Tracking and Evaluation of Human State Detections in Adaptive Autonomous Vehicles.

Autonomous Vehicles (AVs) take away the control of their passengers, which could result in uncomfortable or stressful situations through inappropriate driving styles, reducing the overall acceptance of AVs [1]. The detection of such situations through sensory measurements could improve driving quality through intelligent driving style adaptations, leading to an overall more personalized and satisfying driving experience [2, 5]. Such adaptive autonomous vehicles (AAVs) promise great increase of trust and better quality of human machine interaction, leading to more acceptance and higher safety in driving. We propose a software framework that allows an easy recording of various sensor measurements, statistical analysis of the data and generating easy to understand user interfaces. Since the sensory data is collected at time scales of milliseconds and changes in the user and vehicle behavior can stretch over multiple months, an efficient analyses and transparent representation of the recorded data is crucial. Enabling the passenger to see how well the vehicle adapted to his or her needs can further increase trust, while providing in depths analysis of long-term changes to the manufacturer can help to track and improve the vehicles adaption models. The graphical user interface (GUI) allows passengers to access comprehensive metrics, the history of the recorded data and to make changes to the car’s behavior.