A group of researchers from the High School of Telecommunications Engineering of the Universidad Politécnica de Madrid (UPM) has shown that is possible to detect an aggressive driving behavior by just monitoring external driving signals such as the speed or acceleration, since the aggressiveness works as a linear filter on these signals. This model was validated with empirical data under real driving situations with rates exceeding 92%.
Traffic accidents annually produce 1.3 million fatalities and 50 million injuries worldwide. They are currently the main cause of death among young adults between 15–29 years old. Apart from these deaths, traffic accidents cause high economic cost due to property damage and medical expenses, amongst others, estimated at around half a billion Euros per year. The 70% of these accidents are related to human factors and reckless driving which is the result of an excessive or inappropriate speed.
Boosting new behavior patterns could prevent a significant percentage of accidents. One way to promote these behavior changes of drivers it can be the monitoring and characterization of their driving in order to detect inappropriate driving situations and early warn at risk of accident.
The early research works in this field were mainly based on the characterization through intrusive methods by monitoring physiological signals such as heart rate, breathing and stress level. These methods, though effective, are less appropriate because can produce discomfort to the driver and represent an additional cause of distraction.
The question is, can we efficiently characterize the driving through non intrusive methods and that are not perceived by the driver? A group of researchers for the Applied Mathematics for Information Technology and Communications Department of the High School of Telecommunications Engineering of UM has found the answer. Thanks to this research, they have shown that is possible to detect a reckless driving behavior by just monitoring driving external signals, such as velocity and acceleration.
The key is that aggressiveness performs like a linear filter on these signals: scaling its probability distribution and modifying its average value, standard deviation and dynamic range.
The model was empirically verified under real driving situations at UPM facilities. They tested its validity and its generality for diverse driving signals, different drivers and types of roads. The results showed a success rate of 92% for detecting reckless driving behavior from driving signals.
In the future, this real time reckless driving system for early detection will be integrated into smartphones, significantly contributing to increase safety on the roads.