Vehicle Dynamics Data Collection to Characterize the Drivers’ Behavior

Vehicle Dynamics Data Collection to Characterize the Drivers’ Behavior


Naude, IFSTTAR, Serre, IFSTTAR, Ledoux, Cerema


This paper aims to characterize the drivers' behavior with vehicle dynamics data recorded by EDRs. It is focused on the distribution of acceleration levels, the driver’s profile and the link between incidents and driving parameters of the travels.


The availability of data provided by in-board recorders is a great opportunity to acquire knowledge on drivers’ behavior. The objective of this paper is to show how additional data synthesizing the dynamic solicitations of the vehicle during the travels, based on the concept of the friction circle, help to characterize the behavior of the driver-vehicle couple. An experiment with 51 cars of public fleets equipped with a specific Event Data Recorder (EDR) was carried out in three regions of France during one year. 221 drivers were volunteers to take part to the experiment and 154 of them agreed to authenticate themselves with a badge associated to the following driver’s profile: age, year of driving license and gender. The EDRs recorded continuously data on driving behavior and manoeuvers performed by the vehicles as well as GPS trajectories. Data were used to detect critical driving situations called incidents (e.g. near-crash situation) but also to derive dynamic “syntheses” of all the travels. These syntheses aggregate the time spent in every longitudinal and lateral accelerations combination, with 0.1 g intervals, and they can be cumulated for several travels. 2D and 3D graphical representations of these dynamic parameters give a good estimation of the real use of the car by the drivers.
During the one-year experiment more than 100 000 km was travelled allowing recording 338 incidents and 3050 travel syntheses. This paper will focus on results that were derived from the analysis of these syntheses to study:

1) The distribution of acceleration levels
The cumulate accelerations of all the travels of this experiment are similar to the results of a reference experiment conducted in 1992 with the aim of acquiring knowledge on the driver-vehicle behavior in usual driving situations. Accelerations above 0.3 g are very rare, performed only by few drivers and are mostly linked to lateral acceleration.

2) The driver’s profile
Driving profiles of some volunteers are presented linked to the number of their incidents to illustrate the diversity of types of driving. “Behavior ratios” were calculated to quantify the actions of the drivers, the number of incidents per distance travelled, and the levels of accelerations endured by the vehicles. These ratios were calculated for each of the 43 authenticated drivers who drove equipped cars. A brief socio-demographic study was also carried out on the 28 authenticated drivers who travelled more than 250 km. These illustrative results show the relevance of using such data to differentiate the driving behavior of sporty vs calm drivers, men vs women, young vs old, and novice vs experienced drivers.

This work shows that data collection of aggregate accelerations is useful to characterize the drivers’ behavior, which is essential to better understand the occurrence of crashes, and to improve road safety. A similar algorithm applied to motorcycles was also developed in another project, based on a dedicated smartphone and using not only accelerations but also angular speeds, in order to study the specific behavior of motorcycle drivers.


Association for European Transport