The challenge
The traffic management authority faced a governance problem as much as a technical one. Signal timings had been established in a 2014 study and lightly updated by individual engineers ever since. The city had grown, the road network had changed, and the trip patterns the timings were built for no longer reflected reality. Authority leadership understood the problem but had no mechanism to update 612 junctions systematically — each change required a study, and the pace of study could not keep up with the pace of urban change.
Discovery
The first 12 weeks were a measurement exercise: instrumenting 84 representative junctions with traffic counts and turning-movement detection, then quantifying the gap between current signal plans and the analytically optimal plan for each junction. The performance gap was large and consistent — but the diagnostic finding that shaped the programme was that the gap was different at every junction. A centralised re-timing exercise, done once, would have captured less than half the available improvement.
Our approach
System architecture: designed a hybrid edge-and-centre architecture — edge controllers computing optimal signal phases locally, with a central optimiser coordinating corridor green waves during morning and evening peak
Per-junction modelling: built a per-junction model learning actual flow patterns from continuous detection, replacing the assumed patterns from a study completed over a decade earlier
Unified priority integration: integrated pedestrian and emergency-vehicle priority into the same optimiser so each system operates on the same data rather than competing through separate controllers
Public transparency: designed and built a public-facing performance dashboard at corridor level — the same performance data the authority uses internally
The Outcome
The engagement delivered a hybrid edge-and-centre architecture: local edge controllers handle real-time phase decisions, the central optimiser coordinates corridor green waves at peak periods, and a continuous learning loop keeps every junction's timing plan current as the city evolves. Performance data is the same for the authority and the public.
The results
“The platform updates the city without the city having to commission a study every time. That is what makes it sustainable.”
