Award-winning AI project keeps roads safe through winter
A paper on Smart Winter work submitted by Kent County Council and Amey has won an award and received an honourable mention at the XVI World Winter Service and Road Resilience Congress hosted in Calgary.
The Congress, which tackles topics such as new technologies and methods in winter service, recognised the innovative work on the ‘AI Surface Temperature Prediction Model’ which predicts road surface conditions over the winter season.
The model trials a machine learning approach to predict road surface conditions over the 2018/19 winter season. It uses local geographical, traffic usage and weather forecast features along with target road surface temperatures provided by a network of 112 RST sensors installed across Kent’s gritting route estate.
The model is shown to predict route surface temperature to a 90 percent accuracy to within 1ºC, and the paper presents a practical application of the model that enables service operators to offset its precision (efficiency) in favour of recall (coverage), which is important for maximising safety outcomes on the network.
The weather forecast features – in particular, air temperature – were shown to be the most influential features for prediction, but local geophysical features are also shown to influence the model, especially on more challenging marginal nights. Accordingly, the model can be easily applied to produce route-specific forecasts that account for localised factors, enabling authorities to take more nuanced and targeted gritting decisions that make more effective use of limited and costly gritting resources.
The trained model is also extrapolated across the complete estate. A clustering algorithm is applied to re-formulate all of Kent’s gritting domains to maximise their consistency, targeting gritting resources to regions with uniform freeze response that will minimise ‘surplus’ gritting of un-frozen roads. These new domains have now been trialled operationally by Kent for the 2020/21 winter maintenance season.
Kent County Council’s (KCC) Highways Project Manager, Carol Valentine, said: “This project is truly innovative and gives us the opportunity to enhance our services by looking back at previous years.
“Using that data, we can make sure we are getting the most out of our resources while ensuring KCC’s road network is well equipped for whatever may come.
“It is a privilege that our work has received this recognition”.
Principal Strategic Consultant at Amey Consulting, Mark Fisher, said: “As a trusted partner of Kent Council, Amey has had a great opportunity to apply its AI and machine learning expertise to a practical and worthwhile initiative. We have clearly demonstrated the potential offered by the wealth of data, now increasingly available within local highway operations. It's an honour to receive recognition for this work which opens up exciting future possibilities for bringing further intelligence and flexibility to Kent's winter operations”.
The work was awarded best paper in the Winter Service category, presented virtually to Carol Valentine, Highways Project Manager Live Labs and Mark Fisher, Amey Consulting’s Principal Strategic Consultant, by Baroness Vere from the DFT at the PIARC UK Winter Awards Ministerial presentation.