Urban Currents

Network-wide fine-grained passenger loading prediction via dynamic relational spatiotemporal deep graph neural networks

The researchers developed the Dynamic Relational Graph Convolutional Recurrent Neural Network (DRGCRNN) to predict fine-grained, edge-level passenger loads across transit networks. The model integrates three relational graphs representing physical layout, operational information, and dynamic loading, alongside a Positive-Unlabeled learning module to handle missing sensor data. It was validated using real-time vehicle location and passenger count data from the Helsinki commuter rail network, demonstrating high accuracy during peak travel periods.

Why it matters — It shifts transit forecasting from coarse station-level predictions to precise, vehicle-specific load monitoring across multi-line networks, even when faced with incomplete data from sensor failures or canceled services.

Caveat: The model's performance was validated exclusively on the Helsinki commuter rail network, which may not fully represent the complexity or operational patterns of larger, more diverse transit systems.

Transportation Research Part C Emerging Technologies · doi

A bottleneck model with shared autonomous vehicles: Scale economies and price regulations

This study develops a bottleneck model to analyze commuter departure times and mode choices between shared autonomous vehicles (SAVs) and normal vehicles under three pricing scenarios: marginal cost, average cost, and unregulated monopoly pricing. The model demonstrates that average cost pricing can lead to multiple equilibria, where early implementation suppresses SAV adoption and increases costs, while later implementation achieves high adoption and cost reductions without deficits. It also identifies a Downs-Thomson paradox where expanding road capacity under average cost pricing increases overall commuting costs.

Why it matters — It reveals that promoting SAV adoption does not always minimize social costs, showing that optimal policy requires combining pricing regulations with congestion tolls to prevent the excessive adoption and overuse of shared fleets.

Caveat: The findings are based on a theoretical bottleneck model of commuter behavior and have not been validated with empirical traffic or ride-sharing data.

Transportation Research Part C Emerging Technologies · doi

Exploring healthcare in smart cities: projects, deployments and challenges

This review maps the integration of commercial wearable devices into smart city infrastructures, analyzing their clinical requirements alongside large-scale architectural challenges. It compares leading commercial wearables and their monitored vital signs, evaluates three active smart health projects (Brescia Smart Living, City4Age, and Minha História Digital), and provides a correlation table mapping multi-vital sign changes to dozens of pathologies.

Why it matters — It identifies a critical technological gap in continuous cuffless blood pressure monitoring and demonstrates how multi-sensor data fusion can enable early detection of acute conditions like sepsis, shifting urban healthcare from reactive to proactive models.

Urban Informatics · pdf · doi

Urban resilience evaluation and obstacle factor analysis of Beijing-Tianjin-Hebei urban agglomeration

This study evaluated the urban resilience of the Beijing-Tianjin-Hebei urban agglomeration from 2010 to 2022 using the entropy weight TOPSIS method and the Obstacle Degree Model. The analysis tracked resilience trajectories across multiple subsystems, finding a general upward trend in all cities alongside persistent deficits in ecosystem resilience. The loan-to-deposit ratio, emergency knowledge promotion, and foreign investment utilization emerged as the primary obstacles hindering resilience improvements.

Why it matters — It identifies the specific financial, educational, and environmental bottlenecks that limit resilience across a major Chinese urban agglomeration, shifting the focus from general development metrics to targeted structural vulnerabilities.

Caveat: The study relies on macro-level statistical indicators, which may not capture localized or community-level resilience dynamics within the individual cities.

Discover Cities · pdf · doi

Potential impacts of personal autonomous vehicles on travel behaviors: A land use and transportation interaction approach

This study simulates the impact of personal autonomous vehicles (AVs) on travel behavior in Swindon, UK, using the TRANUS land use and transportation interaction framework. The model tests three hypothetical scenarios where AVs are shared within households and friend networks, alongside sensitivity analyses on AV occupancy, speed, and waiting times. The simulation reveals that AVs slightly increase overall travel demand and trip generation while reducing vehicle ownership, travel times, and overall trip segments through shared rides.

Why it matters — It demonstrates how household-level AV sharing can mitigate some of the negative externalities of autonomous transport, showing a reduction in vehicle travel distance and ownership despite a shift away from public and active transit.

Caveat: The findings are based on a simulated environment modeled on a single British city, which may limit their generalizability to other urban contexts.

Sustainable Cities and Society · pdf · doi

Adjusting to extreme demand volatility: A comparative analysis of air navigation service provider financial resilience in the southern Asia-Pacific

This study analyzes the financial resilience of five corporatized, government-owned air navigation service providers in the southern Asia-Pacific (Australia, New Zealand, Indonesia, Papua New Guinea, and Fiji) during the COVID-19 pandemic using eleven years of audited annual reports (2015 to 2024/25) and nine semi-structured interviews. Using the Cost Flexibility Ratio, it measures how these providers managed a 38% to 70% drop in user-charge revenue against highly rigid cost structures. The analysis reveals that while three providers recovered or exceeded pre-pandemic revenues by 2024, Airservices Australia remained in deficit despite near-full revenue recovery due to differences in labor contracts, investment cycles, and pricing freezes.

Why it matters — It quantifies the structural 'cost stickiness' of air traffic control operators, demonstrating that labor contracts and capital cycles prevent rapid cost-cutting during demand collapses and showing how different government support mechanisms affect long-term financial recovery.

Caveat: The findings are based on a small sample of five corporatized, fully state-owned providers in a specific geographic region, which may not represent the financial dynamics of fully privatized or directly government-run air navigation models elsewhere.

Transportation Research Interdisciplinary Perspectives · pdf · doi

Examination of image conditions using multiway analysis of variance for predicting high-crash-risk intersections with image recognition AI

The study trained a convolutional neural network on on-site photographs from Fukuoka City, Japan, to predict high-crash-risk intersections, achieving a mean crash risk score of 0.776. It then used a multiway analysis of variance (ANOVA) to statistically evaluate how three image-capture factors—resolution, sky editing, and camera distance—affected the model's predictions. The analysis revealed that unedited skies and a 10-meter camera distance significantly improved prediction reliability, with the optimal configuration occurring at a 1,280 x 960-pixel resolution.

Why it matters — It establishes a statistically validated framework showing that physical image-capture conditions directly alter AI-based road safety predictions, moving beyond model architecture to offer concrete guidelines for standardizing image collection in transport planning.

Caveat: The empirical findings and optimal image parameters are derived from a single model trained on a dataset from a single city.

Transportation Research Interdisciplinary Perspectives · pdf · doi

More with trees: Long-term learning and adaptation of urban forest governance in the Mersey Forest

This study analyzes the 30-year evolution of the Mersey Forest, an urban forestry partnership in England, using a mixed-methods approach to identify five distinct developmental phases. It tracks how the initiative adapted to fluctuating tree-planting rates by diversifying its funding, governance structures, and the recognized benefits of urban trees.

Why it matters — It demonstrates how long-term urban forestry programs can survive shifting political and financial climates by balancing structural stability with experimental, boundary-spanning partnerships. This shifts the focus of urban forest evaluation from simple planting targets to the relational and learning processes that sustain governance over decades.

Caveat: The findings are based on a single, highly specific regional partnership in England, which may limit direct applicability to urban forestry contexts with different local government structures.

Urban forestry & urban greening · doi

Uneven skies: How digital infrastructure and local context shape civilian drone adoption

This study analyzed U.S. county-level drone registration data from 2016 to 2019 using spatial-temporal hotspot analysis and Multiscale Geographically Weighted Regression (MGWR) to identify predictors of civilian drone adoption. While global models identify broadband access as the primary correlate, the MGWR model reveals that the positive association of digital infrastructure with drone adoption is stronger in rural and developing regions than in established technology hubs.

Why it matters — It demonstrates that the adoption of physical-digital technologies like drones does not spread uniformly, showing that local digital infrastructure and housing density act as spatially distinct drivers that can exacerbate existing urban-rural and socioeconomic divides.

Caveat: The analysis relies on county-level aggregations of registration data, which may mask finer-grained neighborhood-level disparities in drone adoption.

Journal of Urban Mobility · pdf · doi

MulDet3D: Multiobjective Optimization-Based Unsupervised Object Detection for Multiple Roadside LiDARs

The researchers developed MulDet3D, an unsupervised two-stage clustering framework for 3D object detection using multiple roadside LiDAR sensors. The system integrates reliability-weighted background modeling, multi-LiDAR registration, and adaptive density-based clustering with a multiobjective particle swarm optimization algorithm to automatically tune parameters. Evaluated on two real-world dual-LiDAR datasets, the method achieved average precision scores of 76.71% for pedestrians, 70.23% for small vehicles, and 72.62% for large vehicles.

Why it matters — It enables high-accuracy traffic monitoring and object detection across overlapping roadside sensors without requiring the labor-intensive manual labeling of training data.

Caveat: The framework was only tested on dual-LiDAR setups under normal weather conditions, and its performance in larger sensor arrays, adverse weather, or complex intersections remains unverified.

Journal of Transportation Engineering Part A Systems · doi

Cultivating modernity: public parks, visuality, and knowledge transfer in late nineteenth-century Korea

This study analyzes the historical diffusion of public parks to Korea during the Open Port Period (1876-1910) by examining the 1895 travelogue 'Observation on a Journey to the West' by reformist Yu Gil-jun. Through textual analysis of this early account of Western industrialization, the research identifies three distinct visual frameworks—reformist, empirical, and subjective—used to translate and adapt Western urban design concepts for a Korean context.

Why it matters — It demonstrates that the adoption of Western park design was not a passive imitation, but an active cultural translation where the park was re-imagined from a pastoral retreat into an active vantage point for experiencing and aestheticizing the modern city.

Caveat: The analysis relies primarily on a single historical text by one prominent reformist to understand the broader cross-cultural transfer of park ideology in late nineteenth-century Korea.

Planning Perspectives · doi

Why now, why here: a conjunctural theory of socialist advance in big city America

This paper develops a comparative framework to explain why socialist electoral movements are succeeding in major US knowledge-economy cities but not in peripheral areas. By comparing Los Angeles, New York, and Minneapolis, it identifies four key conditions for socialist breakthroughs: two organic conditions (an educationally rich but housing-poor middle class, and a concentrated counter-hegemonic civil society) and two conjunctural conditions (a majoritarian legitimation crisis and a minoritarian veto architecture).

Why it matters — It provides a theoretical model that explains the specific geography and timing of modern American socialist electoral victories, moving beyond universal Marxist templates and highly localized city descriptions to show how national economic shifts manifest in specific urban political landscapes.

Caveat: The framework is developed using a comparative analysis of only three major US metropolitan areas, which may limit its direct application to smaller cities or different national political contexts.

Urban Geography · doi

Ferguson effects: libidinal geographies of “police disengagement” and late fascist urbanism

This study analyzes five major qualitative research papers on the 'Ferguson effect' and 'police disengagement' through the lens of psychoanalytic theory. It examines how demands for police accountability and constraints on state violence by Black-led movements are psychologically processed by police officers as a crisis of complete powerlessness.

Why it matters — It shifts the critique of the 'Ferguson effect' from empirical debunking to a psychological and geographic analysis, demonstrating how perceived threats to police authority foster alignment between law enforcement and late-fascist urban ideologies.

Caveat: The analysis is conceptual and qualitative, relying on the interpretation of existing qualitative studies rather than new empirical or quantitative measurements of police behavior.

Urban Geography · doi

Towards welfare-oriented recommendations in activity-travel behavior

The study introduces a welfare-oriented framework for activity-travel recommender systems that evaluates suggestions based on net utility, defined as experienced benefit minus travel costs. It formalizes two decision criteria: Positive Utility Probability, which filters recommendations by the likelihood of non-negative net utility, and Regret Minimization, which limits expected regret compared to a user's self-selected alternative. These criteria were evaluated using an agent-based simulation of heterogeneous synthetic travelers navigating a spatial environment with realistic travel costs, congestion, and feedback loops.

Why it matters — It provides a theoretical and operational foundation for recommender systems that prioritize user welfare over simple popularity or collaborative filtering, preventing systems from suggesting activities where unrecoupable travel costs outweigh the benefits.

Caveat: The framework's performance and behavioral feedback loops are evaluated entirely within a synthetic agent-based simulation rather than with real-world user deployment data.

arXiv · pdf · doi

Can Urban Blight Be Accessed with Vision-language Models: A Case Study in Detroit

This study developed a framework using open-source large vision-language models to assess residential blight by evaluating housing attributes like roof integrity, wall damage, and boarded openings from multiple street-view angles. The models generate binary and probabilistic disrepair estimates, which were validated against professional human annotations in Detroit using an XGBoost ensemble stacking approach and a weighted scoring system.

Why it matters — It establishes a low-cost, scalable method for tracking housing stock conditions over time, providing municipalities with a automated alternative to labor-intensive and infrequent manual blight surveys.

Caveat: The framework's performance and validation are demonstrated using a case study restricted to a single city.

arXiv · pdf

Information Technology Curriculum: General or Specialized? An Australia's Census Study

This study analyzes how Australian universities structure their information technology degrees across three categories: general, specialized, and major-based. Using effect size analysis on a census of Australian IT programs, the research measures correlations between these degree structures and institutional factors such as university reputation, degree level, research components, infrastructure, and industry engagement.

Why it matters — It establishes a comprehensive baseline of IT curriculum design across an entire national higher education system, allowing universities to align their degree structures with specific institutional strengths and industry employment demands.

arXiv · pdf

Vulnerability Modeling for the Adaptation of Physical Systems to Climate Change Extremes

This paper proposes conceptual frameworks for modeling the vulnerability of physical infrastructure to climate change extremes, linking hazard intensity directly to physical damage. The methodology is demonstrated through a theoretical case study of power transmission towers exposed to wind gust hazards across France, using vulnerability curves to project potential damage.

Why it matters — It provides a quantitative framework to guide adaptation planning, allowing operators to determine how to modify physical systems to keep damage within acceptable thresholds as climate hazards intensify.

Caveat: The demonstration relies on a theoretical case study with highly simplified, strong assumptions rather than empirical validation on actual infrastructure failures.

arXiv · pdf