A Bayesian Optimization approach for calibrating large-scale activity-based transport models

TL;DR

This work tackles the calibration bottleneck in large-scale activity-based transport models by introducing a Bayesian Optimization framework that uses an improved Random Forest surrogate to guide global parameter search. By modeling the surrogate with an Expected Improvement acquisition and optimizing acquisitions via LBFGS-B, the approach scales to high-dimensional parameter spaces (e.g., 477 behavioral parameters in Tallinn) without requiring exhaustive ABM runs. The case study demonstrates substantial alignment with baseline data, achieving an OD error of 15.92 vehicles per day and a 4% error in total trips, highlighting the method's practicality for disaggregate, behaviorally grounded transport analysis. The contribution advances automated, scalable calibration for ABMs and provides open-source data, enabling broader adoption and application to policy analysis and new-technology scenarios.

Abstract

The use of Agent-Based and Activity-Based modeling in transportation is rising due to the capability of addressing complex applications such as disruptive trends (e.g., remote working and automation) or the design and assessment of disaggregated management strategies. Still, the broad adoption of large-scale disaggregate models is not materializing due to the inherently high complexity and computational needs. Activity-based models focused on behavioral theory, for example, may involve hundreds of parameters that need to be calibrated to match the detailed socio-economical characteristics of the population for any case study. This paper tackles this issue by proposing a novel Bayesian Optimization approach incorporating a surrogate model in the form of an improved Random Forest, designed to automate the calibration process of the behavioral parameters. The proposed method is tested on a case study for the city of Tallinn, Estonia, where the model to be calibrated consists of 477 behavioral parameters, using the SimMobility MT software. Satisfactory performance is achieved in the major indicators defined for the calibration process: the error for the overall number of trips is equal to 4% and the average error in the OD matrix is 15.92 vehicles per day.

Figures (9)

• Figure 1: Conceptual design of the iterative Bayesian optimization method with Random Forest as a surrogate model and Expected Improvement (EI) as an acquisition function.
• Figure 2: Spatial distributions of residents
• Figure 3: Performance measure progression across 500 iterations for 5 different runs
• Figure 4: Comparison between the starting point (1) and the best simulation in run 2 (182) -- benchmarks against the baseline and residual error; the y-axis represents the percentage for worker_coverage and mode distribution, and the number of legs for all the other benchmarks
• Figure 5: Difference between the number of trips for each OD pair, calculated between the simulated ones and the total obtained instead by upscaling the mobility survey to the whole population. The axis labels include every third district.