealing or a quantum-inspired solver can be used. Bigger cases rely on heuristics and metaheuristics instead. The approach uses Solomon I1 insertion and 2-opt. Relocate is applied too. On top of that, it runs adaptive large neighborhood search and genetic algorithms. For cargo loading, the setup follows a three-dimensional bin-packing problem. An extreme-point heuristic checks cargo dimensions and weight. It considers orientation as well. The method accounts for stacking rules, non-overlap, load balance, and unloading-order constraints. It uses a feedback loop too. That loop sends the reasons for loading infeasibility back to the routing module, which then regroups customers or assigns a different vehicle. In a 50-customer routing case, Solomon I1 with local search found a feasible plan with nine vehicles and a total travel time of 129.0. Adaptive large neighborhood search kept the fleet at nine vehicles but cut total travel time to 92.0. This is about a 28.7% improvement. The results suggest that tying routing and loading together in a closed-loop setup makes plans easier to carry out. It improves vehicle utilization and delivery punctuality, and it makes the decisions easier to explain. This proposed agent offers a practical way to manage land-logistics optimization in one integrated system that can be adjusted on the fly.