AI news ·
Raleigh tests AI to keep intersections moving without adding lanes
Raleigh is testing AI with existing cameras to tune signal timing at busy intersections. Small, human-reviewed tweaks are cutting delays while protecting privacy and budgets.
Raleigh tests AI to fine-tune traffic signals at busy intersections
Raleigh is using existing traffic cameras, AI, and mapping software to make crowded intersections run more efficiently and safely. The goal isn't to erase traffic. It's to squeeze better performance out of signals where people, bikes, scooters, and cars all meet.
City engineers convert live video into data, then review patterns in near real time. The focus is on small, high-leverage tweaks during peak hours that improve flow without widening roads.
How the pilot works
- Computer vision turns camera feeds into data on turning movements, traffic volumes, and travel modes.
- Analytics replaced manual counts, cutting workload to a fraction of what it was.
- Due to computing limits, 10-12 cameras are analyzed at a time, rotated based on active signal studies (including Glenwood Avenue and parts of downtown).
- Engineers adjust signal timing by a few seconds where needed, based on real conditions instead of assumptions.
Why intersections first
Intersections are the bottlenecks. Widening roads often costs more and delivers less than fixing how signals operate where conflict points happen.
By maximizing efficiency at intersections, the city can delay expensive road projects and put funds where they matter most.
Human in the loop, privacy by design
The AI does not automatically change signal timing. Engineers review the data and make the call. During the pilot, staff verified accuracy by running manual counts alongside the software across more than a dozen test cases.
The city does not store video or collect identifying information. Data is limited to anonymous counts (vehicle, pedestrian, bicycle) aggregated in 15-minute intervals.
From data to decisions: the "digital twin" model
Traffic insights feed into a high-resolution 3D model of Raleigh-a digital twin-so staff can compare before-and-after performance and confirm whether changes are working.
This approach helps engineers see how one intersection affects a corridor, not just a single camera at a single moment.
Operational gains you can measure
Small timing shifts during the evening rush-sometimes just a few seconds-are reducing unnecessary delay. The city targets locations under active review, learns quickly, then moves on to the next set of signals.
The process creates a repeatable playbook: analyze, tweak, validate, and scale where results hold up.
Scaling and cross-department value
The pilot runs on existing infrastructure and limited resources. Scaling will require investment, but the learning curve is clear and the benefits extend beyond traffic.
Police, 911 dispatchers, and stormwater teams already use the camera network during emergencies, including flooding, to understand conditions on the ground.
What managers can take from Raleigh's playbook
- Start with what you have. Use existing assets before buying new ones.
- Define a narrow scope. Focus on one constraint (intersections) for faster ROI.
- Keep humans in control. Review AI outputs, then decide.
- Measure before/after. Validate results with parallel manual checks.
- Design for privacy. Aggregate data, avoid storing raw feeds, remove identifiers.
- Create spillover value. Build systems other teams can use during emergencies.
- Scale in phases. Prove value, then invest to expand capacity.
Why this matters for growing cities
Raleigh passed 500,000 residents and is feeling the strain on roads built for lighter loads. Smarter signal timing stretches the network you already have, without more asphalt.
As the city refines its model, expect a clearer path to targeted investments that move people safely and predictably.
Resources
Build AI literacy across your team
If you're planning similar pilots or workforce upskilling, explore role-based options such as the AI Learning Path for Transportation Managers or the AI Learning Path for Training & Development Managers to help staff run pilots and institutionalize analytics workflows.
Bottom line: focus on intersections, keep humans in the loop, protect privacy, and scale what the data proves.