A machine learning approach for predictive modelling of rolling resistance of passenger cars

In recent years, machine learning has extended its field of application to various domains owing to its superlative predictive capabilities. One example of that is its increasing use in the road pavement sector as a means to improve the prediction of pavement performance and enhance pavement sustainability. Framed by this context, this paper presents a machine learning framework for predicting rolling resistance (RR) and excess fuel consumption (FC) in passenger cars. RR plays a critical role in pavement engineering and transportation, considering its impact on vehicle energy consumption and, therefore, environmental and economic impacts. An instrumented passenger car equipped with multiple sensors was used to collect different types of data during a data collection campaign performed at the University Gustave Eiffel's test track. A multi-layer perceptron (MLP)-based FC model was subsequently developed based on the data collected and the pavement surface properties of the road pavement sections comprising the test track. The model developed showed good predictive capabilities, as evidenced by the R-squared value of 0.89. The outcomes of this study can empower the road pavement community to take a significant step towards the achievement of environmental sustainability goals related to the road pavement life cycle management.