Robust optimization approach for system optimal dynamic traffic assignment with demand uncertainty

This research deals with the path-based system optimal dynamic traffic assignment (SODTA) problem with uncertain demands that are assumed to be bounded by a prescribed uncertainty set. A robust optimization approach is adopted to address this problem. The objective is to minimize the total network travel time under the worst-case scenario defined by a pre-determined demand uncertainty set. We formulate the robust counterpart optimization problem of SODTA for a general uncertainty set and show that solving the robust SODTA problem is not more difficult than solving the deterministic SODTA problem for some specific types of uncertainty set. Moreover, a column generation-based algorithmic framework that embeds a scaled gradient projection algorithm is proposed to solve the SODTA problem. Numerical experiments were conducted to demonstrate the effectiveness of the algorithm and to examine the impact of different types of demand uncertainty set on solution quality.