Data driven analysis and forecasting of medium and heavy truck fuel consumption

Abstract

Fuel consumption for transport activities should be as low as possible for ecological and economic reasons. This said, there is no transparency as to which truck model behaves better for a given route and weight. Existing physical models are trained using only specific reference driving cycles. This contribution proposes to use real data from telematics systems in order to extract differences in the fuel consumption of various truck models. ML models are then developed to predict their fuel consumption. Finally, the prediction model is applied to a sample roundtrip and the predicted fuel consumption of different truck models is compared.