Comparative Analysis of Evolutionary, Local Search, and Hybrid Approaches to O/D Traffic Estimation

Abstract

This paper compares the performance of various advanced evolutionary algorithm (EA) techniques to solve the problem of dynamic origin/destination (O/D) estimation. The potential of EA in the dynamic O/D estimation problem lies in their powerful global search and optimization capabilities. This EA-based demand estimation framework is implemented into a model that we call dynamic O/D estimator (DynODE). DynODE is integrated with an existing dynamic traffic assignment (DTA) platform (i.e., Dynasmart-P). The EA-based methods in this research are further augmented with EA parallelization as well as hybridization schemes to further improve the quality and efficiency of the solution. In this paper, we evaluate the performance of the basic EA model, the hybrid EA model, the parallel EA (PEA) model, as well as the parallel hybrid EA model. Additionally, the performance of a local-search method is examined and compared with the EA-based approaches. The comparison is performed on a medium size network. The results from the study show that the PEA model outperforms the other algorithms in terms of speed as well as solution quality. However, all hybrid and PEA runs result in savings in computation resources as well as enhancement in the quality of solution as compared to the basic EA model. Still, the basic EA model outperforms the local-search model.