Efficiency of choice set generation methods for bicycle routes

Halldórsdóttir, Katrín; Rieser-Schüssler, Nadine; Axhausen, Kay W.; Nielsen, Otto A.; Prato, Carlo G. · 2014 · OpenAlex-citations

DOI: 10.18757/ejtir.2014.14.4.3040

archive: archived pipeline: cataloged verified

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

Summary

This study evaluates the efficiency of three choice set generation methods for bicycle route choice: the Doubly Stochastic Generation Function (DSGF), Breadth First Search on Link Elimination (BFS-LE), and the Branch & Bound (B&B) algorithm. Motivated by the need to improve sustainable transport modeling, the research addresses the challenge of generating plausible alternative routes for cyclists using revealed preference data. Unlike previous studies that relied primarily on distance and time, this work extends cost functions to include bicycle-specific attributes: road type, presence of segregated cycle lanes, and land use (scenic or forest areas). The analysis utilized a dataset of 778 GPS-traced bicycle trips from 139 individuals in the Greater Copenhagen Area, mapped onto a high-resolution network comprising over 270,000 directional links. The authors tested four multi-attribute cost functions to assess how well each algorithm replicated observed routes and generated heterogeneous alternatives. Performance was evaluated using coverage measures (the percentage of observed routes replicated at various overlap thresholds), a consistency index, computational time, and path size factors to measure route heterogeneity. Results indicated that both BFS-LE and DSGF generated realistic routes, whereas the B&B algorithm performed poorly. The B&B method failed to generate alternatives for a large percentage of observations within the time abort threshold, particularly for trips longer than 4 km, resulting in low coverage rates (approximately 40–51% at a 70% overlap threshold). In contrast, BFS-LE achieved the highest coverage, replicating up to 67.9% of routes at 100% overlap and 84.8% at 70% overlap when using the combined cost function. DSGF also performed well, with coverage rates between 58.6% and 63.5% at 100% overlap. While BFS-LE was computationally efficient, DSGF produced more heterogeneous route sets. The inclusion of bicycle-oriented attributes significantly improved performance for both BFS-LE and DSGF, confirming that factors beyond distance and time are critical for accurate bicycle route choice modeling. The study concludes that BFS-LE is the most efficient method for high-resolution networks due to its balance of computational speed and high coverage, while DSGF is preferable when route heterogeneity is a priority. The findings underscore the importance of incorporating specific bicycle infrastructure and environmental attributes into choice set generation to accurately reflect cyclist behavior. This approach provides a robust foundation for estimating route choice models that can inform urban planning and sustainable transport policies.

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.

• cyclist safety