Traffic Simulations: Multi-City Calibration of Metropolitan Highway Networks
Chao Zhang, Yechen Li, Neha Arora, Damien Pierce, Carolina Osorio
TL;DR
This work addresses calibrating origin-destination demand for metropolitan highway networks using high-resolution stochastic traffic simulators and abundant path travel-time data $y_p^{GT}$. It introduces a metamodel optimization that couples a differentiable surrogate $m_k(x;\beta_k)$ with a physics-based component $f_A(x)$ to approximate path travel times, enabling efficient optimization with few simulator calls. The approach is demonstrated through six-city, nine-hour calibrations (54 scenarios), achieving an average improvement of $43.5\%$ in ETA nRMSE over the SPSA benchmark and up to $80.0\%$ in some cases, illustrating strong scalability and robustness to initial points. The results underscore the practical potential for digital twins in mobility planning, offering scalable, data-efficient calibration that leverages readily available path ETAs to support congestion pricing, network design, and policy evaluation; future work includes multi-time-interval calibration and higher-order moment analysis of path ETAs.$
Abstract
This paper proposes an approach to perform travel demand calibration for high-resolution stochastic traffic simulators. It employs abundant travel times at the path-level, departing from the standard practice of resorting to scarce segment-level sensor counts. The proposed approach is shown to tackle high-dimensional instances in a sample-efficient way. For the first time, case studies on 6 metropolitan highway networks are carried out, considering a total of 54 calibration scenarios. This is the first work to show the ability of a calibration algorithm to systematically scale across networks. Compared to the state-of-the-art simultaneous perturbation stochastic approximation (SPSA) algorithm, the proposed approach enhances fit to field data by an average 43.5% with a maximum improvement of 80.0%, and does so within fewer simulation calls.