Towards computational models for road-user interaction: Drivers overtaking pedestrians and cyclists

Licentiatavhandling, 2020

Introduction: Crashes resulting from a failed interaction between drivers and vulnerable road users, such as pedestrians or cyclists, can lead to severe injuries or fatalities, especially after failed overtaking maneuvers on rural roads where designated refuge areas are often absent, and impact speeds high. This thesis contains two studies that shed light on driver interaction with either 1) a pedestrian or 2) a cyclist, and oncoming traffic while overtaking. Methods: The first study modeled driver behavior in pedestrian-overtaking maneuvers from naturalistic and field test data, quantifying the effect of the pedestrian’s walking direction and position, as well as the presence of oncoming traffic, on the lateral passing distance and overtaking speed. The second study modeled cyclist-overtaking maneuvers with data from a test-track experiment to quantify how the factors time gap to the oncoming traffic and cyclist lane position affect safety metrics during the maneuver and the overtaking strategy (i.e., flying or accelerative, depending on whether the driver overtook before or after the oncoming traffic had passed, respectively). Results: The results showed that, while overtaking, drivers reduced their safety margins to a pedestrian when the pedestrian was walking against the traffic direction, closer to the lane and when oncoming traffic was present. Results for cyclist overtaking were similar, showing that drivers left smaller safety margins when the cyclist rode closer to the center of the lane or when the time gap to the oncoming traffic was shorter. Under these critical conditions, drivers were more likely to opt for an accelerative maneuver than a flying one. The oncoming traffic had the most influence on drivers’ behavior among all modeling factors, in both pedestrian- and cyclist-overtaking maneuvers. Conclusion: Drivers compromised the risk of a head-on collision with the oncoming traffic by increasing the risk of rear-ending or side-swiping the pedestrian or cyclist. This thesis has implications for infrastructure design, policymaking, car assessment programs, and specifically how vehicular active safety systems may benefit from the developed models to allow more timely and yet acceptable activations.