Urban Currents

Pedestrian Volumes from Push-Button Traffic Signal Data in Oregon: Estimating Models and Assessing Model Transferability

The researchers developed a method to estimate pedestrian volumes using push-button traffic signal data, validated against 2,679 hours of video recording 41,843 crosswalk users across 64 signalized intersections in Oregon during Summer 2022. The final model uses a quadratic relationship based on push-button presses filtered to exclude multiple presses within 15 seconds, achieving a 0.86 correlation with actual volumes and an average error of plus or minus 2.4 pedestrians per hour. The study also demonstrated model transferability by showing consistent quadratic fit lines when compared with historical data from Utah.

Why it matters — It establishes that pedestrian push-button actuations can serve as a highly accurate, low-cost proxy for actual pedestrian volumes without requiring dedicated physical counting infrastructure. Furthermore, the consistency between Oregon and Utah datasets suggests these estimation models can be transferred across different states and time periods.

Caveat: The model's transferability and accuracy are demonstrated specifically for signalized intersections with push-button infrastructure, which may not represent pedestrian behavior in unsignalized or midblock environments.

Journal of Transportation Engineering Part A Systems · doi

UST-GNN: A unified spatial–topological graph neural network framework for urban analytics demonstrated through a case study on urban health prediction

The researchers developed UST-GNN, a graph neural network framework that combines neighborhood connectivity, heterogeneous urban features, and positional embeddings. Evaluated on the MedSAT dataset across 4,835 neighborhoods in Greater London using over 150 environmental and socio-demographic variables, the model improved out-of-sample R-squared by 8.4% to 13.2% over baseline models when predicting six health prescription outcomes.

Why it matters — It provides a unified GeoAI analytical pipeline that simultaneously accounts for spatial autocorrelation and network topology, allowing planners to interpret learned node embeddings geographically and integrate them into urban digital twin workflows for scenario testing.

Caveat: While the framework identifies novel predictors of urban health outcomes, these associations are correlational and require further causal investigation.

Computers Environment and Urban Systems · doi

Aligning LLM agents with human learning and adjustment behavior: A dual agent approach

A dual-agent framework was developed to simulate how travelers learn and adapt their route choices over time. The system pairs LLM traveler agents, which feature a memory system and a learnable persona, with an LLM calibration agent that continuously trains these personas using online data streams. Tested on a real-world dataset from a day-to-day route choice experiment, the framework outperformed existing baselines in individual behavioral alignment and aggregate flow prediction.

Why it matters — It establishes a method for simulating human travel behavior that captures the evolution of underlying cognitive learning processes rather than just mimicking static decisions, enabling more robust generalizations for transportation planning and policy analysis.

Caveat: The framework's performance and generalization capabilities were evaluated using data from a controlled day-to-day route choice experiment rather than observed, open-world urban travel behavior.

Transportation Research Part C Emerging Technologies · doi

Practices of urban care: the grounding of feminist care ethics in Montevideo

This study analyzes how care is integrated into urban space by examining four active initiatives in Montevideo, Uruguay: the Plan Nacional de Cuidados, Cuidados en el B, Fincas Recuperadas, and Reactor. Using a qualitative analysis of institutional documents and project materials, the research investigates how these programs reorganize social reproduction through communal practices and everyday urban arrangements.

Why it matters — It demonstrates how the intersection of national policy frameworks and neighborhood-scale interventions can move care beyond simple social assistance, reframing access to housing, services, and public space as collective responsibilities.

Caveat: The study relies entirely on qualitative institutional documents and project materials rather than direct observational or experiential data from the participants of these care initiatives.

City Culture and Society · doi

Are urban form and function related to urban discourse? Insights from remote sensing and social media data for 15 major cities across the globe

This study analyzed the relationship between urban form, commercial function, and social media discourse across 15 major global cities. The researchers combined multilingual Twitter data processed via topic modeling with Local Climate Zones derived from remote sensing and spatial analysis of commercial areas. The findings show that while certain topics align with specific urban forms globally—such as finance topics concentrating in compact, commercial zones and sports in low-rise, non-commercial areas—most discourse patterns vary by region or remain independent of urban form.

Why it matters — It demonstrates that the relationship between physical urban environments and the digital discourse of their inhabitants is not universal, but instead varies systematically across different regional contexts. This provides a framework for planners to understand how local cultural and regional factors mediate how people experience and discuss different urban forms.

Caveat: The study relies on Twitter data to represent urban discourse, which may introduce demographic biases and exclude populations who do not use the platform.

Habitat International · doi

Fiscally regressive subsidies and the spatial diffusion of electric vehicles in French cities

Using a synthetic event study and a synthetic difference-in-differences approach, this study analyzed the impact of France's 2018 electric vehicle subsidy reform across French municipalities. The analysis revealed a significantly lower rate of electric vehicle fleet growth in communes with higher proportions of lower-income, non-taxable households following a uniform ‐1,500 subsidy reduction. A supplementary ‐1,000 incentive for secondhand electric vehicles failed to counteract this decline in lower-income areas.

Why it matters — The study demonstrates that flat-rate subsidy cuts impose a disproportionate financial barrier on lower-income households, proving that uniform policy rollbacks can actively stall the clean energy transition in poorer communities.

Transportation Research Part A Policy and Practice · doi

30 years of tendering regional rail services in Germany – A data-based analysis of long-term effects and the impact of network characteristics on competition

This study analyzes the long-term effects of competitive tendering in Germany's regional rail market using a newly compiled dataset of tendering procedures spanning 1998 to 2028. It uses cluster analysis on service network topologies and panel regression to evaluate how network complexity affects competition and subsidy levels, finding that competitors hold a 45% aggregate market share, peaking at 63.5% in medium-complexity networks while the incumbent dominates the most complex networks.

Why it matters — It demonstrates that while competitive tendering initially reduced public subsidies, costs have risen faster than inflation since 2020 as the market enters a consolidation phase. Furthermore, it challenges the widespread use of simple, single-line tenders by showing that medium-complexity networks actually yield better competitive outcomes and subsidy performance.

Caveat: The findings are based on the specific institutional and regulatory history of the German regional rail market, which may limit direct generalizability to countries with different railway reform structures.

Transportation Research Part A Policy and Practice · doi

STFusioner: Bidirectional spatiotemporal fusion via dual-branch cross-attention transformers for traffic flow prediction

The researchers developed STFusioner, a traffic flow prediction model that decouples spatial and temporal features using parallel multi-head self-attention before dynamically fusing them via a bidirectional cross-attention mechanism. Evaluated on real-world datasets, the model achieves state-of-the-art performance, yielding average relative improvements of 1.34% in Mean Absolute Error, 0.77% in Root Mean Squared Error, and 1.24% in Mean Absolute Percentage Error compared to existing baselines.

Why it matters — The model introduces a decoupling-coupling paradigm that implicitly learns empirical traffic dynamics, such as shockwave propagation, allowing practitioners to more accurately predict evolving congestion patterns across large-scale road networks.

Transportation Research Interdisciplinary Perspectives · doi

Advanced air mobility in the UAE: climate-resilient risk and reliability of UAVs versus traditional couriers

This study developed a techno-economic model and ran Monte Carlo simulations to compare the reliability and cost-effectiveness of UAV-based freight delivery against traditional truck couriers under the climatic conditions of the United Arab Emirates. The model evaluated performance using an Operational Integration Index (OII) and exponential reliability functions, finding that UAVs had a much lower mean reliability of 0.604 compared to 1.0003 for trucks, and a lower integration readiness score (OII of 0.288 versus 0.421). Spearman rank correlations showed that temperature and wind speed were the strongest negative drivers of UAV performance, while traditional couriers remained largely unaffected by weather.

Why it matters — It quantifies the severe reliability and cost-effectiveness penalties that extreme desert climates impose on drone delivery systems, challenging the assumption that aerial logistics are ready to replace ground transportation in hot, windy environments.

Caveat: The findings are based on a simulation model calibrated to the UAE's climate rather than empirical flight data from active commercial drone operations.

Transportation Research Interdisciplinary Perspectives · doi

Intercity inequality in carbon emission reductions from vehicle electrification in China

Using over 245 million vehicle registration records, this study quantifies fuel-cycle carbon transfers from plug-in electric vehicles (PEVs) across 285 Chinese cities. In 2020, economically developed cities outsourced 41.8% of their PEV-related emissions to less developed cities, causing per-PEV emissions in the receiving cities to exceed those of conventional vehicles by 16.9% to 52.0%. Projections show this intercity emission inequality will remain high through 2030 due to widening gaps in vehicle ownership, only declining as the national power grid decarbonizes.

Why it matters — The study reveals a spatial mismatch where the decarbonization benefits of electric vehicles in wealthy cities are achieved by shifting the carbon and economic mitigation burdens to poorer, power-producing municipalities.

Nature Cities · doi

Estimating Urban Traffic Patterns through GPS Trace Data from Car-Hailing Vehicles: A Temporal–Spatial Analysis

The study introduces a three-layer framework comprising data, interactive software, and processing method layers to estimate urban traffic patterns from car-hailing GPS trace data. The system processes raw traces through data cutting, map matching, and trace correction to generate matrices of vehicle flow and average road speeds. Analysis using this framework revealed that urban road traffic follows a bimodal temporal pattern, whereas car-hailing flow displays a multipeak pattern with high holiday fluctuations, though both share similar spatial hotspot distributions.

Why it matters — It establishes a systematic method to correct and map car-hailing GPS traces, revealing distinct temporal differences between ride-hailing flows and general urban traffic that planners must account for when using commercial fleet data as a proxy for overall traffic.

Caveat: The framework's performance and the generalizability of the observed traffic discrepancies were not validated against independent, ground-truth traffic sensor datasets in the abstract.

Journal of Transportation Engineering Part A Systems · doi

An Even Better Way to Zone: Achieving More Affordable, Equitable, and Sustainable Communities

This paper evaluates the systemic impacts of traditional zoning practices on housing affordability, urban sprawl, and social equity, proposing a reformed framework for land-use planning.

Why it matters — It provides planners with an alternative regulatory approach to zoning that directly addresses the structural drivers of the housing crisis and unsustainable urban expansion.

Journal of the American Planning Association · doi

Revisiting the ‘National Public and Private Building Standard Drawings’ of 1940s China and their postwar urban and rural construction implications

This study analyzes the 1946 'National Public and Private Building Standard Drawings' issued by the Republic of China after World War II. It examines how these technical drawings were produced, their classification criteria, and their policy-driven application in postwar urban and rural reconstruction.

Why it matters — The research recovers the historical role of standardized architectural atlases in state-led construction management, offering a framework that informs both the modern preservation of historical buildings and the design of contemporary administrative offices.

Planning Perspectives · doi

Critical planning futures: new directions in planning theory

This paper outlines new theoretical directions for the future of planning theory, framing the current state of the discipline through a critical lens inspired by Beckettian themes of stagnation and waiting.

Why it matters — It provides a fresh conceptual framework for planning theorists to critique contemporary institutional inertia and reimagine the purpose of planning practice.

Caveat: The paper is purely theoretical and does not present empirical data or quantitative measurements.

Planning Perspectives · doi

What factors drive the vitality of rail transit station areas? An investigation of urban features in Tianjin, China

This study analyzes the spatial and temporal dynamics of human activity around rail transit stations in Tianjin, China, using hourly Baidu Heatmap data and a Geographically and Temporally Weighted Regression (GTWR) model. The analysis reveals that centrally located, commercially dense station areas experience the highest vitality, characterized by a bimodal daily pattern. Factors such as public service density, floor area ratio, and bus route integration show significant, spatially varying influences on local activity levels.

Why it matters — It shifts the focus of urban vitality research from traditional public spaces like parks and streets to transit nodes, providing a localized, hourly measurement of how specific built-environment features attract or deter people throughout the day.

Caveat: The findings are based on a single city, and the reliance on Baidu Heatmap data as a population proxy may introduce demographic biases toward smartphone users.

Transportation Planning and Technology · doi

From Causal Discovery to Implementation: An Agentic AI Framework for E-Scooter Mobility Hub Planning Across 29 German Cities

An agentic AI framework was developed to construct a Causal Template Library using public General Bikeshare Feed Specification (GBFS) data across 29 German cities. The system uses a large language model to orchestrate causal discovery algorithms across 57 city-cluster units, identifying how environmental features drive e-scooter hotspot demand across four city types and two cluster types. The framework was deployed to score candidate sites, recommend infrastructure, and guide the construction of two real-world mobility hubs in Heilbronn, Germany.

Why it matters — It shifts e-scooter planning from proprietary, correlational demand models to open-data causal models that adapt to specific urban typologies. This allows planners to transfer siting templates between similar city types without requiring local proprietary trip data.

arXiv · pdf

Crime reduction through public healthcare: Interpretable machine learning for mental health service impacts in Greater London

The study analyzed street-level crime data, mental health referral records, and socioeconomic metrics across Greater London using traditional statistics, machine learning, and explainable AI. It identified a persistent positive association between crime rates and mental health referrals, revealing a U-shaped relationship where preventive effects occur at lower service levels while demand-driven responses dominate at higher referral rates. A cluster analysis categorized London boroughs into four distinct typologies based on their unique combinations of crime, service access, and deprivation.

Why it matters — It demonstrates that the relationship between mental health services and crime is non-linear, challenging the simple assumption that more service access always correlates with lower crime and showing that policy interventions must be tailored to specific borough typologies rather than applied as one-size-fits-all solutions.

Caveat: The study relies on mental health referrals as a proxy for actual service access, which may not fully capture completed treatments or informal care.

arXiv · pdf

Cellular Predictions on the Move: What about Data?

The study evaluates mobile cellular load forecasting on highways by incorporating population dynamics and mobility data alongside historical traffic time series. Using these additional data sources to represent the potential number and movement of traffic generators yields forecasting accuracy improvements of approximately 60%.

Why it matters — It demonstrates that modeling the underlying human mobility processes driving network demand is far more effective for cellular load prediction than merely optimizing machine learning architectures on traditional, domain-isolated network data.

Caveat: The findings and the 60% performance improvement are validated specifically for a highway scenario, which may not generalize to more complex urban environments with different mobility patterns.

arXiv · pdf

When Agents Meet Electric Bus Fleet Operations: Pricing Behavior, Trade-offs, and Policy Implications in an Aggregator Framework

This paper introduces an agentic aggregator framework that couples an optimization-based electric bus scheduling model with supervisory agents to manage real-time disturbances, tariff adaptation, and schedule evaluation. The system enforces physical constraints across routes, chargers, batteries, and vehicle-to-grid (V2G) exchanges, and was evaluated using a realistic depot case study under profit-based and operation-based coordination modes facing service delays, energy deviations, and price shocks.

Why it matters — It demonstrates that while agentic systems can successfully automate complex fleet-grid coordination and maintain feasible schedules under stress, profit-oriented configurations can extract financial value away from public transport operators, highlighting the need for auditable tariff-setting and explicit value-sharing rules in public-fleet deployments.

Caveat: The framework's performance and financial trade-offs were evaluated using a single simulated depot case study rather than a deployed, multi-depot transit network.

arXiv · pdf

Count data modeling and forecasting of malaria incidence using generalized time series regression

The study modeled monthly Plasmodium vivax malaria incidence in Mumbai, India, using surveillance data from 2012 to 2019 alongside meteorological variables. It compared Poisson, negative binomial, and Generalized Linear Autoregressive Moving Average (GLARMA) models, evaluating forecasts through rolling time series cross-validation. The GLARMA negative binomial model outperformed other approaches by successfully accounting for both overdispersion and serial dependence in the count data.

Why it matters — It demonstrates that seasonal patterns are more strongly associated with malaria incidence than individual climatic variables, and provides a more stable, accurate forecasting framework for urban disease surveillance and early warning systems.

Caveat: The findings and model performance are demonstrated using data from a single urban region, which may limit direct generalizability to areas with different climatic or epidemiological profiles.

arXiv · pdf