Back to Insights


Mobile & Machine Learning - Predicting Potholes Before They Happen

Let's face it, nobody likes potholes! Road defects are a huge issue for local authorities throughout the UK for both cost and safety reasons. British motorists are spending a whopping £1.7bn a year fixing damage to their vehicles caused by potholes. The average cost for a pothole repair in the UK is currently £158 according to a recent survey conducted by

According to the 2017 Annual Local Authority Road Maintenance Survey it would cost a staggering £12.6bn and take over 15 years to fix the current backlog of pothole repairs- an increase from 11.8bn in 2016. Despite the government recently announcing it would offer UK councils an extra £100 million a year to deal with road defects, this is ultimately a drop in the ocean compared to what it costs motorists.

There is currently no standardised approach among local authorities for tackling these road defects. Some authorities rely solely on manual data collection where staff are sent out in a vehicle to record potholes and others simply rely on members of the public reporting defects or making complaints.

In Scotland, local authorities have implemented Mobile LiDAR technology as part of the SCANNER survey, which is a sophisticated piece of machinery that is able to detect the condition of a road surface. However, this is only deployed a couple of times a year and costs over £750,000 a year to run. Whilst it covers major local authority roads annually, less popular roads are not scanned for 2-4 years, with some not receiving a survey for up to 10 years. This is not an efficient or cost – effective method for longer term monitoring or for predicting road defects.

The problem is that most repairs carried out by local authorities are reactive. If potholes or road defects are detected earlier - or better - predicted to form, councils can carry out lower cost repairs at that point rather than incurring substantial costs for resurfacing the whole area or a full-depth structural repair.

The Solution

In 2016, xDesign won a challenge with CivTech Pilot Accelerator and Transport Scotland for funding to drive innovation in the public sector. xDesign rose to the challenge to create a state of the art mobile application that leverages big data, and machine learning to predict potholes and road defects before they appear. This led to the inception of “Road Intelligence”- an xDesign spin off company in 2017.

There were a number of important collaboration partners involved in building the application including The Data Lab, EPCC, Construction Scotland Innovation Centre & The University of Edinburgh Seismology Department.

The Road Intelligence mobile application works by collecting accelerometer and gyroscopic data on every ‘bump’ the vehicle encounters from rough roads to larger potholes. This road defect data is then profiled and then fed into a cloud based machine learning model which runs proprietary algorithms to analyse the data, grade defects for severity and provide physical profiling.

The app can be deployed to multiple users across the road network-(for example engineers, staff, contractors or members of the public) which means the network can be analysed much more frequently than it is currently. This will allow local authorities to be much more proactive when it comes to fixing road defects through continuous data streaming.

How Does It Work?

Road intelligence is available for Android or iPhone. There are 2 versions - passive (place in car and go) and active, which allows engineers to collect additional details about road defects should they require it. - Drive

As the vehicle is being driven we use the accelerometer, GPS and gyroscopic data to capture information regarding road defects. - Data transmitted in the background

A single defect is over 250 points of data. All of this is then sent to the cloud which allows us to use an infinite amount of power to process and make sense of the data. - Cloud computing crunch numbers

The app then attempts to match the patterns of data sent from the vehicle against known patterns of defects and non defects (eg. speed bumps) and grades the size. - Machine based learning

As multiple vehicles travel over multiple defects, the system improves automatically through the continuous stream of data, both for specific defects and as a system overall. - Visualisation

The data collected is the displayed visually in a web portal - mapping your entire road network- showing the status of individual roads and information about defects.

How Will Road Intelligence Help?

Road Intelligence is designed to continuously collect data to predict potholes before they are even formed. This will help local authorities and road network organisations to substantially reduce their costs through decreased compensation pay-outs and maintenance costs. It will also help to reduce the amount of citizen complaints, limit the number of road accidents and help authorities gain a better understanding of their road network infrastructure.

We are currently in the process of piloting Road Intelligence in Scotland with a number of local authorities and we are looking forward to continuously test the technology and analyse the results!

Want more information? Check out the website to schedule a demo or contact us and we'd be happy to talk you through it in more detail!