Aurora

About the research

Friction is the ultimate metric for measuring the ability of a driver to control a vehicle on the road and inclement weather is the primary factor that influences roadway friction. Many state departments of transportation (DOTs) use roadway friction measurement devices as guidance for alerting the traveling public and for snow removal activities. However, these devices are not available universally along the highways and large gaps in friction information can result, especially where friction measuring devices are stationary.

An increasing number of mobile friction measurements are becoming available, from DOT fleet vehicles and more recently from private vehicles. These measurements provide additional road (friction) state given current weather, but more importantly provide an opportunity to create improved modeling of friction impacted by weather events. Improving highway friction forecasts using forecasted weather conditions would be beneficial to state DOTs for a variety of planning purposes, including Variable Message Signage, Variable Speed Limit adjustments, chain-up and chain-down timing and more. Improved friction modeling may also allow more accurate estimates of friction conditions along roadways where friction measurements are sparse.

The objective of this research project is to gather friction measurements from stationary (RWIS) and mobile (vehicle) sources as well as concurrent weather data and then develop a predictive roadway friction machine learning model. The project team expects to gather the appropriate data and develop a new machine learning model during the first year of this research, then test the new model on weather forecast data from the following winter season in the second year. A summary of the project’s efforts will include an examination of the friction model performance as well as a comparison of forecast quality between stationary and mobile friction locations.