Freeway truck travel time prediction for freight planning using truck probe GPS data

Wang, Zun; Goodchild, Anne V.; McCormack, Edward · 2016 · DOAJ

DOI: 10.18757/ejtir.2016.16.1.3114

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 critical gap in freight planning regarding the accurate prediction of truck travel times. Existing travel time prediction models, such as the Bureau of Public Roads (BPR) and Akçelik functions, are primarily designed for passenger vehicles and often fail to account for the distinct performance characteristics of heavy trucks. Consequently, current planning tools either treat trucks identically to passenger cars or apply crude adjustment factors, leading to inaccurate forecasts that hinder effective freight project prioritization. Motivated by the increasing availability of truck probe GPS data in North America and Europe, the authors propose a pragmatic, multi-regime speed-density relationship approach to generate more accurate, truck-specific travel time predictions. The methodology utilizes empirical truck GPS data for speed observations and dual-loop detector data for traffic volume to derive segment density. The approach involves four steps: identifying traffic regimes using k-means cluster analysis, fitting speed-density relationships for each regime, estimating travel time, and evaluating accuracy. The authors tested two and three-cluster configurations to segment traffic into free-flow, intermediate, and congested phases. For each cluster, they fitted linear, logarithmic, or exponential models based on adjusted R-squared values. The study was validated through two case studies on high-capacity freeway segments in Washington State: a 3-mile stretch of Interstate-5 in Seattle and a 2-mile segment of Interstate-405 in Bellevue. Data collected between May and July 2012 was split into training and testing sets, with Mean Absolute Percentage Error (MAPE) used to compare the proposed method against the BPR and Akçelik models. The results demonstrate that the proposed speed-density approach significantly outperforms traditional methods. In the Interstate-5 case study, the two-cluster model achieved a MAPE of 6.16%, while the three-cluster model achieved 5.55%. In contrast, the BPR and Akçelik models yielded MAPEs of 11.52% and 11.60%, respectively. Although the three-cluster model offered slightly higher accuracy, the authors noted that the improvement was marginal relative to the increased analytical effort, suggesting the two-cluster approach is sufficient for practical application. The findings indicate that the proposed method generates substantially less deviation between estimated and observed truck travel times than existing standard models. The significance of this research lies in its provision of a transferable framework for long-term freight planning. By leveraging readily available GPS and loop detector data, transportation agencies can generate more reliable truck travel time estimates. This improved accuracy supports better decision-making in freight project prioritization, benefit-cost analysis, and travel demand modeling, ultimately enhancing the efficiency of freight infrastructure investments. The study confirms that truck-specific modeling, rather than passenger-vehicle approximations, is essential for accurate freight performance measurement.

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.

• telematics crash prediction