Driving behavior at a roundabout: A hierarchical Bayesian regression analysis

The paper models and compares driving behavior and vehicular emissions at a roundabout. Four drivers drove a vehicle instrumented with a GPS data logger over a study route. Second-by-second vehicle positions were recorded for various runs. Speed profiles of drivers were modeled using a Bayesian inference methodology. Circulating speed and maximum accelerations were simulated from the speed profile models and were compared across drivers. In addition, vehicular emissions were estimated using past experimental data. It is found that speed profiles differ significantly across drivers, as do the mean speeds at the circulating path of the roundabout. Acceleration events correspond to significantly higher emissions since during acceleration more than required fuel is injected into the combustion chamber of the engine. An emissions hotspot was defined as group of consecutive locations on the route where the sum of absolute values of acceleration was more than 95 percentile. Emissions at these hotspots were more than 25% of the emissions for a given speed profile.