ProxiCycle : Passively Mapping Cyclist Safety Using Smart Handlebars for Near-Miss Detection

Research Background and Issues

• Issues and Challenges: The authors identified that many potential cyclists are deterred from choosing cycling as a mode of transportation due to concerns about traffic safety, particularly the fear of collisions with motor vehicles. Existing methods primarily rely on sparse and delayed traffic accident data or self-reports from cyclists, lacking real-time, standardized, and scalable approaches to quantify and assess road safety.
• Research Significance: Cycling, as an active mode of transportation, has positive impacts on personal health and environmental sustainability. However, due to perceived safety concerns, cycling has not been widely adopted. In the United States, cycling accounts for less than 1% of all trips, and its safety is closely tied to users' perceived safety, which is especially critical for beginners.
• Research Motivation and Related Work:
  • Existing studies suggest that improved cycling infrastructure can directly enhance cycling safety, but implementation often requires time and funding.
  • Some crowd-sourced cycling safety maps have been proposed, but they lack large-scale quantitative data support, and self-reported data is prone to subjective bias.
  • Therefore, implementing a passive, scalable, and easily installable sensing method can address the shortcomings of existing approaches while promoting cycling adoption and safety improvements.

Solution

• Methodology and Innovation:

  1. A smart bicycle handlebar sensing device named ProxiCycle is proposed, featuring a dual-sensor design to measure the distance of motor vehicles overtaking cyclists during rides.
  2. The system passively collects data without requiring users to actively mark incidents, offering a low-cost and scalable solution for innovatively measuring close-pass events by vehicles.
  3. This new method can aggregate user data to provide real-time, high-resolution insights into urban road safety.

• Implementation Steps and Key Technologies:

  1. Hardware Design:
     • The device is installed at the end of the bicycle handlebar and uses time-of-flight (ToF) infrared sensors to accurately measure the distance between vehicles and bicycles.
     • It includes lateral and rear-facing sensors to detect the trajectory of vehicles approaching from behind and overtaking.
     • The hardware is miniaturized, low-cost (<$25), supports Bluetooth Low Energy (BLE) communication, and is battery-powered.
  2. Signal Processing Pipeline:
     • After raw data collection, a streaming signal processing pipeline detects sensor events (e.g., dynamic changes triggered by vehicles entering the sensor's field of view).
     • A distance threshold and the sequential triggering of rear and lateral sensors are used to identify "close-pass" events.
  3. Validation and Deployment:
     • Controlled experiments validate the accuracy of sensor readings.
     • A pilot study with 7 participants in real-world cycling environments, followed by a long-term deployment with 15 participants, was conducted to collect two months of "close-pass" data across an urban area.

Research Outcomes

• Key Findings:

  1. In technical evaluations, the device demonstrated high accuracy in both laboratory and real-world environments, achieving F1 scores of 0.915 and 0.868 for close-pass events (<1.3 meters) and safe-pass events (1.3–3 meters), respectively.
  2. During the long-term deployment involving 2,050 overtaking events, the device successfully mapped hotspot distributions of overtaking incidents across the city.
  3. The study revealed that locations of close-pass events significantly correlated with historical bicycle accident data from the past five years (Pearson correlation coefficient = 0.6, P<0.0001) and were highly consistent with perceived safety survey responses (PR-AUC 0.85).

• Advantages:

  1. Compared to traditional methods of collecting bicycle accident data, ProxiCycle offers a more real-time and dense data collection mechanism.
  2. The device is easy to install, uses passive data collection, and is low-cost, making it suitable for large-scale deployment.
  3. The data is privacy-preserving, as it does not collect video or images, only measuring physical distances.

• Limitations and Future Directions:

  1. Coverage Limitations: The current method only captures lateral "close-pass" events and cannot fully reflect other types of hazards, such as "dooring" or "right-hook" collisions.
  2. Behavioral Bias: Cyclists' own behaviors (e.g., riding closer to the center of the road or staying near the curb) may introduce biases in safety data.
  3. Data Aggregation: While individual user data may contain errors, aggregating data across users can filter and enhance reliability.
  4. For future improvements, the study suggests integrating environmental and temporal contexts (e.g., traffic flow, weather) or incorporating additional sensors, such as low-resolution video, to expand functionality.

Conclusion

ProxiCycle provides a highly practical solution to address traffic safety issues related to cycling. It generates real-time bicycle road safety data in an automated and low-cost manner, aligning closely with authoritative data. The research outcomes offer a new tool and data support for enhancing the safety and adoption of urban cycling. Additionally, this study serves as an effective case for future city-scale traffic monitoring and improvements.