Advanced optimization methods for road construction

点击翻译按钮获取中文摘要

Road construction represents a significant portion of supporting an efficient road network. Highway quality roads cost as much as $20 million per km while forestry roads costs from $1,000 to $1 million per km. Reducing road construction costs while maintaining safety standards is very desirable. Road design has also become increasingly more complex. It now has to take into account environmental impact, mitigation of the impact of natural or man-made disasters. Reducing road construction costs may play a major role in controlling future budgets and maintaining quality of life, both in Canada and in any developed country. The road design problem is usually split into three subproblems. First, the horizontal alignment considers a satellite view of the potential road site, selects a handful of corridor alternatives, and determines the exact path within that corridor. Next the vertical alignment considers which hill to cut or ravine to fill to build a road that goes through the horizontal alignment while satisfying safety and regulation constraints. Finally, the earth moving problem considers how much material to move to realize the target vertical alignment. Our goal is to automate much of the current manual process to save design engineers time on computational tasks so (s)he can concentrate on considering as many constraints as possible. We plan to fully automate the full road design process from corridor selection to earthmove decisions. Our approach focuses on improving computation time to solve the vertical alignment model so we can perform an extensive search for the cheapest horizontal alignment. We will incorporate additional features into current models to more precisely reflect the building process. While the complete problem is out of reach of current state-of-the-art optimization solvers, the latest optimization algorithms coupled with carefully approximation will provide much chearper road designs, allow engineers to consider a wider range of alternative paths, and provide insight into the robustness of the solution. The resulting models and algorithms will be implemented and deployed in a commercial road design software, and made widely available.