Urban Currents

Learning-based model predictive control for passenger-oriented train rescheduling with flexible train composition

The study develops a learning-based model predictive control (MPC) framework for real-time, passenger-oriented train rescheduling that incorporates flexible train composition and rolling stock circulation. The model combines pre-trained Long Short-Term Memory (LSTM) networks to predict integer decision variables with nonlinear constrained optimization to solve for continuous variables, using four presolve pruning techniques to reduce computational complexity. The approach was validated through numerical simulations using real-life operational data from the Beijing urban rail transit system.

Why it matters — It establishes a method to dynamically adjust train lengths and schedules in response to fluctuating passenger demand, resolving the computational bottleneck of mixed-integer nonlinear programming in real-time transit operations.

Caveat: The performance and computational efficiency of the model are demonstrated through simulations rather than a live physical deployment.

Transportation Research Part C Emerging Technologies · pdf · doi

Assessing the Association Between Urban Amenities and Home Squatting Residents: Evidence from Location-Based Service Data in Weihai

This study identified young home squatting residents aged 26 to 36 in Weihai, China, using location-based services data to analyze their spatial distribution. Using a Multiscale Geographically Weighted Regression model, the research measured how the density of these residents relates to different types of urban amenities within a 500-meter neighborhood. The analysis revealed that traditional physical amenities, real-virtual hybrid amenities, and virtual support amenities show spatially heterogeneous relationships with home squatting density, varying distinctly between the Old Town and the Economic Development Zone.

Why it matters — It provides empirical evidence on how home-centered lifestyles interact with physical and digital urban infrastructure, showing that even highly home-bound residents rely on distinct spatial patterns of local amenities.

Caveat: The findings are based on a single case study of Weihai, China, and focus exclusively on a narrow age cohort of 26 to 36 years old.

Applied Spatial Analysis and Policy · pdf · doi

What remains? Unveiling traces of institutionalization of socially innovative multi-actor collaboration in urban food governance in Madrid (2015–2023)

This study analyzes the evolution of socially innovative multi-actor collaboration in Madrid's urban food governance from 2015 to 2023, a period marked by significant political shifts. Using a newly developed seven-dimension framework that merges critical political science and planning theories, the research evaluates how actor networks and institutional spaces transform through document analysis, participatory research, and interviews.

Why it matters — It demonstrates how hybrid actors, intermediate organizational structures, and supra-local networks can sustain collaborative governance models and preserve social innovations even during periods of political disruption and discontinuity.

Caveat: The findings are based on a single-city case study of Madrid's food sector, which may limit their direct applicability to other urban policy domains or cities with different political dynamics.

Cities · doi

Built environment pathways to cardiometabolic risk in a 20-minute neighbourhood context: A Bayesian network study

This study constructed a Bayesian network model using longitudinal data from the North-West Adelaide Health Study across three waves over a ten-year period. It mapped built environment indices, developed via fuzzy logic and GIS data within a 1600-meter street-network buffer of residences, to cardiometabolic risk. The model explained 32.1% of the variance in HbA1c, demonstrating that food and public facility access connect to HbA1c indirectly through fruit and vegetable intake, physical activity, and BMI.

Why it matters — It demonstrates that the built environment's influence on cardiometabolic health operates through complex, indirect behavioral and biometric pathways rather than direct relationships, establishing BMI and physical activity as critical bridging nodes for urban health interventions.

Caveat: The study relies on data from a single metropolitan region, which may limit how well these specific behavioral and environmental pathways generalize to other urban contexts.

Cities · pdf · doi

The journey to work as observed accessibility: Job access within employment subcenters in greater Mexico City

This study analyzes the determinants of functional commuting linkages between workers' residences and three major employment subcenters in the Mexico City Metropolitan Zone using Spatial Autoregressive Combined models. It utilizes 2019 secondary employment subcenter data alongside public transport commuting trips from the 2017 Origin-Destination Survey to evaluate observed job access. The findings reveal distinct accessibility regimes: the Central Agglomeration acts as an inclusive hub driven by formal labor integration and transit connectivity, Santa Fe serves as a socially selective enclave for higher-income commuters, and Naucalpan functions as a proximity-based industrial subcenter.

Why it matters — It demonstrates that job accessibility is not a uniform metropolitan-wide condition but is instead governed by subcenter-specific regimes, shifting the analytical focus from jobs that are theoretically reachable to those that workers actually secure.

Caveat: The analysis relies on a subsample of public transport commuting trips, which may not capture the accessibility dynamics of private vehicle commuters.

Cities · doi

Systemic resilience of the urban educational facility system: A spatiotemporal framework coupling chronic stresses and acute shocks

This study develops a three-dimensional framework based on Complex Adaptive Systems theory to evaluate the resilience, adaptability, and robustness of educational facilities. Applying this framework to Kumamoto City, Japan, the research models the facility system's performance under demographic shifts from 2020 to 2070 alongside earthquake and flood scenarios, identifying five distinct resilience patterns through K-means clustering.

Why it matters — It establishes a method to simultaneously evaluate long-term demographic contraction and sudden natural disasters, revealing that central urban schools risk resource obsolescence while peripheral schools face functional overload during evacuation events.

Caveat: The empirical findings and spatial patterns are derived from a single case study of Kumamoto City, which may limit direct generalization to cities with different demographic or geographic profiles.

Sustainable Cities and Society · doi

Comparative assessment of heat mitigation strategies across hotspot local climate zones in a hot arid city

This study presents an integrated framework evaluating urban heat mitigation strategies in Doha, Qatar, using a high-resolution LiDAR-GIS approach to classify local climate zones (LCZs) with 95.7% accuracy. Microclimate simulations using ENVI-met in two hotspot areas (LCZ 2 and LCZ 3) during midsummer show that tree canopies provide the greatest localized cooling, reducing air temperatures by 3–5°C and mean radiant temperatures by over 15°C. Conversely, building energy simulations for July reveal that building envelope interventions yield the largest reductions in cooling demand (15–25%), whereas tree cover saves less than 3%.

Why it matters — It establishes a transferable, cross-scale methodology that directly measures the trade-offs between outdoor thermal comfort and indoor building energy savings, demonstrating that the most effective strategy for cooling outdoor pedestrian spaces is the least effective for reducing indoor air conditioning loads.

Caveat: The microclimate and energy simulations are based on two specific hotspot neighborhoods in a single hot, arid city, which may limit the direct transferability of the specific temperature and energy reduction percentages to other climate zones.

Sustainable Cities and Society · doi

Will high-speed rail compete for long-distance travel in a car-dependent society? Insights from a Canadian high-speed rail survey

This study models potential passenger shift to high-speed rail (HSR) along Canada's Toronto-Qu ebec City corridor using a two-stage sequential weighted multinomial logit model. Based on a bilingual survey of 6,813 respondents in October 2025, the analysis calculates an implied value of travel time of $31.5 per hour and projects that HSR could capture 26% of corridor trips under a medium-adoption scenario, drawing riders primarily from private cars and conventional rail.

Why it matters — It quantifies how travelers in a highly car-dependent North American market weigh cost, time, and service quality, isolating group travel penalties and car-habituation effects that are often overlooked in European or Asian HSR studies.

Caveat: The ridership projections rely on stated preference survey data rather than observed travel behavior, which may introduce hypothetical bias regarding actual HSR adoption.

Transportation Research Part A Policy and Practice · pdf · doi

Citizen-based decision support for sustainable public transport system selection: Evidence from Vietnam

The study developed a multi-method decision framework that integrates the Analytic Hierarchy Process for citizen-derived weighting with four ranking methods (SAW, TOPSIS, VIKOR, and PROMETHEE II) to evaluate public transit alternatives. Tested in Ho Chi Minh City, Vietnam, the framework incorporates Monte Carlo simulations to introduce stochastic weight perturbations, evaluating decision robustness through new score-based and ranking-based stability metrics.

Why it matters — It establishes a quantitative method to measure how sensitive public transit decisions are to the inherent noise and variability of participatory citizen inputs, reducing single-method bias for urban planners.

Caveat: The framework's empirical validation is limited to a single case study in Ho Chi Minh City.

Journal of Urban Mobility · doi

Cities that green themselves: urban environments offset vegetation loss from expansion

Using a new conceptual framework applied to 998 cities globally from 2001 to 2022, this study quantified how much the enhanced growth of remaining urban plants offsets the vegetation lost to physical expansion. The analysis reveals that this positive growth offset compensates for approximately 41% of vegetation loss when measured by greenness, and 31% when measured by productivity. Projections indicate that targeted greening of gray infrastructure, such as rooftops and walls, could almost entirely counteract the vegetation loss caused by urban sprawl.

Why it matters — It quantifies a globally prevalent but previously unmeasured ecological trade-off, proving that urban environmental conditions significantly boost the productivity of remaining vegetation and demonstrating that strategic greening of built surfaces can fully neutralize the loss of natural land.

npj Urban Sustainability · pdf · doi

Beyond inequality to inequity: rethinking spatial and population disparities in green cooling effects across China’s major cities

This study quantified greenspace coverage, cooling efficiency, and cooling capacity at a 1-km resolution across 29 major Chinese cities using remote sensing data and spatial modeling. The analysis revealed that realized cooling capacity diverges from greenspace coverage due to spatial variations in cooling efficiency, with higher-priced neighborhoods receiving greater cooling capacity and areas with larger elderly populations experiencing weaker alignment in half of the studied cities.

Why it matters — It demonstrates that equal greenspace coverage does not guarantee equitable thermal benefits, shifting the planning paradigm from measuring physical green infrastructure distribution to assessing actual service delivery across different demographic groups.

Caveat: The analysis relies on 1-km resolution spatial modeling and remote sensing data, which may not capture microclimate variations or localized human-greenspace interactions.

npj Urban Sustainability · pdf · doi

Identification of Priority Attributes for Improvement of Inland Water Transport Service: Importance–Satisfaction Analysis Using Best–Worst Scaling

This study identifies priority areas for improving inland water transport services in Kolkata and Howrah, India, by applying best-worst scaling to capture commuter perceptions. Using a hierarchical Bayesian multinomial logit model alongside fuzzy c-means clustering and gap analysis, the researchers evaluated how commuters rank the importance of and their satisfaction with various transit attributes. The analysis reveals that qualitative factors, particularly safety and security, emerge as top priorities, while fare levels have a low perceived importance.

Why it matters — It demonstrates that transit improvement strategies must jointly analyze importance and satisfaction rather than relying on importance rankings alone, while also proving that service priorities vary significantly across different socioeconomic and trip-based user segments.

Caveat: The findings and priority attributes are derived specifically from the commuter demographics and operational context of the Kolkata and Howrah twin-city inland water transport system.

Journal of Transportation Engineering Part A Systems · doi

Hybrid Congestion Classification Framework Using Flow-Guided Attention and Empirical Mode Decomposition

The study introduces FLO-EMD, a hybrid traffic congestion classification framework that combines dense optical flow-guided spatial attention with empirical mode decomposition of motion traces. The model refines RGB features toward motion-relevant regions and fuses them with decomposed temporal components to classify light, medium, and heavy congestion. Tested on 1,050 five-second video clips from four surveillance networks, the framework achieved a 97.5% overall test accuracy and a weighted F1-score of 0.9742.

Why it matters — It overcomes the limitations of vision-only models that bias toward static infrastructure and signal-only models that lack spatial context, providing a highly accurate method to jointly capture roadway scene context and nonstationary traffic motion.

Journal of Transportation Engineering Part A Systems · doi

Integrating technology and sustainability: Behavioral drivers for autonomous electric vehicle adoption

This study proposes and validates a behavioral intention model for Autonomous Electric Vehicle (AEV) adoption using survey data from university students in the United States. The model integrates functional, emotional, and value-based predictors, finding that trust acts as the primary mediator and the strongest direct predictor of adoption intention, alongside significant roles for green identity and effort expectancy.

Why it matters — It establishes that trust and environmental identity are more critical than simple functional utility in driving early-stage consumer acceptance of uncommercialized, software-driven mobility technologies.

Caveat: The findings are based on a sample of U.S. university students, which may not represent the broader population of potential early adopters.

International Journal of Sustainable Transportation · doi

Multi-granular metro passenger flow forecasting via MMoE and hierarchical multi-source data fusion

The researchers developed MG-MMoE, a unified forecasting framework that predicts metro passenger flows across multiple granularities, including station-level inflow/outflow and inter-station origin-destination/destination-origin flows. The model integrates a multi-gate Mixture-of-Experts architecture to share knowledge dynamically across tasks, alongside a hierarchical fusion strategy that incorporates POI similarity, temporal features, and station attributes. The framework was validated using two real-world metro datasets, outperforming existing baseline models across several prediction horizons.

Why it matters — It replaces separate, isolated models for different flow types with a single, unified system that captures the complex, asynchronous dependencies between station-level and network-level passenger movements.

Transportmetrica B Transport Dynamics · doi

From Mobile Data to Business Insights: An End-to-End Analytics Framework for Large-Scale Urban Mobility Analysis and Decision Support

The researchers built an end-to-end data platform and modular analytics framework using Google BigQuery and Vertex AI to process anonymized mobile application location data. The system executes ETL pipelines to generate data products for mobility profiling, frequent trajectory mining, area of influence analysis, traffic anomaly detection, and origin-destination pattern analysis. The processed insights are delivered to stakeholders through interactive Power BI dashboards.

Why it matters — It provides a reusable, modular architecture that translates raw, large-scale smartphone location data into standardized analytical building blocks, allowing urban planners and commercial decision-makers to run diverse spatial-temporal analyses without building custom pipelines from scratch.

Caveat: The abstract does not disclose the specific sample size, temporal duration, or geographic location of the mobile dataset used to validate the platform.

arXiv · pdf

Development of a Bio-Inspired Routing Algorithm According to Values of Solidarity and a Freirean Perspective of Engineering

A new variant of the Vehicle Routing Problem, termed the Señoritas Routing Problem, was developed to support a bicycle delivery cooperative of cis women and trans people in São Paulo, Brazil. The algorithm incorporates individual rider constraints on weight, volume, and distance, alongside a solidarity objective solved via a genetic algorithm. Testing three fitness formulations showed that a progressive constrained formulation eliminated all constraint violations and reduced the standard deviation of route lengths among riders from 7.92 km to 0.81 km, with only a moderate increase in total distance.

Why it matters — It demonstrates that operations research and routing algorithms can be successfully reoriented to prioritize workload equity and worker solidarity over pure cost or distance minimization.

Caveat: The algorithm was co-designed for and tested on the specific operational constraints of a single small-scale cooperative, which may limit its immediate generalizability to other logistics models.

arXiv · pdf

The limits of visitation entropy as a summary of mobility patterns

This study systematically evaluates visitation entropy using synthetic and empirical human mobility trajectories, revealing that sequence length and the number of unique locations visited explain 90.7% of its variance. The analysis demonstrates that shorter sequences systematically underestimate entropy due to finite-sample bias, which distorts comparisons between demographic groups.

Why it matters — It exposes a fundamental measurement flaw in a widely used mobility metric, showing that apparent entropy differences between genders, commuters, and urban-rural residents are largely artifacts of sequence length and location counts rather than actual behavioral differences.

Caveat: The study identifies the limitations of the metric but relies on the assumption that controlling for sequence length and incorporating structural network measures can sufficiently correct these biases.

arXiv · pdf

Understanding electricity consumption behaviour through Inverse Reinforcement Learning

This study models household electricity consumption in Italy as an agent-based decision process using Inverse Reinforcement Learning to extract reward functions across different consumer clusters. By analyzing household responses to the energy crisis and heatwaves from summer 2021 to summer 2023, the model tracks how cooling behaviors shifted across different socioeconomic groups, built environments, and daily consumption profiles.

Why it matters — It demonstrates that household responses to climatic and economic shocks are highly heterogeneous and durable, proving that the timing of daily electricity use is a distinct dimension of behavioral adaptation that operates independently of a household's socioeconomic status.

Caveat: The findings are derived from clustered consumption profiles rather than individual household-level tracking, which may smooth over some localized behavioral variations.

arXiv · pdf