Cross-entropy optimization model for population synthesis in activity-based microsimulation models

Abstract

Microsimulation techniques for disaggregate activity-based travel demand models are expected to synthesize a desired number of fully specified individual activity patterns. A data prerequisite for this technique is the so-called synthetic population, a key input for activity-generating procedures. The iterative proportional fitting method has been widely used to estimate the multiway demographic tables for different geographic areas. This study proposed a cross-entropy optimization model in which generalized constraints for different demographic characteristics of the synthetic population could be included. A quasi-Newton algorithm was devised to solve the proposed problem. Encouraging results obtained from the model application suggested that the proposed method held much promise for generating a more realistic synthetic population with different types of demographic characteristics and could be generally applied in different geographic areas.