A SimOpt approach to balance consistency of historical OD matrices and accuracy of traffic counts

Origin-Destination (OD) matrix calibration plays a crucial role in transport modeling, as it ensures that travel demand estimates are consistent with historical data and accurately reflect empirical observations. Therefore, we propose a refined Simulation-Optimization approach that improves the accuracy of transport simulation assignments with respect to observed traffic flows, while preserving the consistency of historical demand patterns. A key aspect of the calibration process is determining the parameter α, which governs the trade-of between the historical OD matrix and the observed traffic count data. Our proposed approach identifies the value of α that better maintains both the integrity of historical OD matrices and the accuracy of simulated traffic flows, by employing a bi-objective algorithm integrated with a macrosimulation model developed with PTV Visum tool. We also introduce a novel evaluation method for the calibration procedure that considers the variation of the α value and jointly considers cosine similarity and GEH statistics. We tested our method on a real-world case study focused on the Catania (Italy) transport network, demonstrating its robustness, scalability, and practical applicability. Notably, numerical results indicate that the optimal value of α does not necessarily correspond to the intuitive midpoint (α = 0.5), but instead reflects a more complex balance between historical consistency and empirical accuracy.