Appending GPS traces with road grade data to estimate driver behaviour risk

Abstract

Road gradients have various impacts when it comes to the performance of a vehicle and its driver. Literature shows how steep grades can cause an increase in emission rates and accident rates, but the actual behaviour and risk of a driver on different road grades are neglected. To find reliable road grade values to use with a behavioural model, this dissertation proposes a smoothing method to be used on the elevation pro le of a freely available Digital Elevation Model, namely the 1 Arc second Shuttle Radar Topography Mission dataset.

A behavioural model from literature is used to determine the risk of the drivers within an accelerometer dataset from 45 trucks for a day's travel, using all the records and not only those that fall within certain thresholds of harsh events. The model also allows for road grade to be added as a contextual variable, where records are filtered on fi ve different grade categories ranging from steep downhill to steep uphill roads. These categories each have their own risk space and ranking of driving performance. The majority of drivers have relative constant rankings, but the results show that some drivers behave differently with changes in gradient. Therefore, road grade can influence the behaviour of drivers and can be a useful addition to de fining and understanding the risk of a driver.