AI for Traffic Signal Optimization in Mid-Size Cities
AI signal optimization moved from research to billable project in mid-size cities. The implementation realities and what determines whether it earns its.
Traffic signal optimization is one of the AI use cases that produces measurable wins in mid-size cities — meaningful travel time reduction, emissions reduction, fewer rear-end crashes. The deployments need integration with the city’s existing ATMS and a multi-quarter commitment to data-driven operation.
What works in mid-size deployments and what determines success.
What the AI actually does#
Two families of approach:
Adaptive signal control. Real-time optimization based on detected demand. Picks phase order and durations dynamically. Tools like SCATS, SCOOT have been adaptive since the 1980s; modern AI variants improve on them with deeper learning and richer inputs.
Coordinated corridor optimization. Multi-signal arterial coordination for “green wave” behavior. Pulls from corridor-wide demand data; sets coordinated plans.
Predictive optimization. Uses upstream sensor data and historical patterns to set the next 5–15 minutes of signal behavior, not just reactive control.
The interesting modern systems do all three with reinforcement learning policies trained on the corridor’s actual traffic.
Where it earns its place#
Mid-size cities (100k–500k population) with multi-intersection arterials where the existing fixed-time plans haven’t been re-timed in years. Easy wins here.
Corridors with significant demand variation. Time-of-day, day-of-week, event-driven. Adaptive control beats fixed-time decisively.
Cities with budget for the supporting infrastructure. Detection upgrades, communication backhaul, ATMS integration.
Where it doesn’t#
Tiny networks (1–10 signals). Manual re-timing is fine.
Highly congested networks where signal optimization can’t fix capacity problems. AI re-shuffles green time; doesn’t add capacity.
Cities without operations capacity. AI signals need monitoring; without TMC staff, the system runs blind.
The data-pipeline reality#
For credible AI signal control:
- Reliable detection (loops, video, radar) at all approaches
- Communication backhaul (fiber or robust wireless) to ATMS
- Real-time data flowing from intersection to central
- Historical data retention for model training and audit
- Operator interface for override and monitoring
Most mid-size cities have ATMS systems. The data discipline that makes AI work is what’s often missing.
The integration question#
AI signal control must integrate with:
- The city’s ATMS (Centracs, MaxView, Intelight, custom)
- Detection systems
- Connected-vehicle infrastructure where deployed
- Emergency vehicle preemption
- Pedestrian and transit priority
Tools that don’t integrate with the operator’s existing screens get ignored.
Our data engineering practice handles this integration for city transportation departments.
Where AI doesn’t replace the engineer#
Traffic engineering is licensed work. AI signal optimization is a tool; the licensed engineer signs off on the operational concept and reviews ongoing performance.
What we ship for city transportation departments#
For signal optimization engagements:
- Detection data pipeline into ATMS
- AI optimization layer integrated with operator interface
- Performance dashboards (travel time, delay, emissions, safety)
- Manual override and monitoring workflows
- Annual performance review and re-training
The results that hold up#
Independent evaluations of AI signal optimization in mid-size cities consistently show:
- 10–25% reduction in corridor delay
- 5–15% reduction in stops
- Modest emissions reductions
- Small but real safety improvements (fewer rear-end crashes)
Results vary by corridor; the best wins come on busy arterials with variable demand.
The funding context#
Federal and state funding for smart-mobility projects is available; cities increasingly find AI signal optimization fundable. The cities that have built the underlying data infrastructure can deploy quickly; the ones that haven’t spend the first year on the data layer.
AI signal optimization is one of the clearest urban-AI wins. The infrastructure has to be ready. Our team builds the data and integration layer for city transportation programs. Tell us about the city.