Robust Traffic Assignment via Convex Optimization

The static traffic assignment problem with deterministic demand is often formulated as a linear, or, more generally, convex optimization problem. It has long been recognized that various uncertainties may affect the input data, such as origin-destination demands, or network topology. In turn, these uncertainties may greatly deteriorate the optimality of solutions to the traffic assignment problem. Thus, it is desirable to obtain a traffic assignment that is robust with respect to uncertainties affecting the model. Recently, new approaches to decision-making under uncertainty have been proposed, under the name of robust optimization. The methodology has been successful in many areas of engineering, such as communications, filter design, control systems, and also in machine learning and statistics. The goal of this project is to evaluate the potential benefits of using a robust optimization approach in the context of traffic assignment, both for static and dynamic problems. It is expected that the approach will provide a traffic assignment methodology that provides solutions that are far more robust than the original ones, yet give up relatively little in terms of performance.