Simulation models are critical for management and control of highway. A critical issue of applying simulation models is the calibration of parameters, referring to determine simulation parameters such that the simulation outputs best fits real-world observations. Existing methods typically focus on isolated scenarios, e.g., free-flow sections, limiting the applicability of the parameters to diverse highway environments such as toll plaza and accident sections. Additionally, such scenario-specific approaches lead to inconsistent representations of driving behaviors across a full highway corridor. To tackle this issue, this paper presents an integrated framework for multi-scenario calibration of driving behavior parameters, considering three representative highway scenarios: toll plaza, free-flow section, and accident section. The method combines global sensitivity analysis with simulation-based optimization. Specifically, variance-based Sobol indices are first used to identify influential parameters across multiple highway scenarios, reducing calibration complexity. Then, a joint calibration model incorporates cross-scenario shared parameters and scenario-specific adjustments, optimized through a Differential Evolution algorithm. Experimental results demonstrate that the proposed framework improves calibration accuracy and robustness compared to single-scenario approaches. Sensitivity analysis reduces computational cost while maintaining performance, with jointly calibrated parameters achieving consistent replication of traffic metrics across all scenarios. This study provides a systematic approach for developing reliable simulation models applicable to complex highway environments.
Shi et al. (Fri,) studied this question.