This research is an explorative study to look for the potential to predict traffic density from driver behaviour using signals collected from the Controller Area Network (CAN) bus. The hypothesis is that driver behaviour is influenced by traffic density in such a way that an approximation of the traffic density can be determined from changes in the driver behaviour. Machine learning will be employed to correlate a selection of commonly available sensors on cars to the traffic density. Challenges in the processing of the data for this purpose will be outlined. The data for this study is collected from five passenger cars and nineteen trucks driving on the A28 highway in Utrecht region in the Netherlands. This study is restricted to straight roads in order to isolate the steering behaviour attributable to the traffic state influences rather than following the curve in the road. The results are encouraging that the correlation between driver behaviour and traffic density can be established. An overall accuracy of over 95% is achieved with a precision of 92%. The recall rate however is low most likely caused by over-fitting due to the unbalanced data set. The results still look promising and more training data should improve the results. This research is part of the broader project VIA NOVA which aims to investigate the use of car-sensor data for traffic and road asset management.

Congestion, Traffic Density, Driver Behavior, CAN Bus, Probe Vehicle Data, Supervised Machine Learning
dx.doi.org/10.1016/j.procs.2019.04.016
Procedia Computer Science
Stochastics

Koch, T. (2019). Predicting traffic phases from car sensor data using machine learning. In 10th International Conference on Ambient Systems, Networks and Technologies (Vol. 151, pp. 92–99). doi:10.1016/j.procs.2019.04.016