The growing penetration of telematics systems and connectivity in vehicles has enabled a large variety of possible value-added services for drivers and service providers. In this project we focus on utilizing this abundance of data to model how drivers behave in various driving contexts. We use the state of the art Machine Learning algorithms to build models that can serve different applications with the focus on only using vehicle sensors while avoiding the usage of intrusive sensors such as in-car cameras. For example by modelling distinct driving patterns of each individual we can identify the driver, which has applications in detection of stolen vehicles. In another use-case we work on distraction detection as an safety application. This time the model captures changes between attentive and distracted driving therefore we can detect such events and possibly issue a warning to the driver. Lastly in a more general sense more comprehensive driver models can augment current simulation tools which play a crucial role in development of advanced driver assistance systems (ADAS) and autonomous vehicles.
Sasan Jafarnejad, German Castignani, Thomas Engel (2018). Revisiting Gaussian Mixture Models for Driver Identification. Accepted for presentation at IEEE ICVES.
Sasan Jafarnejad, German Castignani, Thomas Engel (2018). Non-intrusive Distraction Detection Based on Car Sensor Data. VEHITS.
Sasan Jafarnejad, German Castignani, Thomas Engel (2017). Towards a Real-Time Driver Identification Mechanism Based on Driving Sensing Data. IEEE ITSC.