Driver unique acceleration behaviours and stability over two years
The identification of characteristic individual driving behaviours is an emerging challenge that occurs within longitudinal studies of drivers to distinguish different drivers of a shared vehicle. It also has application in the insurance industry where insurance risk and associated owner premium depends on the diversity or lack thereof of drivers for a vehicle such as a vehicle driven/never driven by secondary drivers that have higher risk driving behaviours. Lastly, emerging self driving vehicles could allow the owner to personalize the vehicle behaviour to drive more like them increasing owner acceptance of the technology. In this paper, a big data set of driving data for 14 drivers is analyzed - a single year of data includes over 250,000 km and almost 5000 hours of driving for the 14 drivers. Analytics methods are presented that identify acceleration events within the data for the drivers and it then proposes a two-phase relationship model for these events that is indicative of unique drivers' behaviour. The results show that the two-phase acceleration relationship for maximum and mean acceleration allows 84.6% and 80.2% of the 91 driver pairs that can be formed from the 14 drivers to be distinguished (p<5%). The paper shows the stability of two-phase acceleration and deceleration relationships for the 14 drivers as the second year of events for each of the 14 drivers have a mean correlation with the first year relationships of 0.971 or higher.
|Keywords||Acceleration, Big data, Data analytics, Driving signature, Finite difference equations, Global Positioning System (GPS)|
|Conference||5th IEEE International Congress on Big Data, BigData Congress 2016|
Wallace, B. (Bruce), Goubran, R, Knoefel, F. (Frank), Marshall, S. (Shawn), Porter, M. (Michelle), & Smith, A. (Andrew). (2016). Driver unique acceleration behaviours and stability over two years. In Proceedings - 2016 IEEE International Congress on Big Data, BigData Congress 2016 (pp. 230–235). doi:10.1109/BigDataCongress.2016.36