Machine learning multi-objective optimization for time-dependent green vehicle routing problem

This study explores the Multi-Objective Time-Dependent Green Vehicle Routing Problem (MOTDGVRP) within the framework of energy economics, focusing on optimizing fuel consumption and transportation efficiency. Our model minimizes total distance, transportation time, and fuel consumption, all of which are critical for reducing energy costs and environmental impact. We introduce methodologies for calculating travel time and fuel consumption, considering time-dependent speed variations across different periods. Key factors influencing fuel consumption include vehicle load, dynamic traffic speeds, and distance traveled, reflecting real-world energy use in logistics. Given the NP-hard nature of the problem, we employ Non-dominated Sorting Genetic Algorithm 2 (NSGA-2) and an NSGA-2 enhanced with Machine Learning (MLNSGA-2) to optimize routing decisions. The originality of this study lies in the integration of machine learning (ML) in vehicle routing optimization,which enhances solution quality and accelerates computational performance. While ML applications in routing are growing, their use in Vehicle Routing related models remains novel. Additionally, the model accounts for time-dependent speed variations, addressing real-world traffic dynamics that significantly impact fuel consumption and delivery efficiency. The combination of ML-enhanced optimization with time-sensitive routing presents a new approach to energy-efficient transportation. From an energy economics perspective, our findings provide valuable insights for optimizing energy use in logistics, reducing operational costs, and promoting sustainable transportation. The integration of machine learning-driven optimization offers a scalable method for enhancing energy efficiency in supply chains, contributing to both economic and environmental objectives.