Glasgow's joggers, take note: the next car that fails to give way to you on Great Western Road may not have a driver to glower at.
A new study led by the University of Glasgow has found that self-driving cars will need to learn a whole new visual vocabulary if they are to share the road safely with runners — and the team has designed exactly that.
Researchers at the University's School of Computing Science, working with colleagues at KAIST in South Korea, used augmented-reality headsets to stage life-sized encounters between pedestrians and a simulated autonomous vehicle (AV). The results, they say, should reshape how driverless cars are trained to behave around the world's most popular form of exercise.
Runners take more risks
Twenty-four volunteers walked or ran towards a junction while a virtual driverless car approached. The findings were striking: runners were far more likely than walkers to push on across the road, less likely to slow down to weigh up the traffic, and on three occasions were “struck” by the simulated car.
In two of those near-misses, the runner saw a red signal from the car but chose to dart ahead anyway, misjudging its speed in order to keep their pace.
“I'm a runner myself, and I've noticed that crossing roads while running feels different than when I'm walking,” said Ammar Al-Taie, one of the paper's corresponding authors, who carried out the research at Glasgow before moving to KAIST. “I'm much more motivated to keep moving because slowing to let a car pass and accelerating again takes a real physical effort.”
Runners, he added, “do seem more tolerant of risk if it helps them keep moving. That makes them a riskier class of road user for self-driving cars to deal with.”
A new language of light
The Glasgow team's answer is a system of external lights mounted on the car — what engineers call an external Human-Machine Interface, or eHMI. Think of it as the driverless equivalent of the friendly nod or wave a human driver gives at a zebra crossing.
Two designs were tested in the trials. A simple “LightRing” of red and green around the vehicle proved easy for both walkers and runners to read at a glance. A more sophisticated “CyanBand” of animated lights, which swept inwards as the car braked, worked well for walkers but baffled runners, who had no time to decode the animation.
Building on that, the team has proposed a new design called DualBeam: two rows of lights along each side of the car, using amber — a traffic-cone orange — to signal “I'm not stopping”, and purple, closer to a royal or stage-light hue, to signal “on you go”. The unfamiliar colours, they argue, sidestep the split-second confusion that red and green can cause for a runner already focused on their pace.
DualBeam could also be paired with an early-warning system, sending a discreet buzz to a runner's smartwatch or earbuds when an autonomous vehicle is approaching.
Glasgow at the front
Professor Stephen Brewster, who leads the work at Glasgow, said the gap in research had been hiding in plain sight. “Running is the most popular physical activity in the world, with more than 600 million people estimated to run recreationally,” he said. “Self-driving car journeys are booming, with a million AV trips a month in the USA alone. Clearly, it will be increasingly important to ensure that runners and AVs can share the roads safely.”
For a city that fields everything from the Great Scottish Run to a thousand quiet Parkrunners along the Clyde, the message from Gilmorehill is reassuring: someone is already thinking about the runners.
The paper, Running into Traffic: Investigating External Human-Machine Interfaces for Automated Vehicle-Runner Interaction, was presented at the CHI 2026 conference in Barcelona last month.
