You Shall Not Pass: Warning Drivers of Unsafe Overtaking Maneuvers on Country Roads by Predicting Safe Sight Distance

TL;DR

This work tackles unsafe overtaking on rural two-lane roads where long-range perception and V2V communication are unavailable. It introduces a sensor-based overtaking warning assistant that predicts the required sight distance for safe passing using in-car sensors and 3D map data, without autonomous overtaking, and tests two UI designs (monitoring and scheduling) in a VR driving study. The study reports that the assistant promotes a more patient driving style and high usability, though perception of usefulness and timing vary with user preferences; model accuracy improves when acceleration assumptions are aligned with observed behavior, and robustness analyses show the method is viable under realistic sensing constraints. The authors open-source the PSD prediction models and release a large driving dataset to support replication and further development. Overall, the work advances safe-overtaking support on non-connected rural roads and provides practical guidance on UI design and model parameters for driver-centric warning systems.

Abstract

Overtaking on country roads with possible opposing traffic is a dangerous maneuver and many proposed assistant systems assume car-to-car communication and sensors currently unavailable in cars. To overcome this limitation, we develop an assistant that uses simple in-car sensors to predict the required sight distance for safe overtaking. Our models predict this from vehicle speeds, accelerations, and 3D map data. In a user study with a Virtual Reality driving simulator (N=25), we compare two UI variants (monitoring-focused vs scheduling-focused). The results reveal that both UIs enable more patient driving and thus increase overall driving safety. While the monitoring-focused UI achieves higher System Usability Score and distracts drivers less, the preferred UI depends on personal preference. Driving data shows predictions were off at times. We investigate and discuss this in a comparison of our models to actual driving behavior and identify crucial model parameters and assumptions that significantly improve model predictions.

Figures (26)

• Figure 1: HUD with the scheduling-focused UI. (Top) Overtaking now is not safe (warning icon). The next overtaking opportunity (in red due to potential oncoming traffic) is in 243m. (Bottom) Overtaking now is not safe. There is not enough room between two vehicles ahead (red warning icon to the left). The next overtaking opportunity (in yellow due to a second lane opening) is in 866m.
• Figure 2: Static distances in a typical overtaking maneuver. $L_1$ and $L_2$ are the distances to the vehicle to be overtaken in phase (i) and after phase (iii), respectively. $L_{oing}$ and $L_{oen}$ are the lengths of the overtaking car and the vehicle to be overtaken, respectively.
• Figure 3: Comparison of the three acceleration models to measured data from the driving simulator ("real data"), in dependence of velocity. The clearly visible steps in the measured data stem from switching gears.
• Figure 4: Typical speeds traveled at various speed limits as reported in the works by Kockelman NAP22048, Mannering Mannering2007EffectsOI, and Haglund et al. HAGLUND200039. Our linear model (red) covers the upper end to be on the safe side and anticipate a higher speed rather than a lower one.
• Figure 5: Safety distances when overtaking with oncoming traffic. $L_{sm}$ denotes the static distance between the overtaking vehicle (which traveled a distance of $d_{tot}$ during the maneuver) and the oncoming vehicle (which traveled a distance of $d_{opp}$ during the maneuver) after the overtaking is finished.