An AI coordination layer that sits above existing traffic infrastructure - optimising the whole network, not just individual intersections. Software-only. No roadworks.
The problem
SCATS - the traffic signal system used in 216 cities across 32 countries - adjusts cycle lengths, phase splits, and offsets at each intersection based on local detector data. It does this well.
But it has a documented architectural limitation: it distributes delay equally across intersection approaches rather than minimising total delay across the network. Improvements at one intersection can - and do - create worse outcomes elsewhere.
Nobody is responsible for optimising the network as a whole, because no tool exists to do so. Until now.
A right-turn phase was added to ease side-road congestion. The side road improved - but the arterial got significantly worse.
| Route | Before | After | Change |
|---|---|---|---|
| Main arterial | 240s | 350s | +46% |
| Collector road | 320s | 250s | -22% |
Source: Council travel time data, 2025
The solution
HOTMESS uses sensors cities already own to optimise the network as a whole - without replacing anything, disrupting traffic, or touching a single road surface.
No roadworks, no new hardware, no intersection visits. HOTMESS sits above existing SCATS infrastructure as a pure software layer.
Integrates SCATS loop data, CCTV feeds, and Bluetooth journey times - all already deployed, none currently used for automated analysis.
A recommendation engine, not autonomous control. Traffic operations teams stay in command. HOTMESS suggests; humans decide.
Optimises total network delay, not individual intersections. Addresses the equisaturation gap that SCATS was never designed to solve.
Phased approach
A graduated rollout designed around safety, transparency, and measurable outcomes at every stage.
Digital twin of a real urban network. Train reinforcement learning models on real SCATS data. Validate against historical performance.
In progressRun alongside live SCATS on a pilot corridor. Generate recommendations without acting. Measure what would have happened versus what did.
NextRecommendations surfaced to the operations team in real time. Human approval required for all actions. Build trust through transparent, auditable decision-making.
FutureGraduated autonomy for validated, low-risk optimisations. Human override always available. Safety-critical systems architecture throughout.
Long-termWe're looking for pilot partners, research collaborators, and forward-thinking transport agencies.
Get in touch