2023 Volume 79 Issue 4 Article ID: 22-00301
This paper presents an efficient algorithm for solving dynamic user equilibrium (DUE) traffic assignment problems with one-to-many origin-destination demand. We first formulate the DUE problem, with the point queue model, as a linear complementarity problem (LCP). Then, we transform the LCP into a quadratic programming problem (QP). The QP enables us to build an efficient algorithm based on the convex combination method (Frank–Wolfe algorithm). Numerical experiments reveal that the proposed algorithm can (i) solve extremely large-scale problems to which no conventional methods can be applied, (ii) calculate the solution more than 1000 times faster than a general-purpose solver for optimization problems, and (iii) yield an extremely accurate equilibrium solution.
