ABSTRACT
During the last decade, the interest for vehicular opportunities and potentialities has heavily grown, due to the numerous advantages given by the technological progress. In particular, road safety has been one of the major objectives to be reached. Road accidents represents the main cause of deaths each year: in most cases they are due to reckless driving styles employed by drivers (pedal pressures, steering speed, etc.). The real-time identification of potentially dangerous driving styles is an important element for road safety, as it gives the opportunity to take precautions in terms of safety distance, speed, etc. The main goal of this work is represented by the characterization of the different driving styles in various environments, with the possibility of highlighting potentially dangerous behaviors. Our approach is based on the deployment of a Smart-Device (SD, a phone, a tablet, etc.) to acquire, process data and perform the characterization (nowadays, at least one SD is present in a vehicle). The SD gives the possibility to acquire information from the Internet (e.g. weather data), from its own sensors (such as the gyroscope), from the built-in GPS and from the Electronic Control Unit (ECU) via On Board Diagnostics II (OBD-II) standard [1], [2], [3]. In particular, the SD allows to interface with the vehicle OBD-II via Bluetooth. There are many dedicated devices, like the KiWi
work, the term “driving style” refers to the way a driver lead the vehicle. In our work, we started from some existing studies [7], [8], about driving styles recognition and, then, we refined them by taking into account the characteristics of different road environments (urban, suburban or highway). It is important to observe that, in different environments with equal speed and acceleration, the same behavior can be considered in different ways. For example, a driver who travels at 35 km/h in urban environment is classified with normal behavior, but if the speed of 35 km/h is maintained on the highway, the behavior is no longer normal, but it can indicate the presence of traffic jams or abnormal situations. For this reason, it is not possible to use the same fuzzy speed sets in all environments (an analysis will be made for each case). After the implementation of the Android application, all the analyses have been carried out through the Fuzzy Logic and Statistical Toolboxes of MATLAB. This paper is organized as follows: Section II presents an in-depth overview on state-of-the-art of similar approaches in VANETs; Section III introduces the considered scenario, while Section IV offers a deep description of the proposed scheme. Section V validates the proposed scheme and, then, conclusions are summarized in the last section.
