Projects:2019s1-124 Development of a Tool for Naturalistic Measurement of Vehicle-Cyclist Passing Distances
Members: Scott Adamson Robert Broadhead Max Telford
Supervisors: Braden Phillips Jamie Mackenzie (CASR)
Motivation
While cycling is a mode of transport to be encouraged, cyclists are far more likely to be killed or hospitalised from road accidents than other commuters [1]. Having accurate data on where, how often, and how close cars overtake cyclists is an important step towards improving the laws and infrastructure that keep cyclists safe.
Aim
This project is sponsored by the Centre for Automotive Safety Research (CASR) and seeks to improve and validate the capabilities of an existing bicycle mounted device for measuring vehicle-cyclist passing distances. The three areas of focus are the sensor choice, device design and verification.
Sensor Choice
Existing Limitations:
Measurement Frequency and Size
Sensor considerations:
Measurement Frequency
Resolution/Accuracy
Cost
Durability
Power Consumption
Size
Weight
Communication
Interference
Other Output Data
Available technologies and sensors:
Ultrasonic and LiDAR
Conclusion:
On-road testing results showed both LiDAR sensors were able to detect 38% more overtakes than the ultrasonic sensor. A particular LiDAR sensor was chosen as it is not limited by frequency and maintained accuracy in all lighting conditions. This will allow the sensor housings to be considerably smaller whilst maintaining durability and detecting more overtakes.
Verification
Verifying the accuracy of the data collected adds value to existing and all future trial results.
The steps taken to verify the device are as follows:
- Mount a Raspberry Pi camera to the bicycle
- Sync the camera with the sensor data
- Analyse the video and accuracy of the sensors
- Build and implement a detection algorithm
Conclusion:
Testing revealed video recordings as costly and time consuming method, but allowed for more efficient sensor detection algorithms to be developed.
