Internal validation of near-crashes in naturalistic driving studies: A continuous and multivariate approach

Road safety of passing maneuvers: A bivariate extreme value theory approach under non-stationary conditions

“…A major improvement of this study is that it links the probability of accident occurrence to the frequency of conflicts estimated from observed variability of accident proximity, using a probabilistic framework and without using accident records. The generic formulation of the application of EV to road safety analysis was then proposed by Tarko (2012) and it was only very recently applied to other accident types and data sets ( [10], [5]). This formulation relies usually, but not exclusively, on time-based surrogate measures and estimates the probability of accident occurrence using the EV fitted distribution of such measures (see Appendix Appendix A for formulation details).…”

“…Aiming at estimating the probability of a head-on collision for a single passing maneuver, the minimum TTC was considered as a head-on accident surrogate measure. The data was then filtered to account only for values smaller than 1.5s ( [4], [10], [28]), leading to a total of 463 observations. Knowing that 9 maneuvers ended with actual head-on collisions, the empirical probability of a head-on collision in a passing maneuver given that a critical TTC (i.e.…”

Modeling driver behavior in dilemma zones: A discrete/continuous formulation with selectivity bias corrections

“…Selectivity bias is present in studies involving individual drivers, due to the fact that only one decision path (the one actually selected by the driver) is able to be observed. Several prominent publications have shown that a failure to account for this bias in the data can have deleterious effects in any corresponding analysis or model output (Fisher, 2011;Jonasson and Rootzén, 2014). Furthermore, a very limited amount of research currently exists within the transportation realm that attempts to correct for these biases and other impacts of heterogeneity across observations in driving data (Bastin and Cirillo, 2010;Zhou and Lyles, 1997).…”

Large‐scale road safety evaluation using extreme value theory

“…In the latter, the BM approach was used to analyse the safety of signalised intersections, using data collected with video cameras in 18 locations for 8 h. Some years later, Zheng et al in 2014 [15,16] applied both BM and POT approaches to assess the safety of lane-change manoeuvres in freeways; collecting data with video cameras in 29 locations for ∼3 h. EVT applications to road safety involved the use of driving simulator to collect data, as in the case of Tarko [8], who investigated road departures with POT approach; Farah and Azevedo [3], who analysed head-on collision during passing manoeuvres with both BM and POT approaches; and lastly, Orsini et al [17], who dealt with entering-circulating collisions in roundabouts with both BM and POT approaches. There have also been examples of naturalistic driving experiments, as in the case of Jonasson and Rootzen [18], who analysed rear-end crashes and Asljung et al [19], who analysed the safety of autonomous vehicles.…”