Artificial intelligence hold out promise of improving safety of cycling, it is claimed

A trial of how artificial intelligence (AI) and datasets could reduce cycling accidents in the US has been held out as promising by a tech news site.

John Brandon, writing for the Venture Beat website, said the trial by Tome Software will test real cyclists at the University of Michigan's Tech Lab at Mcity, focusing on what the company says is the most common type of accident for cyclists, a car hitting a bike from the side or behind. The idea is to create a system that can alert both the cyclist and driver when a bike is in one of these danger zones, he said.

"Machine learning... will correlate several datasets, and examine factors like time of day, road characteristics (such as the speed limit), and existing crash data from cars and cyclists in that area," he continues. But Tome Software is keen to avoid false warning alerts, telling Brandon that it is "really tough" for a computer to understand whether a cyclist is on a road or a cycle lane.

The journalist is hopeful that safety can be improved by the project, also covered by the TechCrunch site, which said the bike manufacturer Trek had approached the software company to help "get automakers interested in working with the cycling community".