A Multi-Objective Approach for the Calibration of Microscopic Traffic Flow Simulation Models

Cobos, Carlos; Paz, Alexander; Luna, Julio; Erazo, Cristian; Mendoza, Martha · 2020 · OpenAlex-citations

DOI: 10.1109/access.2020.2999081

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of calibrating microscopic traffic flow simulation models, a task complicated by high-dimensional search spaces, parameter interactions, and the need for low-cost, comprehensive solutions. Existing approaches often rely on mono-objective strategies that combine multiple metrics into a single weighted function, leading to information loss and an inability to simultaneously optimize all objectives. Furthermore, previous studies have lacked systematic statistical comparisons of alternative algorithms. To address these gaps, the authors propose an adapted Multi-Objective Global-Best Harmony Search (MOGBHS) algorithm. This approach explicitly considers multiple objectives, handles both continuous and discrete variables, and facilitates easy tuning and parallelization. The study aims to determine a robust calibration strategy by comparing MOGBHS against seventeen other metaheuristics across varying levels of model complexity. The methodology involves minimizing two objective functions: the normalized root mean square (NRMS) error for traffic volumes and speeds. The authors implemented MOGBHS with memetic capabilities, incorporating local search strategies such as Hill Climbing, Simulated Annealing, and Iterative Local Search to enhance exploitation. The algorithm utilizes parallel threading to maximize exploration without increasing computing time. Performance was evaluated using three CORSIM-based traffic models of increasing complexity: a low-dimensionality hypothetical network (McTrans), a medium-dimensionality network in Reno, Nevada, and a high-dimensionality network on Interstate 75 in Miami, Florida. The proposed MOGBHS was compared against mono-objective algorithms (SPSA, GASA), single-state multi-objective algorithms, and other multi-objective metaheuristics like NSGA-II and SPEA-2. Statistical significance was assessed using nonparametric Friedman and Wilcoxon tests, with calibration success defined by meeting specific error thresholds for at least 85% of links. The results demonstrate that the adapted MOGBHS dominated all alternative algorithms in every test case. It provided the most stable and diverse solutions, achieving superior convergence and minimization of errors compared to competitors. The statistical tests confirmed the dominance of MOGBHS, validating its effectiveness across models of different dimensionalities. The algorithm’s ability to efficiently order non-dominated solutions and generate a comprehensive Pareto Front allowed for better visualization of trade-offs between volume and speed errors. Additionally, the parallel implementation proved effective in managing the computational intensity of running multiple simulation models simultaneously. The significance of this work lies in providing a generalized, efficient, and statistically validated framework for calibrating microscopic traffic simulation models. By demonstrating the superiority of a multi-objective approach over mono-objective and other multi-objective alternatives, the study supports the adoption of MOGBHS for complex transportation planning tasks. The findings suggest that explicit multi-objective optimization yields better calibration outcomes, enabling practitioners to select solutions that balance various performance metrics effectively. This contributes to the broader field by offering a robust tool that addresses previous limitations regarding algorithm selection, hyperparameter tuning, and the lack of comprehensive comparative analysis in traffic simulation calibration.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• traffic density