Vaisala Xweather and NIRA Dynamics have developed a dataset that combines weather forecasting and computer vision with real-time data, writes Nick Flaherty.

The integration draws on data from NIRA Dynamics from billions of data points gathered from connected vehicles. This is paired with Vaisala Xweather’s machine-learning road-weather forecasting models, which achieve accuracy levels that are 50% higher than those of publicly available forecasts.

Autonomous driving will become increasingly dependent on accurate, real-time weather and road condition data.

Automated driving functions are currently disconnected based on crude weather and road-condition estimations, such as a temperature drop to under 4 C or a change in tyre rolling resistance.

Real-time data from connected vehicles on tyre grip, road friction, surface quality and weather conditions are combined with weather forecasts to give an aggregated dataset. This is then used with computer-vision algorithms to assess road conditions to optimise maintenance efforts such as minimising winter road salt.