Australia's traffic signal technology is, in most places, unchanged since the 1980s. It runs on fixed sequences: the light follows a preset timer, and sensors can nudge the timing a little, but the light has no real awareness of what is actually in front of it. The result is something anyone who has sat at a red light watching an empty road will recognise.
The Moreton Bay trial replaces that with a system that classifies every road user in real time: cars, trucks, cyclists, pedestrians. It then decides the most efficient way to clear the intersection, without being locked to a predetermined order. The hardware costs about AU$170,000 per intersection to install. That is not a small number for a council budget, but it is also not large relative to the cost of road widening, which AI traffic management can sometimes make unnecessary.
The performance case is strong. Pittsburgh has run a comparable system across dozens of intersections since 2012. Travel times fell by roughly 25%, waiting at red lights dropped by more than 40%, and emissions went down by about 21%. When Pittsburgh's developers first launched the system, pedestrians immediately complained they had been forgotten. Engineers added a maximum wait time for pedestrians after that feedback. The lesson has been learned elsewhere, but not always.
Here is where this becomes a decision that matters beyond road engineering. An AI traffic system does not have a single objective. It has whatever objective someone programs into it. Cities like San Jose in California have used AI signal priority to give buses more green time; bus travel times on those routes improved by over 50% and ridership rose 15% in early 2024. Copenhagen uses smart signals to detect large groups of cyclists and immediately clear their path, even if it holds cars a little longer. Torino gives its trams signal priority and saves transit passengers an average of five minutes per trip. These are deliberate choices about what a city values.
If Moreton Bay, and later other Australian cities, default to optimising for car throughput, the system will make driving faster and nothing else will change. If they choose to give buses and pedestrians equal weight in the algorithm, the city gradually becomes easier to move through without a car. The technology is neutral. The priority setting is not.
Cybersecurity is the part of this story that deserves more attention than it typically gets. Connected traffic lights have a documented vulnerability history. A critical flaw discovered in traffic controllers used across more than 60 countries left an open debugging port that required no password to access. Researchers in the Netherlands were able to trick AI traffic systems into responding to cyclists who did not exist, by reverse-engineering a public cycling app. Simulation studies show that changing a single light red for just two to three minutes creates knock-on congestion across other parts of the same network minutes later.
Australia has no national standard yet for how connected traffic infrastructure should be secured or audited. That gap needs to close before AI traffic systems are deployed at scale. The bare minimum is a fallback to fixed timing if the AI is unavailable, and a clear human override process.
For business operators, the practical impact of this technology comes in two waves. The first is operational: if AI lights reduce average travel times by even 15% in urban areas, delivery schedules, field service routing, and employee commute reliability all improve. The second wave is slower and more structural: cities that use this technology to make public transport and cycling more attractive will, over years, see different patterns of where people live, shop, and work. That reshapes retail catchment areas, last-mile logistics, and commercial property values.
The Petrie intersection is a quiet suburban test site. But the choices made there about what the system is told to optimise will be the template for every intersection that follows.