The column generation (CG) algorithm has been widely applied to traffic assignment problems due to its capability of circumventing path enumeration. Incorporating bounded rationality (BR) and dynamics, this paper proposes four tolerance-based strategies for extending the CG algorithm to the bounded rational dynamic user equilibrium model (BR-DUE): (i) a tolerance-based minimum disutility path search strategy is developed to allow travelers seeking satisfactory paths; (ii) a self-adjusted convergence threshold strategy is applied for fast convergence at the intermediate iterations; (iii) a varied temporal resolution scheme, combining exploration and exploitation, is suggested to assign flows to narrow time regions rather than to the whole time horizon; and (iv) a path search skipping strategy is introduced by comparing the lower bound of travel disutility and the minimum disutility of used time-dependent paths. With these strategies, an efficient tolerance-based column generation (TBCG) algorithm for BR-DUE is developed. Numerical examples are provided to illustrate the efficiency of the TBCG algorithm, of which the speedup factor is around two compared with using the orginal CG algorithm.
|Title of host publication||Proceedings 98th Annual Meeting of the Transportation Research Board|
|Publication status||Published - 2019|
|Event||98th Annual Meeting of the Transportation Research Board - Walter E. Washington Convention Center, Washington, D.C., United States|
Duration: 13 Jan 2019 → 17 Jan 2019
|Conference||98th Annual Meeting of the Transportation Research Board|
|Period||13/01/19 → 17/01/19|