Forecasting road maintenance need with AI and IoT

Northern cities spend millions on winter maintenance each year to ensure roads are safe and traffic can operate efficiently. We interviewed road maintenance personnel to identify operational challenges they face in winter... [ view full abstract ]

Northern cities spend millions on winter maintenance each year to ensure roads are safe and traffic can operate efficiently. We interviewed road maintenance personnel to identify operational challenges they face in winter maintenance.

The main challenge we identified was predicting road slipperiness. Gaining an accurate overview of the road network slipperiness is difficult as ice and water tends to be localized. Day to day maintenance need assessments are done by surveying roads manually from a vehicle. As road networks are thousands of kilometers long, whole networks can't be surveyed. Instead the overview at any given time is based on an estimation. Thus sudden, localized changes in road condition may go undetected for days due to surveying methods. While traditional weather forecasts give a fairly accurate overview of long term weather in a large area, they are unreliable for short term district or street level forecasts according to winter maintenance workers.

We propose a novel solution based on cutting edge AI research as well as IoT- & open data and cloud computing. Our model has been tested on data from Oulu and Helsinki getting recall rates of 98.7% and 80.4% respectively when predicting slipperiness. For comparison the recall rate of the Finnish Meteorological Institute's forecast is 0.7%. The data we have used is roadside weather station data, road network topographies, vehicle amounts and most importantly IoT measurements from vehicles. Our proposal includes a cloud platform for displaying the forecast and current overview on a webmap as well as subscribing to mobile notifications.

We expect this to provide benefits for cities, public and private street maintenance contractors as well as citizens. Specific benefits include better surveying prioritization, easier supervision of results, ability to proactively focus maintenance resources, efforts and procedures. As a result we expect the system to reduce accidents and damages, lower costs and increase wintertime street quality.