AR boosts cyclist safety around autonomous vehicles

Known as CycleARcade, the platform is the latest development in the university’s efforts to make cycling safer when autonomous vehicles (AVs) arrive on our roads.

AR headsets allow cyclists to see and interact with simulated AVs in real-world environments, with different scenarios and setups easily swapped out. As described in two separate research papers, CycleARcade was used to test different warning systems for cyclists as well as to assess how cyclists in different countries might react to AVs on their roads. 

“Ultimately, we’re aiming to thoroughly explore the ways in which cyclists and autonomous vehicles can speak the same language on the roads to keep both as safe as possible,” said research lead Professor Stephen Brewster from Glasgow’s School of Computing Science.

“Human drivers and riders have developed a sophisticated series of signals to help decide who has the right of way or who has priority in a change of lanes, for example, and it’s vital that cyclists can have the same level of trust and understanding with self-driving cars.

“CycleARcade is a powerful tool to help explore how that new language can be developed, using real bikes in real physical spaces, with virtual elements that can be tweaked or replaced in real-time.”

The first paper outlined how CycleARcade was used to test new designs for interfaces which could alert cyclists to nearby AVs and provide information about the cars’ intentions. Using a focus group of 20 cyclists, the Glasgow team developed and tested three virtual displays – RoadAlert, rearview and Gem - to give riders information about vehicle behaviour around them.

“What we found in this study is that you don't need to alert cyclists about all vehicles equally,” said lead author Ammar Al-Taie, also from the School of Computing Science.

“Cyclists need focused awareness of vehicles that pose the greatest risk, like those approaching from behind or vehicles that won't yield, while being able to maintain attention on the road ahead. RoadAlert was the design that brought those qualities together most effectively for our study participants.”

In the second paper, computing scientists and psychologists from Glasgow University and colleagues from the KTH Royal Institute of Technology in Sweden described how CycleARcade was used to study the road safety expectations of cyclists in three different countries, namely Sweden, Oman and Scotland. It was found that the different levels of cycling infrastructure across the three countries had a significant influence on cyclist expectations and behaviour. 

“These findings clearly show that cyclists learn to share the roads with cars differently from country to country, which suggests that self-driving cars might need to adapt their communication methods to better speak the language of the local roads,” said Al-Taie.

The team’s papers, titled ‘Around the World in 60 Cyclists: Evaluating Autonomous Vehicle-Cyclist Interfaces Across Cultures’ and ‘evARything, evARywhere, all at once: Exploring Scalable Holistic Autonomous Vehicle-Cyclist Interfaces’, will be presented at the CHI Conference in Yokohama, Japan at the end of April.