Calibration of Vehicular Traffic Simulation Models by Local Optimization*

Davide Andrea Guastella, Alejandro Morales-Hernández, Bruno Cornelis, Gianluca Bontempi · 2025

Read Calibration of Vehicular Traffic Simulation Models by Local Optimization* by Davide Andrea Guastella, Alejandro Morales-Hernández, Bruno Cornelis, Gianluca Bontempi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-11585.txt.

Introduction

Traffic simulators help cities test road changes before spending money, but the simulator must first be tuned to match real counts. Davide Andrea Guastella, Alejandro Morales-Hernández, Bruno Cornelis, and Gianluca Bontempi present a tuning method in 2025 that works locally, street by street.

Their stochastic method splits the calibration task across neighborhoods, which lets it scale to large networks and approach real-time speed, a step toward live digital twins of cities. It needs only traffic counts, the most widely available data. The method was tested on a model of Brussels with data from real road sensors. In the open SUMO simulator, the tuned model ran about sixteen percent more accurate than models tuned by leading rival methods.

Catalog record

CREATORDavide Andrea Guastella, Alejandro Morales-Hernández, Bruno Cornelis, Gianluca Bontempi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-11585

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Calibration of Vehicular Traffic Simulation Models by Local Optimization* Davide Andrea Guastella, Alejandro Morales-Hernández, Bruno Cornelis, Gianluca Bontempi Abstract Simulation is a valuable tool for traffic management experts to assist them in refining and improving transportation systems and anticipating the impact of possible changes in the infrastructure network before their actual implementation. Calibrating simulation models using traffic count data is challenging because of the complexity of the environment, the lack of data, and the uncertainties in traffic dynamics. This paper introduces a novel stochastic simulation-based traffic calibration technique. The novelty of the proposed method is: ( i ) it performs local traffic calibration, ( ii ) it allows calibrating simulated traffic in large-scale environments, ( iii ) it requires only the traffic count data. The local approach enables decentralizing the calibration task to reach near real-time performance, enabling the fostering of digital twins. Using only traffic count data makes the proposed method generic so that it can be applied in different traffic scenarios at various scales (from neighborhood to region). We assess the proposed technique on a model of Brussels, Belgium, using data from real traffic monitoring devices. The proposed method has been implemented using the open-source traffic simulator SUMO. Experimental results show that the traffic model calibrated using the proposed method is on average 16% more accurate than those obtained by the state-of-the-art methods, using the same dataset. We also make available the output traffic model obtained from real data. (*) Preprint version. This paper has been published on Springer Transportation Journal (eISSN 1572-9435). Please cite DOI: 10.1007/s11116-025-10593-x Introduction City administrations aim at sustainable development, where reducing the human footprint is a priority. Tackling sustainable development in densely urbanized environments poses the challenge of driving disruptive changes in complex environments. A major challenge in the traffic domain is providing traffic management experts with smart methods to analyze and evaluate traffic policies and avoid congestion, a major cause of CO 2 emissions [ 1 ] . On the one hand, the impact of the control strategies on road infrastructures is not observable until they are deployed in the real world [ 2 ] . On the other hand, testing control strategies in real-life settings are expensive, risky, and often unfeasible [ 3 ] . In this context, urban traffic simulation models have become an indispensable asset, providing an in-silico environment where it is possible to design and assess alternative control strategies before the deployment. The most important requirement for a traffic model is its realism. If the dynamics of simulated traffic are close to reality, then traffic management experts can reliably assess the consequences of control strategies in the real world [ 4 ] . This paper addresses the problem of estimating traffic simulation models using traffic count data acquired from traffic monitoring devices. We propose a local optimization method that calibrates the simulated traffic to match the input traffic data as much as possible. First, we partition the city network into a set of non-overlapping regions, where a region delimits a part of the road network. We create simulated vehicles starting from each region. The number of vehicles is progressively calibrated to minimize the difference between the real and simulated traffic. Although our proposal can be implemented with any microscopic traffic simulator software, in this work we use the open-source traffic simulator SUMO. The novelties of the proposed method are: ( i ) it performs local traffic calibration, ( ii ) it allows calibrating simulated traffic in large-scale environments, ( iii ) it requires only traffic count data. The remainder of this paper is organized as follows: first, we motivate the local approach for addressing traffic calibration. Then, we discuss different state-of-the-art methods to address traffic calibration. Following, we formalize the problem addressed in this paper. The following section introduces the proposed technique for calibrating traffic models using traffic count data. We then illustrate the experimental setup and compare the results obtained from the proposed technique and state-of-the-art methods for calibrating traffic. Finally, we summarize the main findings of our work and outline future research directions. Motivation The intuition behind the local calibration is to constrain both in space and time the task of calibrating the traffic to the observed data. This is challenging, as the calibration problem is underdetermined [ 5 ] : the available data can be used to define a virtual representation of traffic that is faithful to input information but not to the real dynamics of traffic [ 6 ] . Our objective is to estimate traffic that matches the input data, without necessarily ensuring the accuracy of vehicle trajectories considering the real traffic demand or addressing uncertainties deriving from traffic counts. By isolating parts of the road network, it is possible to vary traffic parameters locally, analyze the local error, and evaluate the impact of local parameters on nearby regions. A further motivation to perform local traffic calibration is related to computational advantages, as it is possible to distribute the load of the traffic calibration over different computational modules, one for each region and time interval. This is effective when considering traffic count data streams in large-scale environments. Related Work The implementation of realistic traffic simulation models has been addressed from different perspectives and has employed various strategies to enhance model accuracy, scalability, and applicability in diverse urban settings. Sha et al. [ 7 ] propose a Bayesian optimization framework for the calibrating problem of traffic simulation models. Bayesian optimization is an approach to optimize expensive objective functions. Hence, the benefits of the suggested calibration method because traffic simulations are generally expensive in terms of simulation time and resource consumption. The key idea of Bayesian optimization is to choose more promising values for future evaluation based on previous results. This is achieved by constructing a surrogate model which is a probability model of the original objective function and is easier to optimize (i.e. it is computationally efficient). The proposed method calibrates simulation parameters such as the vehicles’ maximum speed, acceleration, deceleration, and minimum gap after the leading vehicle for four types of vehicles, resulting in 16 parameters to be calibrated (one per vehicle type). The objective function to minimize is the sum of the root mean square percentage error (RMSPE) of traffic count and speed. The authors validate their proposal using a model of a section of the New Jersey Turnpike and compare their results to those obtained by the Simultaneous Perturbation Stochastic Approximation (SPSA), a stochastic approximation algorithm well suited for optimizing functions involving a high number of parameters. The minimum value of the objective function achieved using Bayesian optimization and SPSA is 6.87% and 7.41%, respectively. While Bayesian optimization guarantees good results when considering several variables in a large-scale scenario, the SPSA technique showed convergence issues [ 8 ] . In addition to the inherent limitations of Bayesian optimization-based methods (e.g., computational complexity is proportional to the number of observed points), the primary drawback of this study is in the actual problem design. Since the authors optimize several parameters for four different vehicle types, they do not consider the traffic dynamics observed throughout the day. Therefore, the calibrated model uses the same parameter values for each vehicle without considering that these may

Read the complete text