Estimating Path Choice Models through Floating Car Data

Comi, Antonio; Polimeni, Antonio · 2022 · Crossref

DOI: 10.3390/forecast4020029

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 path choice models for transportation scenario assessment, a process traditionally hindered by the high cost and limited reliability of survey-based data collection. The authors propose leveraging Floating Car Data (FCD), specifically GPS tracking, to derive reliable datasets of actual user paths. The study aims to demonstrate that FCD can effectively support the calibration of route choice models for both passenger and freight transport, offering a scalable and robust alternative to traditional methods while maintaining applicability in large-scale assignment procedures. The methodology employs a general procedure for scenario assessment where FCD is used to define the choice set for each origin-destination pair. Rather than generating hypothetical routes, the authors assume the choice set consists of all observed routes within a significant time period. A map-matching algorithm associates GPS points with road links to identify specific paths and calculate travel times. The study utilizes a Multinomial Logit (MNL) model within the Random Utility Theory framework. This model was selected for its analytical tractability and suitability for assignment procedures, particularly in large-scale networks where route overlap is minimal. The systematic utility function incorporates travel time, toll costs, and the percentage of the route on highways. Model calibration is performed using maximum likelihood estimation, followed by validation to ensure the model accurately reproduces observed choices. The approach is applied to two case studies in Italy. Case Study 1 focuses on heavy goods vehicles (laden weight >18 tons) in the Veneto region, analyzing 59,226 trips exceeding 50 km from a database of over 5,300 vehicles. Case Study 2 examines private cars in the Lazio region, utilizing data from 49,105 vehicles and 231,958 trips. The calibrated parameters for both models were statistically significant with correct signs. For freight vehicles, the value of time (VOT) was estimated at approximately 24 €/h, with a rho-squared ($\rho^2$) of 0.26 and an 84% accuracy rate in reproducing observed choices. For cars, the VOT was approximately 10 €/h, with a $\rho^2$ of 0.39 and a 92% accuracy rate. These results align with values reported in existing literature. The significance of this work lies in demonstrating that FCD provides a feasible, reliable, and efficient method for calibrating route choice models without the resource-intensive requirements of traditional surveys. By using observed paths to define choice sets, the method simplifies data processing and reduces computational complexity. The study concludes that FCD is particularly effective for large-scale networks where route overlap is negligible, allowing for the successful application of simpler models like MNL. This approach facilitates the broader implementation of path choice simulations in transportation planning and policy assessment, ensuring that models are both empirically grounded and practically applicable.

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.

• exposure measurement