Improved Partial Linearization Algorithm for Solving the Combined Travel-Destination-Mode-Route Choice Problem

Combined travel demand models (CTDM) that integrate trip generation, trip distribution, modal split, and traffic assignment have been developed to resolve the inconsistency problem between the level-of-service and flow values of the sequential four-step travel demand forecasting procedure. In this paper, an improved partial linearization algorithm for solving the logit-based combined travel-destination-mode-route choice model formulated as a convex mathematical programming is developed. The improvements mainly focus on exploring recent advances in line search strategies to minimize the computational efforts required to determine a suitable stepsize that guarantees convergence. Specifically, the quadratic interpolation and the self-regulated averaging schemes are examined. Numerical results show that the self-regulated averaging line search scheme is more effective and efficient for solving the convex mathematical programming with a complex objective function in terms of solution quality and computational effort.