Road De-icing under a Rolling Horizon

Abstract: Road de-icing is part of the winter road maintenance sector, an area of interest to the Uppsala-based company BM System. In order to solve the road de-icing problem, BM System developed an optimisation engine which makes use of a local-search algorithm. In a given geographical area, new data is generated periodically from hourly weather forecasts, creating the necessity to calculate a new solution to the road de-icing problem with each new incoming forecast. The aim of this thesis is to take a different approach of solving the road de-icing problem. We describe a simplified version of the original problem as a constrained optimisation problem and propose a modelling approach towards finding a solution. We design a model and extend it with initial-solution heuristics, which help improve the quality of the solutions found, for example by making use of the rolling-horizon timeline of the weather data. We evaluated the model experimentally on real-life data under different comparison points. The solutions calculated by a simplified version of BM System's optimisation engine were provided as a reference point for comparison. Results show that our approach for solving a simplified version of the road de-icing problem is competitive in both solving time and solution quality with the company's simplified optimisation engine. In the future, it can be extended to encompass more complex versions of the problem that appear in real life.