The hi-tech battle to banish Britain’s potholes 

UK roads are poorly designed and badly built. Could the answer lie in AI — which predicts problems before they appear — and rupture-fixing robots? 

Rhys Blakely , Science Correspondent , The Times. Friday, February 14 2025

There are three big issues with the UK’s roads, according to Ioannis Brilakis, an engineering professor at Cambridge University. First, they are poorly designed — to lower specifications than those used in other countries. Second, they are built as cheaply as possible by contractors beholden to an “it meets the specifications” philosophy that demands they do as little work as they can get away with. Finally, almost all maintenance is reactive, with repair crews deployed only once potholes have reached bone-shuddering proportions. 

“As a result, the roads here degrade a lot faster than in other countries,” he said. 

His analysis may not surprise motorists as they bounce along the pockmarked streets of Britain and while Brilakis would like to see them built to a higher standard in the first place, he and his team are also working on another solution. The first steps will include Cambridge becoming the first city in the UK to have every road scanned and then scrutinised by an AI that is being trained to predict where potholes will appear. 


After that, the aim is to build a road-repairing robot that can mend small problems before they turn into big ones. 

Potholes can be fatal, with more than 250 cyclists either killed or seriously injured due to road defects between 2017 and 2023. According to the AA, they have contributed to a surge in vehicle breakdowns, and the annual cost of pitted and worn-out roads to the economy was estimated last year at £14.4 billion by the Centre for Economics and Business Research. 

Big lorries do lots of the damage, but the problem is likely to worsen as electric cars, which are around 30 per cent heavier than conventional ones, become more common, and as climate change brings wilder weather. 

Yet local authorities still largely rely on the public to point out potholes. The authority will then typically ask a contractor to repair them. “So they get in the van, they go there with their tools to fix the pothole,” said Brilakis. 

“And when we’ve spoken to them, they reported to us that on the way to the pothole, they see another 30 potholes, but they’re not supposed to touch them. And then, quite often, when they get to the pothole they’re supposed to fix, it’s not there — because it’s been misreported. This is a frequent thing. The council might have very limited resources, and those tiny resources are being deployed inefficiently.” 

Identifying damage before it evolves 

On average, all classes of roads in England and Wales are resurfaced once every 80 years, while the typical lifetime is roughly 20 years. Brilakis and his team hope to change how life-extending maintenance is carried out. Working with the Greater Cambridge Partnership, they will scan the roads of central Cambridge, using a system mounted onto a special vehicle that combines lidar, which uses pulses of laser to measure distances, with cameras. 

His group have developed AI that can analyse this data and identify different types of damage. It can distinguish, for instance, between potholes and the cracks from which they evolve. It can also highlight “rutting” (the grooves that form in wheel tracks), shoving (bulging and rippling where vehicles pull their brakes) “spalling” (where the surface of a concrete road flakes away) or “bleeding” (where patches of shiny asphalt form on the surface, meaning that tyres lose grip). 

The end result will be a “digital twin” of Cambridge’s roads where every defect should be mapped. It will be put to a number of uses. One will be testing the accuracy of another source of information on the state of the UK’s roads: the mountains of data gathered by modern sensor-laden cars. 

Equipped with their own cameras, motion detectors, radar and lidar systems, a family saloon can now collect a continuous stream of information on the state of a highway, from rutting and shoving to fallen street signs and damaged lane markings. 

Car manufacturers are already selling this data to governments in other countries. Potentially, it could be analysed by the AI and used to constantly update a master digital map, flagging each fault in each UK highway as it appears. Brilakis and his team will investigate how useful and accurate it really is. “Imagine a world where, rather than us inspecting the roads once a year or once every two years, as soon as three vehicles drive on top of a specific defect, it’s already detected and reported,” he said. 

Fixing potholes before they appear with the help of AI 

If the AI can use the car data to spot the minor cracks that precede potholes, targeted proactive maintenance could be possible, which should be cheaper than reactive repairs. 

In another laboratory in Cambridge, Professor Abir Al-Tabbaa and her team are researching new “self-healing” materials that could be used. Potholes appear when water enters a small crack, freezes and expands, breaking up the asphalt used to coat the road. Part of the challenge is producing a material that is fluid enough to penetrate a very small crack, sets quickly and is strong enough to seal it for years. “Cars have advanced dramatically in recent decades; the roads have not,” Al-Tabbaa said. 

One solution that she and other researchers have looked at involves mixing tiny capsules containing a “healing agent” with the asphalt. “A crack will just propagate through these capsules and rupture them and the healing product comes out and seals the crack,” she said. Another approach, showing promise in her group’s lab, involves adding glass fibres to the sticky bitumen used to fix early-stage cracks. According to Dr Damian Palin, tests show strength improvements of around 20 per cent. “We’re trying to minimise the number of times we have to repair the roads,” Al-Tabbaa said. “Because, ultimately, you want to stop road closures. That’s where most of the carbon emissions [related to road building and upkeep] came from – diversions and traffic congestion.” 

Pothole-prevention robots — ‘It’s like a health check’ 

The end goal is to automate the process. Robotiz3d, a spin-out construction company from the University of Liverpool, is developing a self-driving robot designed to spot, categorise, and repair road defects. It uses a “laser profilometer” (another method of using a laser to measure distances) to create highly accurate 3D profiles of potholes and other flaws, with millimetre-level precision. The data can then be analysed by an AI to assign severity levels and prioritise repairs. 

There are still regulatory hurdles to overcome before fully autonomous robots can roam the streets repairing roads, but a Robotiz3d prototype recently tested by Hertfordshire County Council carries its own bitumen-based material to seal cracks proactively. 

It can travel at 50-60mph when scanning for damage, but slows to about 5 mph when performing what Lisa Layzell, the company’s chief executive, calls “precision surgery” on the road, filling small fissures. She sees a future where fleets of these robots are sent out at night to mend minor damage before it worsens. 

“That’s our vision: smart cities, where you have these robots roaming the streets. When they find a crack, they’ll stop and repair it,” she said. 

“We know 75 per cent of the cracks will evolve into potholes in two to three years. If you fill them, only 1 per cent will continue to deteriorate. It’s like a health check: if you find small problems and intervene early, they don’t evolve into life-threatening problems later on. And it’s much cheaper.”