On the design of virtual sensors using vehicular sensor data

Abstract

Vehicles' internal control systems rely on physical sensors to acquire data about their operation and malfunctions. Nowadays, manufacturers invest in better and more diverse sensors, however, there are aspects still unmonitored. Virtual sensors were proposed to replace failing physical sensors, but their applications have grown to include new aspects of a vehicle. The number of high-quality sensors available in modern cars enabled the creation of virtual sensors that extract new information from raw data.This work presents the design of vehicular virtual sensors, which is divided into three stages, namely, collection, operation and presentation of new data. To detail this process, in the first stage a data collection and its preprocessing are presented. The prepared data is then fed into operators, divided into four categories with proofs of concept that show their behavior. Finally, to illustrate the presentation stage, three examples of virtual sensors are provided.