Amey’s Smart Winter initiative uses Artificial Intelligence models to predict road conditions 24 hours ahead of time, analysing geographical data along with road surface temperature data from a network of sensors. The accuracy rate is 90 percent to within 1ºC. This data analysis has allowed for a wholesale redesign of Kent’s gritting domains to maximise their effectiveness and efficiency through targeting resources to stretches of road that need them most. Unnecessary, ‘surplus’, gritting of roads that do not freeze is reduced and more susceptible routes are treated more quickly. This creates safer conditions for communities across the county and keeps people moving, benefiting both individuals and the local economy.
The Smart Winter modelling collaboration with Kent County Council was first trialled in 2018 but fully launched in winter 2019/20. More recent initiatives, including the re- programming of gritter paths to optimally traverse each domain, have further enhanced the programme. Additionally, through its new Live Labs collaboration, Amey Consulting has also implemented a new digital highways platform (‘HADMS’) for management of Kent’s estate and assets, which includes monitoring performance and compliance of Kent’s gritting vehicles. HADMS is now used operationally by Kent to actively monitor performance of its winter fleet operations.
The programme came out of Amey’s strong partnership with Kent County Council. The Council had the vision to recognise the emerging potential presented by using new active in-field sensor hardware to bring intelligence to maintenance operations.
As a trusted partner of the Council, Amey suggested leveraging these new sensors alongside the use of micro-weather forecasting and geospatial data to train an AI model that would enable prediction of road surface conditions. The two immediate objectives for this piece of work were to provide more accurate and granular forecasts that account for localised factors and to enable these to be produced across the entire road network rather than being limited to only those locations with active physical sensors.
The key objective for Kent was to make sure it was using its gritting fleet effectively. Gritting is an essential but costly service, and it is critical to deploy resources in the most cost-efficient way while also ensuring the safety of road users. The Amey solution allows the Council to achieve both by making more effective, intelligence-led gritting decisions. In 2019/20 Kent laid down around 12,500 tonnes of grit. This, along with labour and plant costs, represents a significant expenditure.
Additionally, Amey has created a digital model of Kent’s gritting network which has been used to simulate and optimise the locations and movements of its gritting fleet. This has been adapted so that the Council’s winter services team can test alternative scenarios such as varying the salt spread rate or excluding specific routes to suit forecast weather conditions. Updated routes can then be uploaded to the gritting fleets’ navigation systems.
The HADMS compliance monitoring platform is integral to applying the findings of the data analysis such as new gritting routes or domains. This, alongside rolling out changes to Kent’s fleet navigation service provided by Exactrack, means that coverage and performance can be continuously monitored and any issues or areas for improvement identified immediately.
In total there are over 100 sensors located across the Kent route geography capturing road surface condition data. These were installed by Winter Sense with Exactrack providing the vehicle tracking and navigation system. The data gathered by sensors sits alongside micro-weather data as well as geospatial and terrain data that is used to train the AI model and predict route surface conditions. This micro-weather data was provided by the Dark Sky service using weather station and other data to generate granular weather statistics.
The AI model applies machine learning to adapt to the data provided by these sensor readings and other sources (weather, terrain etc.). The model uses a very high volume of samples – in this case gathered over the course of winter 2019/20 – to learn the physical response behaviour of these sensorised locations on the network. Once trained in this way, the model can then be readily applied to other non-sensorised locations by leveraging the same source data, allowing Kent to forecast route conditions across the full extent of its road network.
These systems are ground-breaking in the winter maintenance space. Gritting domains have been changed and gritting routes revised to maximise effectiveness and efficiency. The Amey team is now in the process of digitising new gritter paths to more accurately fit each domain, and is working with Kent County Council and its weather forecast provider DTN Meteogroup, to help them move towards route-based rather than domain-based gritting decisions across the road network. This means that rather than gritting entire sections (domains) of the network, Kent can potentially pinpoint and treat only those routes and sections of road that need treating. This can be supported by adaptive routing and navigation, opening up the possibility in the future for the county’s gritting fleet to be programmed to tackle different permutations of route forecasts, improving flexibility and allowing for real time reaction to changing conditions.
These changes place data at the centre of decision making and together they represent a step change in the winter management of Kent’s road network. This new approach delivers benefits all round. Resources are used efficiently, targeting gritting where it is most needed while necessary road treatment is reduced. But most importantly it benefits communities across the county, helping to keep road users safe and traffic moving freely over thought the winter months.
Mark Fisher, Principal Strategic Consultant, Amey Consulting