Iranian Geography and Urban Planning Association Sustainable city 2476-6631 Articles in Press 2026 08 09 Intelligent Multi-Objective Optimization of Urban Waste Collection Routes Using Machine Learning and the NSGA-II Algorithm (Case Study: Tabriz City) Intelligent Multi-Objective Optimization of Urban Waste Collection Routes Using Machine Learning and the NSGA-II Algorithm (Case Study: Tabriz City) 250447 10.22034/jsc.2026.528909.1853 FA Masoud Haghlesan Assistant Professor, Department of Architecture and Urban Planning, Ilk.C., Islamic Azad University, Ilkhchi, Iran. Mehrdad Najafi Ph.D. Student in Urban Planning, Department of urban planning, Ta.C., Islamic Azad University, Tabriz, Iran 0009-0004-0530-4107 MohammadSadegh Jedari Sefidgari Ph.D. Student in Urban Planning, Department of urban planning, Ta.C., Islamic Azad University, Tabriz, Iran 0009-0000-9790-8012 Journal Article 2025 06 18 Abstarct<br /> Rapid population growth and the physical expansion of metropolitan areas, efficient urban waste management has become a critical environmental, economic, and administrative challenge. Among its components, the collection and transportation of waste account for a significant portion of operational costs and energy consumption. Inefficient route allocation for waste collection leads to resource misallocation, increased fuel usage, environmental pollution, and disruption in municipal services. This study aims to propose an intelligent and realistic multi-objective framework for optimizing the routing of municipal waste collection fleets in Tabriz, Iran.<br /> To achieve this, the urban transportation network was extracted from OpenStreetMap data, with the geographic positions of waste bins and disposal sites precisely identified. A robust database incorporating diverse route allocation scenarios was then compiled to support decision modeling. Using multilayer perceptron neural networks, a predictive model was developed to estimate operational costs, including total distance, travel time, and fuel consumption. Subsequently, the NSGA-II algorithm was applied to simultaneously optimize the three objective functions, generating a set of non-dominated solutions in the form of a Pareto front.<br /> Empirical results demonstrated that the proposed neural network model achieved high predictive accuracy (with R² values exceeding 0.96 for some metrics), and the use of NSGA-II led to substantial improvements in operational performance: fuel consumption was reduced by up to 43%, collection time by 35%, and total traveled distance by 35%. This research underscores that combining machine learning with evolutionary optimization can offer an effective solution for intelligent and sustainable urban waste management in large cities.. Abstarct<br /> Rapid population growth and the physical expansion of metropolitan areas, efficient urban waste management has become a critical environmental, economic, and administrative challenge. Among its components, the collection and transportation of waste account for a significant portion of operational costs and energy consumption. Inefficient route allocation for waste collection leads to resource misallocation, increased fuel usage, environmental pollution, and disruption in municipal services. This study aims to propose an intelligent and realistic multi-objective framework for optimizing the routing of municipal waste collection fleets in Tabriz, Iran.<br /> To achieve this, the urban transportation network was extracted from OpenStreetMap data, with the geographic positions of waste bins and disposal sites precisely identified. A robust database incorporating diverse route allocation scenarios was then compiled to support decision modeling. Using multilayer perceptron neural networks, a predictive model was developed to estimate operational costs, including total distance, travel time, and fuel consumption. Subsequently, the NSGA-II algorithm was applied to simultaneously optimize the three objective functions, generating a set of non-dominated solutions in the form of a Pareto front.<br /> Empirical results demonstrated that the proposed neural network model achieved high predictive accuracy (with R² values exceeding 0.96 for some metrics), and the use of NSGA-II led to substantial improvements in operational performance: fuel consumption was reduced by up to 43%, collection time by 35%, and total traveled distance by 35%. This research underscores that combining machine learning with evolutionary optimization can offer an effective solution for intelligent and sustainable urban waste management in large cities..