A. Benedetto, C. Benedetto, M.R. De Blasiis
This paper focuses on the studies and activities in the field of road safety and accident prediction carried out by the research group of the Department of Sciences of Civil Engineering Section Road Infrastructures. The group has been working on this topic since the eighties but only at the beginning of nineties the research activities reached meaningful results The new assumption formulated from the first investigations and demonstrated by the experimental studies declined the following criterion: traffic safety does not only involve kinetic and dynamical problems under the domain of road and mechanical engineering but also a multidisciplinary approach on the basis of the following assumptions: (1) traffic flow and mobility motivations influence risk thresholds and driving behaviours, (2) variability of environmental conditions affects safety standards, (3) human factors, perception reaction mechanism, decision making processes impact on driving manoeuvres and finally on safety. Traditional approaches have many points of weakness connected to the inadequacy of the accident database and to the low reliability of statistical methods for the diagnosis of the cause. Techniques based on artificial intelligence have been investigated. Neural networks are advanced mathematical models showing real and promising results, but their applicability is not easy and absolutely not efficient. Hazard analysis approach has given important, very applicable and reliable results.
Keywords: road safety; neural networks; accident prediction; safety standard; driving simulation; human factors; road safety performance