Data Collection for Traffic and Drivers’ Behaviour Studies: A Large-scale Survey

Abstract

Studies of driving behaviour are of great help for different tasks in transportation engineering. These include data collection both for statistical analysis and for identification of driving models and estimation of modelling parameters (calibration). The data and models may be applied to different areas: i) road safety analysis; ii) microscopic models for traffic simulation, forecast and control; iii) control logics aimed at ADAS (Advanced Driving Assistance Systems). In this paper we present a large survey based on the naturalistic (on-the-road) observation of driving behaviour with a view to obtaining microscopic data for single vehicles on long road segments and for long time periods. Data are collected by means of an instrumented vehicle (IV), equipped with GPS, radar, cameras and other sensors. The behaviour of more than 100 drivers was observed by using the IV in active mode, that is by observing the kinematics imposed on the vehicle by the driver, as well as the kinematics with respect to neighbouring vehicles. Sensors were also mounted backwards on the IV, allowing the behaviour of the driver behind to be observed in passive mode. As the vehicle behind changes, the next is observed and within a short period of time the behaviour of several drivers can be examined, without the observed driver being aware. The paper presents the experiment by describing the road context, aims and experimental procedure. Statistics and initial insights are also presented based on the large amount of data collected (more than 8000km of observed trajectories and 120hours of driving in active mode). As an example of how to use the data directly, apart from calibration of driving behaviour models, indexes based on aggregate measures of safety are computed, presented and discussed.