Abstract
Researchers have developed S-Door, a dooring-prevention system designed to protect cyclists and pedestrians by passively monitoring Bluetooth Low-Energy (BLE) signals emitted by their smartphones. By leveraging existing vehicle hardware and a novel localization algorithm called TRI² (Triangulation-Trilateration), the system alerts drivers to potential collisions without requiring new sensors or user opt-in. Preliminary results show a 100% detection rate in close-proximity scenarios.
Introduction
"Dooring"—the act of opening a vehicle door into the path of a vulnerable road user (VRU) like a cyclist—is a persistent urban safety hazard. In cities like Chicago, these accidents occur daily, often resulting in serious injury. Despite the prevalence of Advanced Driver Assistant Systems (ADAS), current technologies struggle to prevent dooring for three primary reasons:
- Cost: High-end sensors like LiDAR or radar are prohibitively expensive for entry-level consumer vehicles.
- Line-of-Sight: Cameras and ultrasonic sensors often fail to see around corners or behind parked cars where VRUs might be hidden.
- Friction: Many existing V2X (Vehicle-to-Everything) solutions require cyclists to "opt-in" by installing specific apps or hardware, which limits adoption.
S-Door addresses these challenges by utilizing the BLE transceivers already integrated into modern vehicles for keyless entry and infotainment systems.
Methodology: The TRI² Algorithm
The core of the S-Door system is the passive observation of Advertising Data (AD) packets. Most smart devices constantly broadcast these packets to remain discoverable. S-Door extracts the Received Signal Strength Indicator (RSSI) and, where available, the Angle-of-Arrival (AoA) from these packets to locate the device.
To overcome the limitations of standard localization, the authors introduced the TRI² (Triangulation-Trilateration) algorithm. Standard trilateration uses distance (RSSI) from multiple receivers but often fails when a target is outside the "bounds" of the receivers (e.g., behind the car). Triangulation uses the direction (AoA) of the signal.
TRI² combines these by creating "pentagons" of probability for each receiver—merging a circle sector (from AoA) with a rectangle (representing the signal's path). The intersection of these shapes from multiple on-vehicle receivers provides a robust estimate of the VRU's location, even when they are approaching from behind. This is particularly effective for the "narrow vehicle" problem, where the small baseline between receivers makes standard trilateration imprecise at long distances.
Experimental Setup
The authors evaluated S-Door using a custom-built prototype. The hardware consisted of three Nordic Semiconductor nrf52833-DK boards mounted on the roof of a subcompact SUV, connected to a Raspberry Pi 4. The team recorded data for VRUs walking, running, and biking toward the vehicle, conducting 10 trials for each scenario to collect time-series distance data.
Key Results and Discussion
In all tested scenarios, S-Door achieved a 100% True Positive Rate (TPR) when the VRU was in close proximity to the vehicle. This is critical because the logarithmic relationship of RSSI means the system becomes more accurate as the VRU gets closer—exactly when the risk of dooring is highest.
However, the system occasionally suffered from False Positives (FP) at greater distances. Because RSSI can be "noisy," the system sometimes predicted a VRU was closer than they actually were, triggering an early alert. For a safety-critical system, the authors argue that a "cautious" alert is preferable to a missed detection (False Negative).
Limitations and Future Work
As a Work in Progress (WIP), S-Door faces several hurdles before commercial deployment:
- Environmental Factors: Future work must evaluate how rain, snow, or dense urban "multipath" interference affects BLE signal reliability.
- Legacy Support: While BLE 5.1+ supports Angle-of-Arrival, many legacy smartphones do not yet broadcast the necessary Constant Tone Extension (CTE) fields.
- Sensor Fusion: To reduce the False Positive Rate, the authors suggest fusing S-Door data with existing low-cost sensors like side-view cameras or ultrasonics.
By utilizing signals already in the air and hardware already in the car, S-Door represents a significant step toward making urban environments safer for everyone without requiring expensive new infrastructure.
Summary updated