AI digital twin cuts traffic delays by 14% in Tees Valley pilot
TL;DR: Tees Valley Combined Authority has built an AI-powered digital twin of its road network that reduced delays by 13.7% during a six-month pilot. The system uses GPS bus tracking and roadside sensors to predict congestion and adjust traffic light timing automatically. It is one of the first UK local authority projects to show measurable results from AI-driven traffic management.
A virtual replica of the Tees Valley road network is using AI to anticipate traffic problems before they escalate. The system pulls data from GPS-tracked buses and roadside sensors, then adjusts traffic signal timing across 11 congestion hotspots without waiting for human intervention.
The results from the first phase are concrete: a 13.7% reduction in delays over six months. Transport officials say the AI consistently finds better routing solutions than human operators, and does so faster.
How the system works
The digital twin collects real-time data from multiple sources and builds a continuously updated model of traffic conditions. When congestion builds at one location, the AI evaluates the knock-on effects across the network and adjusts signal cycles accordingly.
Stephen Harker, leader of Darlington Council and transport lead at TVCA, noted that while human oversight remains in the system, the AI has demonstrated it can make quicker and more effective decisions. The software identifies where traffic should be diverted and recalculates signal timing in ways that human operators were not matching.
A test case for UK councils
For other UK local authorities exploring AI applications, the Tees Valley project offers something rare: published performance data from a real deployment rather than a vendor demo. Digital twin technology has been discussed widely in UK transport planning, but few projects have moved past the proof-of-concept stage with measurable outcomes.
The Green Party’s response was cautiously supportive but flagged a structural limitation. Vehicles on North East roads clocked up 13 billion miles in 2024, a rise of more than 1.5 billion over the past decade. Green councillor Matthew Snedker argued that AI traffic optimisation cannot solve congestion driven by increasing car volumes, comparing the approach to “buying bigger trousers to solve obesity.”
Looking forward
The next phases will expand the digital twin to cover additional routes and incorporate freight, active travel, and environmental data. The project’s value to other councils will depend on whether those expanded datasets improve performance further, and whether the cost of implementation makes sense for smaller authorities with fewer congestion hotspots.