An improved TAPAS algorithm for the traffic assignment problem

Recently, Bar-Gera proposed a new algorithm for the traffic assignment problem (TAP) that promises to produce highly precise solutions by identifying and keeping a list of paired alternative segments (PAS) on origin-based subnetworks. The algorithm was named the traffic assignment by paired alternative segments (TAPAS) algorithm. This paper presents an improved TAPAS algorithm (termed iTAPAS) that enhances the original TAPAS mainly in two aspects: (1) a new PAS identification method is proposed; (2) each PAS is only associated with one origin. iTAPAS not only accelerates the solution convergence performance by improving the efficiency of the PAS identification and flow shift operations, but also simplifies the algorithm implementation by removing the PAS matching and branch shift steps. In this paper, detailed algorithmic procedures of TAPAS and iTAPAS are discussed. The numerical results from our implementation and comparison reveal that iTAPAS is nearly two times faster than TAPAS in solving large-scale networks. An in-depth interpretation of the comparison results is also subsequently followed by examining the PAS set information of both algorithms.