2026-09-04

7 arXiv categories· 96 journals· 2706 candidates — 6 worth your time· 114 without an open abstract

Headline

A multimodal benchmark dataset and vision-centric classifier for mapping gated communities

Today, together

tag shift
no tag ran above its 30-day average
shares references
no items share references today counted over the 6 papers published above
on the same shoulders
6 of the day's papers cite Accessibility evaluation of land-use and transport strategies: review and research directions (Karst Geurs, Bert van Wee 2003); 6 cite Travel demand and the 3Ds: Density, diversity, and design (Robert Cervero, Kara M. Kockelman 1997); 4 cite Travel and the Built Environment (Reid Ewing, Robert Cervero 2010) counted over the 117 of this day's 120 named papers whose reference lists we have — standing on the same older work is not the same as being about the same thing
first appearance
Finnish Environment Institute; University Transportation Research Center none of them on a paper we published in the 84 days before this issue

Two counts, not one counted twice: every row here but “on the same shoulders” counts the papers published above, and that one counts every paper this day named — those and the ones under Also published today, whose abstracts we could not read.

A global framework was developed to estimate relative urban cycling intensity by weighting crowdsourced Strava Global Heatmap data with population and point of interest counts. Researchers also created GBA-GCs, a multimodal benchmark dataset of over 37,000 residential compounds in China's Greater Bay Area, to map gated communities and analyze their equity implications. Finally, a dynamic charging system was modeled to manage mobile electricity suppliers transferring power to electric vehicles in motion within a platoon.

codedatapreprint

Minwei Zhao, Weiming Zhang, Jiawang Du, et al.
arXiv cs.CV

satellite imagery · China

Urban Boundaries, Social Barriers: A Benchmark and Vision-Centric Framework for Mapping Gated Communities and Equity Implications

The researchers developed GBA-GCs, a multimodal benchmark dataset containing 37,444 residential compounds in China's Greater Bay Area, featuring aligned boundary polygons, high-resolution satellite imagery, Chinese metadata, and expert-verified gated/open labels. They also built the Multimodal Classifier for Gated Community (MCGC), a vision-centric framework based on DINOv3-SAT that fuses imagery, text, and structured data using cross-attention and adaptive gating. Applying this model to the metropolitan scale revealed spatial clustering of gated communities, privatized green spaces, and reduced pedestrian connectivity.

Why it matters — It provides a reproducible, automated method to map gated communities at a metropolitan scale, replacing qualitative or small-scale manual audits with a high-performing multimodal machine learning approach.

Caveat: The classification framework and benchmark are specifically tailored to the architectural and administrative context of Chinese residential compounds (fengbi xiaoqu), which may limit direct transferability to other international contexts.

arXiv · pdf · code

datapublished: Journal of Transport Geography

Robert Klein, Elias Willberg, Silviya Korpilo, et al.
Journal of Transport Geography · was arXiv

University of Helsinki

Urban Transport and Accessibility · Human Mobility and Location-Based Analysis · Urban Green Space and Health · spatial analysis · University of Helsinki · General Electric (Finland)

A global framework to estimate urban spatial cycling patterns based on crowdsourced data

The researchers developed a framework to estimate relative urban cycling intensity by weighting openly accessible Strava Global Heatmap data with population and point of interest (POI) counts. They validated this approach against official cycle count data from 29 cities across Europe and North America, finding strong correlations (Spearman's rho > 0.7 for all, and > 0.8 for most cities), with POI-weighting performing best.

Why it matters — It provides a low-effort, globally consistent method to estimate spatial cycling volumes in cities where official physical counter infrastructure is sparse or non-existent.

Caveat: The framework's estimation accuracy is dependent on local cycling rates, performing better in cities that already have a higher cycling modal share.

Journal of Transport Geography · doi

published: Transportation Research Record Journal of the Transportation Research Board

Zhexi Fu, Joseph Y. J. Chow
Transportation Research Record Journal of the Transportation Research Board · was arXiv

University Transportation Research Center

Electric Vehicles and Infrastructure · Transportation and Mobility Innovations · Vehicle Routing Optimization Methods · University Transportation Research Center

On the Modular Platoon-based Vehicle-to-Vehicle Electric Charging Problem

This study models a dynamic charging system where a fleet of mobile electricity suppliers transfers power to electric vehicles in motion within a platoon. The authors formulated a mixed integer linear programming model to minimize energy consumption and travel time, testing it against genetic algorithms and benchmark scenarios on a modified Sioux Falls network.

Why it matters — The dynamic platooning approach demonstrates that in-motion vehicle-to-vehicle charging can reduce energy consumption by up to 11.07%, travel time by 11.65%, and total costs by 11.26% compared to traditional charging station stops, particularly benefiting long-distance routes in areas with sparse charging infrastructure.

Caveat: The findings and performance metrics are based on numerical experiments simulated on a single, simplified network model rather than a real-world deployment.

Transportation Research Record Journal of the Transportation Research Board · doi

published: International Journal of Geographical Information Systems

Ruixiang Liu, Zhenlong Li, Ali Khosravi Kazazi
International Journal of Geographical Information Systems · was arXiv

GeoInformation (United Kingdom)

Geographic Information Systems Studies · Semantic Web and Ontologies · Advanced Graph Neural Networks · GeoInformation (United Kingdom)

Towards intelligent geospatial data discovery: a knowledge graph-driven multi-agent framework powered by large language models

A knowledge graph-driven multi-agent framework was developed to improve geospatial data discovery across distributed and heterogeneous platforms. The system uses a unified geospatial metadata ontology as a semantic mediation layer to align metadata standards, constructs a metadata knowledge graph of datasets and their relationships, and deploys multiple collaborative large language model agents to parse user intent, retrieve graph data, and synthesize answers.

Why it matters — It replaces traditional keyword-based catalog searches with an intent-aware, semantically aligned retrieval process, demonstrating a measurable increase in ranking quality, recall, and discovery transparency.

International Journal of Geographical Information Systems · doi

preprint

Mayleen Cortez-Rodriguez
arXiv stat.AP

Disaster Management and Resilience · Agricultural risk and resilience · Disaster Response and Management · causal inference

Natural Disasters and the Nonprofit Sector

This study analyzes the causal effect of natural disaster damage on county-level nonprofit outcomes using a panel dataset spanning 1991 to 2021. Utilizing causal inference methods tailored for panel data, the analysis evaluates changes in nonprofit revenue and assets following disaster events.

Why it matters — The findings challenge previous literature by demonstrating that natural disaster damage has no causal effect on nonprofit financial outcomes, overturning prior assumptions of small but positive associations.

Caveat: The analysis is conducted at the county level, which may aggregate or obscure highly localized impacts on individual nonprofit organizations.

arXiv · pdf

preprint

Fang He, Tao-yang Fu, Wang-chien Lee
arXiv cs.LG

Traffic Prediction and Management Techniques · Human Mobility and Location-Based Analysis · Automated Road and Building Extraction · transformer · street network data

TraveL: Transformer-based Multi-view Path Distributional Representation Learning

The TraveL framework uses a Transformer-based model to encode road network paths and departure times into distributional representations that capture varied traveler behaviors and regional segment correlations. Tested on synthetic and real-world datasets, the model improves travel time distribution estimation by 14.7% in Mean Kolmogorov-Smirnov distance, path similarity prediction by 16.7% in Mean Absolute Error (MAE), and destination prediction by 3.97% in MAE.

Why it matters — It shifts path representation learning from static vector embeddings to distributional representations, allowing models to capture the inherent variability of human travel behavior and complex spatial dependencies across different regions.

arXiv · pdf

Still cited

Zhenlong Li, Huan Ning (2023), Autonomous GIS: the next-generation AI-powered GIS 1 of today's items cite it · 11 of 3240 in the archive stand on it

Also published today

These appeared today in journals we track. Their abstracts are not openly available, so we cannot summarise them.