Benefit evaluation of V2X-enhanced braking in view obstructed crossing use cases

TL;DR

This work tackles crash avoidance when line-of-sight is blocked by obstacles at intersections. It proposes a 2-stage braking system that uses V2X to trigger a partial, early brake, followed by a sensor-triggered AEB for full braking, aiming to overcome sensing limitations while satisfying ASIL constraints. Crash severity is modeled via logistic regression on GIDAS data to quantify injury risk under various impact scenarios, enabling the evaluation of safety benefits across 35 obstructed-crossing use cases involving cars and bicycles. Through a lightweight simulator, the study demonstrates that a V2X-enhanced 2-stage brake can achieve complete crash avoidance with a 2 s TTC threshold and near-complete avoidance with 1.5 s TTC, significantly reducing injury risk, especially for bicycle opponents. The results highlight substantial practical impact for real-world automated braking systems, while acknowledging non-idealities such as false positives and communication uncertainties needing future investigation.

Abstract

If a crash between two vehicles is imminent, an Automatic Emergency Brake (AEB) is activated to avoid or mitigate the accident. However, the trigger mechanism of the AEB relies on the vehicle's onboard sensors, such as radar and cameras, that require a line of sight to detect the crash opponent. If the line of sight is impaired, for example by bad weather or an obstruction, the AEB cannot be activated in time to avoid the crash. To deal with these cases, a 2-stage braking system is proposed, where the first stage consists of a partial brake that is triggered by Vehicle-to-everything (V2X) communication. The second stage is composed of the standard AEB that is triggered exclusively by an onboard sensor detection. The performance of this V2X-enhanced 2-stage braking system is analysed in obstructed crossing use cases and the results are compared against the use of an AEB-only system. The benefit is quantitatively assessed by determination of the crash avoidance rate and, if the crash cannot be avoided, by estimation of the crash severity mitigation.

Figures (6)

• Figure 1: Decision process for 2-stage brake in the ego vehicle.
• Figure 2: Regression models describing the correlation between the probability of severe/fatal injuries and impact velocity for ego vehicle, opponent bike and opponent passenger car. Side of opponent passenger car is separated into impact zones AC and B. The standard deviation is shown as an enclosing tube for each model.
• Figure 3: SCP use case crossing layout with ego and opponent vehicles, building and parking car view obstruction as well as distances $d_\mathrm{ego}$, $d_\mathrm{opp}$ from ego and opponent vehicle paths to view obstructions.
• Figure 4: Depiction of unbraked impact position in percent of opponent length.
• Figure 5: Ego vehicle with simplified sensor model, mounted at the vehicle front and defined by sensor angle and range, as well as communication system model, defined by communication range.