In this paper we propose and assess the accuracy of new fuel consumption estimation models for vehicle routing. Based on real-world shipping operation data consisting of instantaneous fuel consumption, time-varying speeds observations, and highfrequency traffic, we propose effective methods to estimate fuel consumption. By carrying out nonlinear regression analysis using supervised learning methods, namely Neural Networks, Support Vector Machines, Conditional Inference Trees, and Gradient Boosting Machines, we develop new models that provide better prediction accuracy than classical models. We correctly estimate consumption for time-dependent point-to-point routing under realistic conditions taking into account freight transportation operations during peak hour traffic congestion, stop-and-go driving patterns, idle vehicle states, and the variation of vehicle loads. Our methods provide a more precise alternative to classical regression methods used in the literature, as they are developed for a specific situation (fleet, drivers, geography, etc). Extensive computational experiments under realistic conditions show the effectiveness of the proposed machine learning consumption models, clearly outperforming macroscopic and microscopic consumption models such as the Comprehensive Modal Emissions Model (CMEM) and the Methodology for Estimating air pollutant Emissions from Transport (MEET). Based on sensitivity analyses we show that MEET underestimates real-world consumption by 24.94% and CMEM leads to an overestimation of consumption by 7.57% with optimized parameters. Our best machine learning model (Gradient Boosting Machines) exhibited superior estimation accuracy with a gap of only 1.70%.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.