Risk-Aware Multi-Capacity Routing for Shift-Feasible Cosmetic Distribution

Abstract

Urban cosmetic distribution requires routing decisions that satisfy delivery time windows, vehicle capacity, and working-hour feasibility. Conventional routing models often rely on deterministic travel-time assumptions and single-capacity constraints, which may hide lateness risk and volume overload. This study proposes a risk-aware multi-capacity routing framework for shift-feasible cosmetic distribution. Customer and depot coordinates were used to estimate geographical distance using the Haversine formula, adjusted by a circuity factor, and converted into travel time using a baseline speed. The model applies split-delivery preprocessing for oversized customer-date records, enforces weight and volume capacity, and distinguishes delivery trips from physical vehicle requirements within a 09:00–17:00 working shift. Three routing models were evaluated: deterministic single-capacity routing, distance-based multi-capacity routing, and risk-aware multi-capacity routing. The results show that six oversized records were split into seven additional delivery units, eliminating unserved and infeasible customers. The single-capacity baseline produced nine over-volume cases, with maximum volume utilization reaching 242.15%. The multi-capacity model eliminated all capacity violations while maintaining comparable lateness performance. The risk-aware model reduced late customers by 13.65%, lateness rate by 14.17%, and total lateness by 5.17% compared with the baseline, but required 11.46% more delivery trips and 10.42% more physical vehicles. These findings indicate that the proposed framework improves delivery reliability while revealing the trade-off between service performance and resource usage.