Monianturipohjainen tienpinnan tilan arviointi

Road weather describes the atmospheric characteristics and road conditions that affect the safety of road traffic. It is the result of local weather conditions, the characteristics of the built environment, and road operations, including traffic and winter maintenance. Road weather stations measure different aspects of road weather using various atmospheric and road surface sensors, providing road traffic authorities and operators with actionable information on road conditions. In cases where precise measurements of surface layer thicknesses are not available, road weather stations use internal models to estimate derived quantities, such as road state classification. The overall objective of this thesis was to improve the road state classification provided by road weather stations equipped with a limited set of sensors. A simplified physics-based model was implemented, combining atmospheric measurements and measurements from the road surface, such as road salt and the electrical properties of the surface, with concepts from existing road weather models. Reference data collected from two road weather stations using state-of-the-art remote road surface sensors were used to evaluate the model. In addition, particle swarm optimization was studied as a method for fitting the model’s parameters to the reference data to improve the classification accuracy for individual road weather stations. The developed road state model provided a reasonable classification accuracy, demonstrating the applicability of physics-based modeling for estimating the present road state using local observations alone. Major sources of uncertainty included winter road maintenance and determining the exact composition of water during winter. Particle swarm optimization showed potential for automated tuning of model parameters; however, further work is required to achieve consistent results.