Through workshops with Highway’s England staff and visits to the regional network control centre, we developed the technical scope for the feasibility study. This was followed by an extensive data gathering exercise to bring together the different forms of data pertaining to the M40 motorway network area which formed the basis for the analysis.
Data included network road sensors, CCTV camera, satellite data and relevant meteorological information. On top of this we spent time reviewing technical literature on previous attempts to predict fog using different data sources and how appropriate they were given the spatial and temporal specifications needed for a solution to be utilised by Highway’s England.
Modelling work was undertaken using a combination of different analytical techniques. A comparison of COBS fog messages with weather station data, CCTV and traffic flow data suggest that these data provide a promising basis for the development an improved fog prediction system. In particular, it was discovered that a potential ‘fog signature’ in the data could form the basis of a future machine learning system (see image below). Visibility measurements also showed potential for alerting drivers to the risk of fog several hours before it can be reported by officers on the road.
The project demonstrated that advances in data fusion provide a ready-made framework for the development of a data intensive ‘intelligent’ fog prediction system using machine learning to provide early warning alerts well in advance of likely reporting times provided by traffic officers on the network. This intelligent system for fog prediction implemented ‘by software’ as opposed to installing roadside equipment would be universally deployable and would offer the potential for improving safety and reducing the operational costs required to maintain and service new roadside hardware.
Tell us about your data challenge