A Probabilistic Method to Estimate Emissions and Fuel Consumption using Connected Vehicle Data from a Mixed Fleet1

Abstract

Real-time estimation of emissions using microscopic emission models are not trivial even though speed-acceleration data can be inferred from connected vehicles (CVs). This is because models for estimating emissions/fuel require information on vehicle category (which may vary by models) and some models also require soak times (the duration for which the vehicle was at rest before the current trip). Such data are not transmitted by CVs. This study developed and demonstrated an approach for the real-time estimation of emissions/fuel consumption (at a location) from speed profiles from a limited number of CVs in a traffic stream and limited information about CVs (i.e., missing details about vehicle categories and soak times). The broad approach involved applying a microscopic emission model on a large dataset of speed-acceleration profiles of vehicles of multiple types and trips with different soak times obtained from travel surveys. These were then used to create aggregated probability distribution profiles for various speed-acceleration “bins”. The estimated profiles by bins and by vehicle type are then applied on speed-profile that may be typically received by roadside units (RSUs). Using real world trajectory data from Atlanta, the application of the approach was demonstrated for different market penetration levels of CVs.