Computing Dynamic User Equilibrium on Large-Scale Networks Without Knowing Global Parameters

Abstract Dynamic user equilibrium (DUE) is a Nash-like solution concept describing an in dynamic traffic systems over fixed planning period. DUE challenging class of problems, connecting network loading models and notions system one concise mathematical framework. Recently, Friesz Han introduced integrated framework for computation on large-scale networks, featuring basic fixed-point algorithm the effective DUE. In same work, they present open-source MATLAB toolbox which allows researchers to test validate new numerical solvers. This paper builds this seminal contribution, extends it several important ways. At conceptual level, we provide strongly convergent algorithms designed compute directly infinite-dimensional space path flows. An feature our that give provable convergence guarantees without knowledge global parameters. fact, propose are adaptive, sense do not need priori parameters delay operator, even operators non-monotone. We implement schemes standard instances, compare them with strategy employed by Han.