Application of the Elitist Ant System and Clustering Methods for Solving the Constrained Vehicle Routing Problem

The rapid growth of e-commerce necessitates automated Vehicle Routing Problem (VRP) solutions that account for real urban topology, particularly in metropolises such as Almaty. This study implements a hybrid «Cluster-First, Route-Second» approach. This approach integrates K-Means clustering and Ant Colony Optimization (ACO) using real road distance matrices (OSRM API) instead of Euclidean metrics. Comparative experiments on verified geodata demonstrated the superiority of the method over the Genetic Algorithm. A 42% reduction in route length and the elimination of topological errors were achieved. Decomposition analysis confirmed the optimality of K-Means for driver workload balancing. Practical application on a fleet of 4 vehicles showed the potential to reduce annual mileage by over 31,000 km. This provides financial savings exceeding 1.44 million tenge and a reduction in CO₂ emissions of 11.8 tons, offering a validated tool for sustainable urban logistics.