Estimating an origin‐destination matrix with fuzzy weights

Abstract

The Fuzzy‐weighted approach (FWA) to origin‐destination (O/D) estimation is tested on an example network and two case studies. This is to illustrate the versatility and robustness of FWA in solving a variety of link‐count inconsistency problems. In the numerical example, constructed from the experience at the People's Republic of China (PRC), vehicular‐traffic‐count fuzziness is traceable to bicycle obstruction. In Case Study one, drawn from an intercity highway corridor in Eastern Washington, USA, traffic‐count inconsistency is related to differences in road surface condition over time. In Case two, the most general of the three studies, urban traffic counts are subject to random human errors in sampling without apparent “causes.” The three studies represent different network geometries, covering both the regular grid network with multiple paths between O and D and the network where there is only one path between O and D. This is part of our experimental design to compare probabilistic (or multipath) against all‐or‐nothing (or single path) traffic assignment in O/D estimation algorithms. Different traffic sampling rates were tested, ranging from 100% link sampling to 33%. Different “fuzziness” of the link data was also tested starting with a very consistent set found in the Eastern Washington highway corridor and increasingly less consistent in the other two studies. In spite of these varieties, the FWA yields a more accurate O/D estimate in all studies when compared with regular O/D estimation algorithms. The same can be said about probabilistic assignment, which yields better accuracy when compared with all‐or‐nothing assignments.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 28 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.