Cycling safety with PassBox

I started PassBox with Jonathan Nolan to help improve cycling and pedestrian safety with targeted research. I developed the PassBox device to measure how close cars pass cyclists on the road. This is an Arduino based device that uses ultrasonic sensors for measuring passing distance, while also collecting GPS data. I designed and fabricated the enclosures. We have fitted these devices to hundreds of volunteers’ bicycles around Australia to collect data on how they are getting passed by other vehicles on the road. Collected data is reduced to passing events using purpose-built software I have developed that synchronises video footage to distance sensor readings and extracts passing events. These passing events are then coded according to infrastructure, the type of vehicle, etc. and are then used to build models assessing the impact of advertising campaigns, explore different cycling infrastructure, and assess the effect of passing distance laws. I have presented our findings at several international conferences, and we have had peer-reviewed journal articles published.

Check out the PassBox map to see all the passing events we have collected.

PassBox device and action camera mounted to a rack

We have also conducted pedestrian safety analysis using novel methods to estimate pedestrian counts at intersections using pedestrian operated signal data in combination with crash statistics.

Technology

• Python: data reduction reduces synchronises video footage to distance sensor readings and produces passing event videos. Object detection via TensorFlow is used to reject false positive passing events. Pandas is also used for data exploration and analysis
• Bash scripts: misc. utility scripts for managing data
• Arduino: electrical prototyping and programming using Arduino devices
• AWS: S3, EC2
• Electronics: design and fabrication of custom electronics including PCB design
• 3D Printing: PassBox device enclosures designed for 3D printing

Publications