"I did not see that coming": A latent variable structural equation model for understanding the effect of road predictability on crashes along horizontal curves.

Driver anticipation plays a crucial role in crashes along horizontal curves. Anticipation is related to road predictability and can be influenced by roadway geometric design. Therefore, it is essential to understand which geometric design elements can influence anticipation and cause the road to be (un)predictable. This exercise, however, is not straightforward because anticipation is individual-specific whereas road geometric design is location-specific; anticipation is latent and measuring it may not be trivial; anticipation may have several stages from the preceding tangent until the midst of the curve; and not all drivers anticipate in the same way and thus there may well be unobserved heterogeneity in the effect of anticipation on crash risk. Despite methodological advancements in crash risk modelling, there is no econometric model that can adequately explain the above complexities. This study aims to fill this gap by developing an econometric model with a new latent variable, named 'predictability' that is measured by individual-specific driving behaviour indicators and predicted by location-specific road geometric factors. The model is specified with random parameters to account for unobserved heterogeneity and is empirically tested by a unique dataset including detailed geometric design and driver behaviour data obtained for 156 curves in the Netherlands. Results indicate that higher exposure and uphill vertical grade are associated with increased likelihood of vehicle crashes along horizontal curves, whereas adequate superelevation and higher predictability are associated with decreased likelihood of those crashes. Pavement friction influences this likelihood too but it has varied effects. Road predictability is influenced by the differences in angle of horizontal curves, vertical grades, and width of consecutive road segments.

Towards common ethical and safe 'behaviour' standards for automated vehicles.

Automated vehicles (AVs) aim to dramatically improve traffic safety by reducing or eliminating human error, which remains the leading cause of road crashes. However, commonly accepted standards for the 'safe driving behaviour of machines' are pending and urgently needed. Unless a common understanding of safety as a design value is achieved, different manufacturers' driving styles may emerge, resulting in inconsistent, unpredictable and potentially unsafe 'behaviour' of AVs in certain situations. This paper aims to explore the main gaps and challenges towards establishing shared safety standards for the 'behaviour' of AVs, and contribute to their responsible traffic integration, by reviewing the state-of-the-art on AV safety in the core relevant disciplines: ethics of technology, safety science (engineering & human factors), and standardisation. The ethical and safety aspects investigated include the users' perception of AV safety, the ethical trade-offs in critical decision-making contexts, the pertinence of data-driven approaches for AVs to mimic human behaviour, and the responsibilities of various actors. Moreover, the paper reviews the current safety patterns, metrics (surrogate measures of safety - SMoS) and their thresholds introduced in existing research for three use cases: mixed traffic of AV and conventional vehicles, AV interaction with pedestrians and cyclists, and transition of control from machine to human driver. The results reveal several knowledge gaps within each discipline and highlights the lack of common understanding of safety across disciplines. On the basis of the results, the paper proposes a framework for further research on AV safety, identifying concrete opportunities for interdisciplinary research, with common goals and methodologies, and explicitly indicating the path for transfer of knowledge between sectors.

Speed behaviour upon approaching freeway curves.

The actual speed behaviour when drivers approach a curve is very relevant to assess the road design and safety but is mostly overlooked in the scientific literature. Most research into curve driving behaviour is focussed at the behaviour inside the curve, although the speed selection is done before curve entry. The main objective of this research is to identify which freeway characteristics play a role in driving speed selection. High Frequency Floating Car Data, detailed reconstruction of the curves and their surroundings, as well as three dimensional sight distance analysis, were used to analyse individual speed profiles on 153 Dutch freeway curves. By defining the positions where the acceleration approaches 0 m/s2 before and after a curve starts, the positions when the driver started and stopped decelerating upon curve entry were defined. Further correlation and regression analysis of those positions revealed that the radius of the curve is indeed a main explaining variable, as well as the speed driven before deceleration starts. Sight distances and cross section characteristics play a further role in determining the position where deceleration starts. Deceleration ends at approximately 135 m after curve start, and the speed in a curve is also correlated with the deflection angle and length of a curve. Sight distances do not play a role in selecting the speed in a curve based on this research. Overall, the findings indicate a non-constant nature and variability of speed behaviour upon curve entry. This can be used for safer freeway curve design and to assess traffic safety based on actual speed behaviour.

Are collision and crossing course surrogate safety indicators transferable? A probability based approach using extreme value theory.

In order to overcome the shortcomings of crash data a number of surrogate measures of safety have been developed and proposed by various researchers. One of the most widely used temporal indicators is time-to-collision (TTC) which requires the road users to be on a collision course. Road users that are strictly speaking not on a collision course actually might behave and take evasive actions as if they were, thus indicating that such near-miss situations might also be relevant for safety analysis. Taking that into account, a more flexible indicator T2, which does not require the two vehicles to be on a collision course, describes the expected time for the second road user to arrive at the conflict point. Recently extreme value theory (EVT) offering two approaches, block maxima (BM) and Peak over Threshold (POT), has been applied in combination with surrogate indicators to estimate crash probabilities. Most of this research has focused on testing BM and POT as well as validating various surrogate safety indicators by comparing model estimates to actual crash frequencies. The comparison of collision course indicators with indicators including crossing course interactions and their performance using EVT has not been investigated yet. In this study we are seeking answers to under what conditions these indicators perform better and whether they are transferable. Using data gathered at a signalized intersection focusing on left-turning and straight moving vehicle interactions our analysis concluded that the two indicators are transferable with stricter threshold values for T2 and that POT gives more reasonable results.

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

Observed accidents have been the main resource for road safety analysis over the past decades. Although such reliance seems quite straightforward, the rare nature of these events has made safety difficult to assess, especially for new and innovative traffic treatments. Surrogate measures of safety have allowed to step away from traditional safety performance functions and analyze safety performance without relying on accident records. In recent years, the use of extreme value theory (EV) models in combination with surrogate safety measures to estimate accident probabilities has gained popularity within the safety community. In this paper we extend existing efforts on EV for accident probability estimation for two dependent surrogate measures. Using detailed trajectory data from a driving simulator, we model the joint probability of head-on and rear-end collisions in passing maneuvers. We apply the Block Maxima method and estimate several extremal univariate and bivariate models, including the logistic copula. In our estimation we account for driver specific characteristics and road infrastructure variables. We show that accounting for these factors improve the head-on and rear-end collision probabilities estimation. This work highlights the importance of considering driver and road heterogeneity in evaluating related safety events, of relevance to interventions both for in-vehicle and infrastructure-based solutions. Such features are essential to keep up with the expectations from surrogate safety measures for the integrated analysis of accident phenomena, which show to significantly improve from the best known stationary extreme value models.

A forward collision avoidance algorithm based on driver braking behavior.

Measuring risk is critical for collision avoidance. The paper aims to develop an online risk level classification algorithm for forward collision avoidance systems. Assuming risk levels are reflected by braking profiles, deceleration curves from critical evasive braking events from the Virginia "100-car" database were first extracted. The curves are then clustered into different risk levels based on spectrum clustering, using curve distance and curve changing rate as dissimilarity metrics among deceleration curves. Fuzzy logic rules of safety indicators at critical braking onset for risk classification were then extracted according to the clustered risk levels. The safety indicators include time to collision, time headway, and final relative distance under emergency braking, which characterizes three kinds of uncertain critical conditions respectively. Finally, the obtained fuzzy risk level classification algorithm was tested and compared with other Automatic Emergency Braking (AEB) algorithms under Euro-NCAP testing scenarios in simulation. Results show the proposed algorithm is promising in balancing the objectives of avoiding collision and reducing interference with driver's normal driving compared with other algorithms.

Risk appraisal of passing zones on two-lane rural highways and policy applications.

Passing on two-lane rural highways is associated with risks of head-on collision resulting from unsafe completion of passing maneuvers in the opposite traffic lane. In this paper, we explore the use of time-to-collision (TTC) as a surrogate safety measure of the risk associated with passing maneuvers. Logistic regression models to predict the probability to end the passing maneuver with TTC less than 2 or 3s-threshold were developed with the time-gap from initiation of the maneuver to arrival of the opposite vehicle (effective accepted gap), and the passing duration as explanatory variables. The data used for model estimation was collected using stationary tripod-mounted camcorders at 19 passing zones in Uganda. Results showed that passing maneuvers completed with TTC less than 3s are unsafe and often involved sudden speed reduction, flashing headlights, and lateral shift to shoulders. Model sensitivity analysis was conducted for observed passing durations involving passenger cars or short trucks (2-3 axles), and long trucks (4-7 axles) as the passed vehicles for 3s TTC-threshold. Three risk levels were proposed based on the probability to complete passing maneuvers with TTC less than 3s for a range of opposite direction traffic volumes. Applications of the results for safety improvements of two-lane rural highways are also discussed.

Effects of parental vigilant care and feedback on novice driver risk.

Vigilant care aims at reducing adolescent risk behaviors while matching parental involvement to the level of alarm signs. This study examined the effect of parent training in vigilant care and technological feedback on driving risk of novice male drivers. A sample of 217 Israeli families was divided into four conditions: a) no-feedback, b) individual feedback, c) family feedback, and d) family feedback plus parent training in vigilant care. Feedback and risk assessment were conducted through in-vehicle data recorders. A significant difference was found in favor of the vigilant care group compared to the no feedback group. When only the drivers in the high risk percentiles were considered, the vigilant care group was found superior to the family feedback group. The findings suggest that parental training in vigilant care may help reduce driving risk.

The contribution of parents' driving behavior, family climate for road safety, and parent-targeted intervention to young male driving behavior.

One of the prominent issues in contemporary research on young drivers deals with the mechanisms underlying parents' influences on their offspring's driving behavior. The present study combines two sets of data: the first gathered from in-vehicle data recorders tracking the driving of parents and their teenage sons, and the second derived from self-report questionnaires completed by the young drivers. The aim was to evaluate the contribution of parents' driving behavior, participation in a parent-targeted intervention, and the teen drivers' perception of the family climate for road safety, to the driving behavior of young drivers during solo driving. The data was collected over the course of 12 months, beginning with the licensure of the teen driver, and examined a sample of 166 families who were randomly assigned to one of three intervention groups (receiving different forms of feedback) or a control group (with no feedback). Findings indicate that young male drivers' risky driving events rate was positively associated with that of their parents. In addition, any type of intervention led to a lower rate of risky driving events among young drivers compared to the control group. Finally, a higher perception of parents as not committed to safety and lower perceived parental monitoring were related to a higher risky driving events rate among young drivers. The results highlight the need to consider a complex set of antecedents in parents' attitudes and behavior, as well as the family's safety atmosphere, in order to better understand young drivers' risky driving. The practical implications refer to the effective use of the family as a lever in the attempt to promote safety awareness among young drivers.

Can providing feedback on driving behavior and training on parental vigilant care affect male teen drivers and their parents?

This study focuses on investigating the driving behavior of young novice male drivers during the first year of driving (three months of accompanied driving and the following nine months of solo driving). The study's objective is to examine the potential of various feedback forms on driving to affect young drivers' behavior and to mitigate the transition from accompanied to solo driving. The study examines also the utility of providing parents with guidance on how to exercise vigilant care regarding their teens' driving. Driving behavior was evaluated using data collected by In-Vehicle Data Recorders (IVDR), which document events of extreme g-forces measured in the vehicles. IVDR systems were installed in 242 cars of the families of young male drivers, however, only 217 families of young drivers aged 17-22 (M=17.5; SD=0.8) completed the one year period. The families were randomly allocated into 4 groups: (1) Family feedback: In which all the members of the family were exposed to feedback on their own driving and on that of the other family members; (2) Parental training: in which in addition to the family feedback, parents received personal guidance on ways to enhance vigilant care regarding their sons' driving; (3) Individual feedback: In which family members received feedback only on their own driving behavior (and were not exposed to the data on other family members); (4) CONTROL: Group that received no feedback at all. The feedback was provided to the different groups starting from the solo period, thus, the feedback was not provided during the supervised period. The data collected by the IVDRs was first analyzed using analysis of variance in order to compare the groups with respect to their monthly event rates. Events' rates are defined as the number of events in a trip divided by its duration. This was followed by the development and estimation of random effect negative binomial models that explain the monthly event rates of young drivers and their parents. The study showed that: (1) the Parental training group recorded significantly lower events rates (-29%) compared to the CONTROL group during the solo period; (2) although directed mainly at the novice drivers, the intervention positively affected also the behavior of parents, with both fathers and mothers in the Parental training group improving their driving (by -23% for both fathers and mothers) and mothers improving it also in the Family feedback group (by -30%). Thus, the intervention has broader impact effect beside the targeted population. It can be concluded that providing feedback on driving behavior and parental training in vigilant care significantly improves the driving behavior of young novice male drivers. Future research directions could include applying the intervention to a broader population, with larger diversity with respect to their driving records, culture, and behaviors. The challenge is to reach wide dissemination of IVDR for young drivers accompanied by parents' involvement, and to find the suitable incentives for its sustainability.

Risk evaluation by modeling of passing behavior on two-lane rural highways.

Passing maneuver on rural two-lane highways is a complex task, which has a significant effect on capacity, level of service and safety. The maneuver is conditioned on the gap between two successive vehicles on the opposing lane. The minimum time to collision, defined as the remaining gap between the passing vehicle and the oncoming vehicle at the end of the passing process, expresses a measure of the risk involved in the passing maneuver. This paper develops a model that explains the minimum time to collision. The model formulation is based on the analysis of drivers' passing decisions on two-lane rural highways using an interactive driving simulator. The simulator enables the collection of vehicle speeds and positions for different road and traffic scenarios. In addition to the driver simulator, participants responded to a questionnaire which collected information about their socio-demographic characteristics. The composed dataset was analyzed and processed to develop a model that predicts the risk associated with the passing behavior. Tobit regression models were found to be more suitable, in comparison to ordinary least square models and Hazard-based Duration models. The explanatory variables tested represent road geometry, traffic conditions and drivers' characteristics. It was found that while the traffic related variables had the most important effect on the measure of risk chosen, factors related to the geometric design and the driver characteristics also had a significant contribution.

Impact of infrastructure characteristics on road crashes on two-lane highways.

OBJECTIVE: The main purpose of this study is to estimate and quantify the contribution of the infrastructure to highway crashes and to develop an infrastructure coefficient, that represents the overall characteristics of the highway and could be used as an independent variable in a crash-prediction model. METHODS: The infrastructure is defined in this study as the highway and its geometric features, including alignment, road-side elements, sight-distances, presence of guardrails, access-points, roadway consistency, and additional variables that measure the overall quality of the highway alignment and elements. The analysis and developments are conducted for two-lane rural highways. The approach taken is to identify the high crash-rate roads, those with crash rates above 0.25 crashes per million vehicle-km, by Smallest Space Analysis. This type of analysis allows the aggregation of higher crash-rate roads versus lower-crash-rate roads only by their infrastructure coefficients, without consideration of their crash records. RESULTS: Crash rates that are attached by Smallest Space Analysis to the group of roads that had less desirable infrastructure features show a high correlation between the same roads and high crash rates vs. identified better infrastructures and low crash rates. Further analysis shows that low crash-rate infrastructure, as defined in this study for two-lane rural highways, can reduce the crash rate by 44% versus high crash-rate infrastructure, at the 99% confidence level, which is almost a certainty. A model for the prediction of crash rates based on a proposed infrastructure coefficient is calibrated and presented. CONCLUSIONS: It is suggested that this model be used in evaluating alternatives for new highways or in improving the alignment and road features of existing highways.