On-Board Diagnostic of the Motor Vehicle through Smartphone

Abstract

On-board car sensors enable continuous driving performance assessment and car health monitoring and tracking. Such sensors provide data related to driving styles, such as speed, pedal positions and injection, and information related to the overall functioning of the engine and other car subsystems, whose continuous monitoring and logging are precious in car maintenance. Indeed, if provided to the driver, for instance, within the car dashboard, these parameters allow the driving style adaptation in terms of optimal goals, such as reducing gasoline consumption. Furthermore, provided to auto mechanics, such data enriches the set of information on which diagnostic and maintenance activities are based, enabling better services and actions.The goal of this paper moves from the observation that sometimes, expert mechanics can successfully predict diagnostic outcomes based on driving tests and without any sensors data inspection: it appears that they may infer the car health status based only on the human senses of touch and hearing. In this study, we investigate this ability to favour its exploitability in innovative car diagnostic AI systems. In the paper we show that an expert system, based only on smartphone data, can predict with good accuracy most car on-board data. This opens to innovative and new applications in non-intrusive car diagnostic systems for car and driver safety.