Intelligence-Driven Three-Dimensional Route Loading for Resource-Efficient Freight Operations

Abstract Digital freight systems become resource-efficient when order, routing and loading data are converted into executable decisions. This study formulates customer-indivisible three-dimensional route loading for homogeneous vehicles as a lexicographic optimization problem that minimizes certified fleet size before route distance. An intelligence-driven hierarchical adaptive large-neighborhood search (H-ALNS) couples dual certified initialization, geographic reassignment, whole-vehicle closure, two-route exchange, three-route ejection, failure-class feedback and adaptive depth-first packing. Every accepted route is reconstructed at coordinate level and checked by a common validator. Sustainability is measured through vehicles required for fixed order demand and fleet intensity per customer and item, linking computational intelligence to resource efficiency rather than treating geometric fill as a proxy. Evaluation covers 600 benchmark instances spanning 20, 60 and 100 customers, 200 and 400 items, and 3, 10 and 100 box types. H-ALNS reduces the aggregate fleet from 9346 to 7915 vehicles, avoiding 1431 dispatches (15.31%). Vehicle demand falls in 340 instances and remains equal in 260, with no increase; positive aggregate reductions occur in all 18 strata and reach 22.30%. Fleet intensity decreases from 22.91 to 19.40 vehicles per 100 customers and from 51.92 to 43.97 vehicles per 1000 items. The results show that adaptive route restructuring produces resource gains when customer regrouping is certified as complete three-dimensional layouts. The resulting decision object—routes, item coordinates, a feasibility certificate and fleet-resource measures—supports resource-efficient freight planning in intelligent logistics systems.