The influence of ADAS on road safety: An analysis of UK crashes

The effects of Advanced Driver Assistance Systems (ADAS) on road safety are still largely unexplored. Most of previous research is focused on the influence on driving behaviour, while studies in which crash data are directly analyzed often disregard the effects of ADAS, mainly because this information is not present in the dataset. Based on this background, the present study started from the United Kingdom dataset about single-vehicle crashes and used both traditional statistical approaches (binary logistic regression) and machine learning techniques (Random Forest) to explore the possible influence of the ADAS on different crash types. Information about vehicles involved in the crashes were manually post-processed to derive information about the possible presence of ADAS on the crashed vehicles. The most recurrent single-vehicle crashes were identified and divided into classes for modelling purposes. Possible correlations of some crash types with ADAS were hypothesized, even if their overall effect seems lower than that of other variables. However, some of these highlighted tendencies suggest pushing forward research to improve the ADAS performance.