Driving Behavior and Traffic Safety: An Acceleration-Based Safety Evaluation Procedure for Smartphones

Traffic safety and energy efficiency of vehicles are strictly related to driver’s behavior. The scientific literature has investigated on some specific dynamic parameters that, among the others, can be used as a measure of unsafe or aggressive driving style such as longitudinal and lateral acceleration of vehicle. Moreover, the use of modern mobile devices (smartphones and tablets), and their internal sensors (GPS receivers, three-axes accelerometers), allows road users to receive real time information and feedback that can be useful to increase awareness of drivers and promote safety. This paper focuses on the development of a prototype mobile application that can evaluate the grade of safety that drivers are keeping on the road by measuring of accelerations (longitudinal and lateral) and warning for users when it can be convenient to correct their driving style. The aggressiveness is evaluated by plotting vehicle’s acceleration on a g-g diagram specially studied and designed, where horizontal and lateral acceleration is displayed inside areas of “Good Driving Style”. Several experimental tests were carried out with different drivers and cars in order to estimate the system accuracy and the usability of the application. This work is part of the wider research project M2M, Mobile to Mobility: Information and communication technology systems for road traffic safety (PON National Operational Program for Research and Competitiveness 2007-2013) which is based on the use of mobile sensor computing systems for giving real-time information in order to reduce risks and to make the transportation system more safe and comfortable.


Introduction
Traffic safety is strictly related to the behavior of each driver and to its individual variability associated to several parameters such as: age, gender, geographic locations, and other factors.Moreover, driving behavior has a great influence on energy efficiency: the difference in terms of fuel consumption and, consequently of gas emissions, between a safe (or calm) driver and an aggressive one is estimated to be higher than 40% (Alessandrini et al., 2012).For this reason, in order to reduce the environmental impact due to the road transportation system, in the last few years the idea of educating drivers to adopt an eco-friendly driving style has been promoted.
An eco-friendly and safe driving behavior could be achieved by reducing or avoiding sudden accelerations and rapid brakings in both longitudinal direction and cornering maneuvers (Yamakado et al., 2009).
The evaluation of real driving scenarios is very complex because it's reliant on many closely interconnected variables depending not only on the different type of drivers, but also on the road environment, the traffic characteristics and the categories of road infrastructure.However, driving behavior assessment has been associated to two main dynamic parameters that, among the others, have been proposed in scientific literature as the most significant for a quantitative evaluation of unsafe or aggressive driving style; these parameters are the longitudinal and the lateral accelerations and decelerations (Shaout & Bodenmiller, 2011;Klauer et al., 2009;Johnson et al., 2011;Paefgen et al., 2012).
By measuring the in-vehicle accelerations in the XY plane it is possible to categorize the driving behavior into two main classes: aggressive drivers and non-aggressive or safe and expert drivers.The driving style of these last ones, in particular, is characterized by the selection of smooth trajectories with a continuous adjustment of the

Results and Discussion
The behavior of each driver was evaluated by calculating the percentage of points that are characterized by values of accelerations higher than the limiting edges of the DSD area.In particular, the methodology proposed in this study aims to the identification of a specific threshoold of this percentage that allows to differentiate an aggressive driver from a safe one.In this way it is possible to use the proposed mobile application for collecting GPS data, calculating the vehicle's accelerations and giving real-time information to road users about their driving style.In order to reach this purpose the application is set for a systematic control of the accelerations for a specific period of time (i.e. 5 minutes); after this time the calculation restarts and new percentages of external points are estimated with a continuous feedback given to drivers.
The results obtained for the twenty tests carried out in this research are reported in Table 1, in which the drivers, the test sites and the percentage of accelerometer data which fall outside the DSD area are specified.The mean value and the standard deviation of the percentage of external points for each driver were also calculated (Table 1).
Slight differences were found in the estimated percentages for the two tests sites; in particular for the test site located in the city of Rende the aggressive driver is characterized by a mean value of external points of about 13.5, with a standard deviation of ±1.2%; the safe drivers registered a value of 5.7%.On the contrary, for the test site located in Montalto, registered values are (11.9±1.1)% for the aggressive drivers and (6.1±0.5)% for the safe ones.
The difference is probably related to the fact that the two tests site are quite different because the first one is characterized by high traffic volumes and it is located in the centre of the city, whereas the other one can be considered a suburban road (lower volumes of traffic, higher operative speeds, etc.).and non-aggressive behavior can be set to 9% of external points.Outcomes of this study are expected to benefit both practitioners and researchers.Moreover, further investigations on different types of roads with higher operative speeds are needed in order to take into account other factors related to the features of the travelled infrastructure.The results presented in this study represent the first prototype of the M2M project.At the end of the project Authors expect a much broader data base that will allow to validate and improve further the results presented in this study.

Table 1 .
Internal and external percentages of accelerometer data in the DSD (* indicates the aggressive driver)