This study presents an integrated optimization framework for the hierarchical pallet and vehicle loading problem arising in real export logistics operations. The problem involves assigning heterogeneous boxes to compatible pallet types under strict type-compatibility constraints, packing each pallet in three dimensions, arranging the loaded pallets within a single vehicle, and selecting the most economical vehicle from an available fleet. A Grouping Genetic Algorithm (GGA) combined with an Extreme Point (EP) decode heuristic is proposed to address all decision levels within a single framework. Each chromosome encodes a variable-length sequence of pallet groups rather than individual item assignments, preserving well-formed pallet structures across generations and naturally enforcing type-compatibility constraints. Four fitness-guided mutation operators targeting volumetric fill ratio imbalances restructure weak pallet groups during the search. The EP heuristic decodes each chromosome into a geometrically feasible loading plan at both the pallet and vehicle levels, with vehicle type selection resolved automatically at the end of each evaluation. The framework was implemented in C++ with OpenMP parallelism and evaluated on eight real shipment instances obtained from TP Electric Electrical Materials Industry and Trade Inc., a manufacturer based in İzmir, Türkiye. The proposed algorithm matched or improved the company's manual loading plan in seven out of eight cases, achieving an average pallet count reduction of 12.5% and an average lower bound gap of 22.8%, compared to 47.3% for the company baseline. These results confirm the practical applicability of the proposed approach for real-world heterogeneous pallet and vehicle loading decisions.

Keywords

Grouping Genetic Algorithm, Three-Dimensional Bin Packing, Extreme Point Heuristic, Pallet Loading, Vehicle Loading Optimization