Inferring Urban Mobility Interactions from Aggregated Dynamics

The paper demonstrates that real‑time urban mobility patterns can be reconstructed from aggregated area‑level counts, without tracking individuals. Using a physics‑informed, uncertainty‑aware framework, the authors infer future origin‑destination flows across twelve datasets from the U.S. and China, achieving accuracy comparable to models that use historical OD matrices. Probabilistic modeling corrects underestimation of sparse corridors, and architectures that preserve spatial heterogeneity before reconstructing pairwise interactions yield more faithful interaction estimates.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.