No More Endless Reds? AI System Speeds Up Detroit’s Traffic Flow

Drivers navigating metro Detroit may soon experience smoother commutes, thanks to an innovative AI project spearheaded by the University of Michigan.

Researchers at the university have developed a system that efficiently retimes traffic lights, aiming to minimize stops and alleviate congestion at intersections.

Craig Bryson, senior communications manager for the Road Commission for Oakland County, highlighted the traditional challenges of retiming signals. “Historically, it was labor-intensive and costly, so adjustments were infrequent. We were fortunate to do it every five to ten years,” he explained.

Many municipalities rarely updated their traffic light timings, even as traffic patterns evolved. The conventional method involved dispatching technicians to manually count vehicles at intersections—a process both time-consuming and prone to human error. “Humans get tired and can miss cars,” Bryson noted.

The University of Michigan Transportation Research Institute is now leveraging anonymized GPS data from vehicles to determine optimal light timings. This data, sourced from automotive companies, allows for a comprehensive analysis of traffic flow.

Zachary Jerome, a postdoctoral research fellow at the institute, emphasized the system's efficiency. “By utilizing data from a small percentage of vehicles, we can retime traffic signals to benefit all road users,” he stated.

While implementing new timings still requires physical input at intersections, Bryson anticipates future advancements will enable remote reprogramming of traffic lights.

Initial tests have shown promising results. At four signals along 8 Mile Road on the Farmington Hills and Livonia border, delays decreased by 30%, and traffic stops reduced by 40%. Similarly, nine signals along 12 Mile Road in Royal Oak experienced a 20% reduction in both delays and stops.

Bryson underscored the significance of these improvements. “Reducing congestion is crucial for safety. Frequent stops can frustrate drivers, increasing the likelihood of risky behaviors like running red or yellow lights,” he said.

Jerome highlighted the cost-effectiveness of the new system. “Optimizing a traffic signal traditionally costs around $4,000 to $5,000. Our system could reduce that to approximately $2,500,” he noted.

Bryson expressed optimism about the system's broader applicability. “This approach is universally applicable to any traffic signal. In the long run, it's inevitable that this will become the standard for retiming signals worldwide,” he predicted.

Jerome added, “As connectivity and AI technology advance, this research will significantly benefit future roadway implementations.”

The project is funded by a $1.4 million grant from the U.S. Department of Transportation.