Population Synthesis by Disaggregating OD Matrices with Time-Progressive Graphs for Agent-based Simulations

Abstract

In the last decades, a shift from traditional travel demand modeling approaches to more fine-granular agent-based simulations occurred. This transition was supported by technological advances in computation capabilities and research regarding the capture of individual travel behaviors. While agent-based simulations have been well established, Origin-Destination (OD) matrices are still widely adopted, especially since much historical data is available in these formats. Furthermore, defining models with OD matrices requires less input than detailed microsimulation models. OD matrix-based modeling frameworks may be adequate for mostly car-and public transit-oriented traffic models; however, OD matrices fail to capture the interaction between different trips and do not regard the traveler as a central entity for travel decisions. These limits become especially apparent when regarding intermodal trips. Hence, we want to bridge the gap between traditional OD matrix-based travel demand analysis and modern agent-based simulations. We propose to compute realistic activity schedules from OD matrices that we refine with external statistics. The algorithm converts OD matrices into a graph, where each path represents a valid activity schedule. With a modified depth-first search, we extract activity schedules from the graph and assign them to agents. The practicability of the algorithm is demonstrated by a runtime analysis and a case study. The results show that the algorithm computes activity schedules while keeping the original flow from the OD matrices.