Modelling overtaking strategy and lateral distance in car-to-cyclist overtaking on rural roads: A driving simulator experiment

Abstract

The involvement of cyclists in road crashes has not been decreasing with the same magnitude as the involvement of other road users. In particular, the interactions between cyclists and motorized traffic can lead to high-severity crashes. To improve the safety of these interactions, a thorough understanding of road user behaviour is first needed. In this study, we focused on drivers overtaking cyclists on rural roads. The two main objectives of this study were to develop models that predicted: (a) drivers' decisions to perform either a flying or an accelerative overtaking manoeuvre in the presence of oncoming traffic, and (b) the lateral comfort distance that drivers maintain from cyclists during the overtaking.A driving simulator study was designed to assess driver decision-making during the overtaking. The 37 drivers who participated in the study each performed seven overtaking manoeuvres with oncoming traffic. Out of the 259 overtaking manoeuvres, 168 were flying and 91 were accelerative. Binary logisticregression models with mixed effects predicted the type of overtaking strategy (flying or accelerative).Driving speeds were found to significantly affect the strategy. The overall performance of the models predicting the strategy was 85-90%. Models were also developed for predicting the lateral comfort distance. The results show that the lateral comfort distance is mostly affected by the longitudinal distance between the subject vehicle and the oncoming vehicle, the longitudinal distance between the subject vehicle and the cyclist, and the presence of an oncoming vehicle-as well as by the drivers' characteristics (sensation seeking in flying overtaking manoeuvres and ordinary violations in accelerative manoeuvres). The root mean square error, which was used to assess the performance of the models, ranged from 0.56-0.62.In conclusion, the models predicting the overtaking strategy performed reasonably well, while the models predicting lateral distance did not provide accurate predictions. The models predicting overtaking strategy may support 1) the development and evaluation of active safety systems, 2) the design of automated driving, and 3) policy making.

Modeling Drivers’ Strategy When Overtaking Cyclists in the Presence of Oncoming Traffic

“…Cyclist overtaking maneuvers can be split into four phases [6], [27], [30], which may help to prioritize and develop safety systems that support the driver in reducing crash risks in those phases: 1) approaching phase: when the ego vehicle approaches the cyclist, and the driver has to decide between a flying or an accelerative maneuver, 2) steering away phase: when the driver steers away to achieve a lateral distance to the cyclist, 3) passing phase: when the ego vehicle passes the cyclist while driving in parallel, 4) returning phase: when the driver has passed the cyclist and steers back into the lane. Several recent studies investigated driver behavior in these phases by leveraging different test environments: simulator studies [31], [32], test-track studies [30], [33], field test studies [27] and naturalistic driving studies [6], [34]- [36]. While simulator and test-track studies have lower ecological validity than field tests or naturalistic driving studies, they can offer more controlled data with higher resolution, which facilitates the development of predictive driver models [5].…”

“…In an overtaking scenario, being able to predict whether a driver would avoid collisions by braking (to initiate an accelerative maneuver) or by steering (to initiate a flying maneuver) can help active safety systems to tune intervention times [37], [38]. In fact, knowing a driver's overtaking strategy can also be valuable information for systems that aim at preventing a head-on collision with the oncoming traffic in the passing phase, as a result of a poorly timed flying maneuver [30], [32], and rear-end collisions with cyclists as a result of a poorly timed accelerative maneuver. In this respect, Farah et al modeled the driver's maneuver choice depending on the ego vehicle speed with logistic regression, based on data from a simulator study.…”

“…In this respect, Farah et al modeled the driver's maneuver choice depending on the ego vehicle speed with logistic regression, based on data from a simulator study. The authors fitted the model on three different distances to the cyclist and concluded that the models perform better at shorter distances [32]. Nevertheless, the model from Farah et al is of more descriptive nature and, therefore, harder to implement in active safety systems that require a predictive model running in real-time [39].…”

“…The model takes as inputs some characteristic distances between the road users ( Fig. 2), which are reported in previous literature as relevant factors that influence driver behavior [30], [32], [36], [41]: d…”

“…In contrast, some flying maneuvers were classified as accelerative maneuvers. From a safety perspective, this behavior is desirable because an accelerative maneuver is a safer choice compared to a flying one, under the assumption that oncoming traffic is present and no vehicle is following closely behind [32]. Further, it was the riskiest flying maneuvers that were classified as accelerative.…”