PhD Research: What AVs need to learn about cyclist braking behaviours?

Automated vehicles (AV) have the potential to prevent conflicts between motor vehicles and cyclists, however existing literature only focuses on vehicle collision avoidance technology (e.g. autonomous emergency braking (AEB)).

Due to the high risk of the cyclist falling during braking due to front wheel lock-up, nose over, fly-off or low velocity/dismount, collision avoidance alone does not ensure the safety of cyclists.

Current thresholds of surrogate safety measures are derived from car-car conflicts, which are not sufficient for the safe braking of a cyclist.

This study aims to answer the following questions:

  1. What are the most dangerous scenarios in which the cyclist has to brake in order to avoid a collision with a motor vehicle?
  2. What are the safety boundaries for safe interactions for the identified scenarios?
  3. What surrogate safety measures can the AV use to reliably assess the risk of the cyclist having to brake strongly?

Information about real-life scenarios (in which there are conflicts between cyclists and motor vehicles) are obtained through the use of the Lidar sensors fitted to ZOE2. When parked in a conventional parking space, ZOE2 can identify cyclists and capture the cyclist’s (natural) trajectory and velocity over a range of 50-100 meters. So far, the trajectory and speed of more than 500 cyclists have been measured at three different locations in Brisbane.

The first stage of this study was conducted as part of Marco Reijne’s internship at CARRS-Q from 1 September 2019 to 4 December 2019. The next stage of this study will be conducted in The Netherlands to identify the cyclist’s safety boundary in identified scenarios and evaluate the performance of different surrogate safety measures in predicting unsafe braking by the cyclist.

ZOE2 is the research platform in the iMOVE and Queensland Department of Transport and Main Roads’ Cooperative and Highly Automated Driving (CHAD) Safety Study. Permission for its use for this study has been provided by Queensland Department of Transport and Main Roads.


Other Team Members

Visiting Intern, Marco Reijne, PhD Scholar at TU Delft, The Netherlands & Professor Arend Schwab, TU Delft, The Netherlands - Supervisor