An AI-Driven Framework Combining K-Means Clustering and VRP Optimization for Sustainable Waste Collection

Abstract

Urban waste management in developing countries faces persistent challenges, including inefficient routing, high operational costs, and significant environmental impact. This study presents a hybrid framework that integrates K-means clustering with Capacitated Vehicle Routing Problem (CVRP) optimization to improve municipal waste collection efficiency within a reverse logistics perspective. Applied to the Technical Landfill Center (CET) of Guelma, Algeria, the model groups 23 urban sectors into operationally coherent clusters before optimizing collection routes. The results show a 63.6% reduction in fleet size, a 69.14% decrease in daily travel distance, and estimated annual CO2 savings of 381 metric tons, while maintaining full-service coverage. Built entirely on open-source tools, the proposed framework offers a computationally efficient and interpretable optimization approach, providing a scalable decision-support tool for sustainable city logistics in resource-constrained settings.