Urban Currents

Spatially uneven service pressure under population decline: evidence from Hungarian primary education

This study analyzes the relationship between demographic decline and teaching intensity in Hungary's primary education system by combining site-level administrative data with census-based demographic indicators. It decomposes the teaching-hours-per-student metric to reveal that elevated teaching intensity in shrinking municipalities is driven by structural constraints, specifically smaller class sizes, rather than an intentional expansion of instructional hours. Additionally, spatial analysis shows a territorial overlap between high teaching intensity and socially disadvantaged student populations, while students with learning and behavioral difficulties exhibit a U-shaped relationship with local population change.

Why it matters — It demonstrates that high per-student teaching intensity in declining regions is a structural consequence of maintaining local school access rather than administrative inefficiency, showing how demographic shrinkage creates uneven service pressures within centralized public systems.

GeoJournal · pdf · doi

Beyond infrastructure: exploring drivers of subjective cycling safety using directed acyclic graphs (DAG)

This study analyzed telephone survey data from 1,530 respondents in Germany using directed acyclic graphs and linear regression to model the causal drivers of subjective cycling safety. The analysis evaluated how gender, age, and social environments influence safety perceptions, finding that women (particularly those with a migration background) feel less safe due to higher anxiety levels, while a pro-cycling social environment increases perceived safety. Self-reported cycling competence was identified as a key mediator across all tested causal models.

Why it matters — It demonstrates that subjective cycling safety is shaped by social networks and self-efficacy rather than physical infrastructure alone, providing planners with specific social levers to encourage cycling adoption.

Caveat: The study relies on self-reported survey data and subjective measures of competence and safety rather than observed behavior or objective safety metrics.

Transportation · pdf · doi

Hybrid human-AI governance framework for accountable decision-making in urban infrastructure management

This study introduces a hybrid human-AI governance framework for municipal infrastructure planning, developed from a thematic analysis of 78 coded responses from 20 infrastructure professionals and six commercial large language models across three decision scenarios. The resulting five-stage workflow uses a RACI matrix to assign data synthesis and option generation to AI, while reserving contextual judgment, ethical oversight, and final authority for human professionals.

Why it matters — It provides a structured, empirically grounded pathway for municipal agencies to integrate large language models into their workflows while mitigating risks like hallucinated outputs and regulatory non-compliance. The framework translates high-level international AI standards into an actionable 18-month implementation roadmap designed to preserve public trust and accountability.

Caveat: The framework is built on a relatively small sample of 20 professionals and six commercial models, which may not capture the full diversity of municipal operational constraints or the performance of open-source alternatives.

Discover Cities · pdf · doi

Analysis of travel mode choice behaviour for short-distance and regular trips in Indian metropolitan cities

This study analyzes travel mode choice behavior for short-distance and regular trips in Delhi using a Multinomial Probit modeling approach. Based on survey data from 600 respondents, the model evaluates how socio-economic factors, travel time, cost, and distance influence decisions across options like private vehicles, metro, buses, auto-rickshaws, and walking.

Why it matters — It quantifies the threshold where travel mode preferences shift in an Indian metropolitan context, demonstrating that income and vehicle ownership are the primary drivers of private vehicle reliance, while metro transit is preferred only as trip distances increase.

Caveat: Although the findings are presented as generalizable to other Indian metropolitan areas, the primary data is restricted to a single survey of 600 respondents in Delhi.

Discover Cities · pdf · doi

Changing residential preferences in the digital age: Evidence on teleworking and online shopping from the Netherlands

This study analyzed how teleworking and online shopping influence housing choices using a stated choice experiment with 1,064 Dutch participants. The researchers applied tenure-specific discrete choice models, specifically a covariate-dependent Generalized Mixed Logit model for homeowners and a mixed multinomial logit model for renters, to evaluate trade-offs regarding commute distances, retail access, and anticipated life events.

Why it matters — It demonstrates that digital behaviors alter physical location trade-offs, showing that teleworking specifically mitigates the perceived penalty of a 20 km commute for homeowners, whereas high levels of online non-grocery shopping paradoxically remain tied to a preference for immediate retail proximity.

Caveat: The findings rely on a stated choice experiment rather than observed residential moves, and the link between online grocery shopping and supermarket proximity preferences may be confounded by income and existing locational differences.

Cities · doi

Urban climate actions: connecting city characteristics to planning responses

This study analyzed the planned climate interventions within the Climate Action Plans of 20 diverse global cities, mapping them against each city's unique physical, climatic, environmental, and socio-anthropological characteristics. The researchers used natural language processing to classify the planned actions and applied multivariate statistical clustering to group the cities by their urban profiles. The analysis reveals that cities facing immediate, severe climate threats prioritize community-level adaptation and hazard awareness, whereas economically wealthier cities focus on integrating climate action into broader economic restructuring and policy frameworks.

Why it matters — It empirically demonstrates how distinct urban profiles dictate the prioritization of specific climate interventions, moving beyond general vulnerability assessments to show how local context translates into actual planning choices.

Caveat: The findings are based on a relatively small sample of 20 global cities, which may limit the generalizability of the identified city clusters and action patterns.

Cities · pdf · doi

Good(s) planning in cities — A review of urban planning scholarship on large freight generators

This systematic review analyzed ten leading urban planning journals to evaluate how planning scholarship addresses large freight generators, which are facilities that produce and attract high volumes of freight trips. The analysis reveals that the spatial and temporal dimensions of moving, storing, and handling goods are rarely treated as explicit objects of urban governance in planning literature.

Why it matters — It identifies how ignoring freight in urban planning systematically displaces logistics facilities to peripheral areas and causes operations to spill onto streets and curbs, providing a structured research agenda to integrate goods and services directly into new residential neighborhood designs.

Cities · pdf · doi

Rural microtransit service to replace fixed route service: A California case study

This study evaluates the Yolobus BeeLine microtransit service in rural Yolo County, California, using rider surveys and ride-along interviews. It identifies that frequent prior use of buses or taxis, physical mobility limitations, and high service satisfaction positively influence microtransit usage, while older age, identifying as White, and frequent solo driving correlate with lower use. The introduction of the service reduced riders' reliance on carpooling, ride-hailing, and driving alone, though financial analysis across different service zones shows that low population and destination density sharply escalate operating costs.

Why it matters — It provides empirical evidence on how rural microtransit alters travel behavior and displaces greenhouse-gas-emitting automobile trips, while quantifying the steep financial trade-offs planners must accept to achieve these accessibility gains in low-density areas.

Caveat: The findings are based on a single case study of one microtransit service in Yolo County, California, which may limit generalizability to rural areas with different demographic or geographic profiles.

Case Studies on Transport Policy · doi

A simulation-based policy evaluation of healthcare access using heterogeneous emergency medical services vehicles and incidents

An agent-based simulation framework was developed to evaluate how different dispatch policies and proportions of specialised ambulances affect emergency medical services (EMS) survival rates and response times. The simulation tested varying levels of dispatch flexibility for specialised vehicles across different incident priorities, revealing that restrictive dispatch rules increase waiting times for non-prioritised cases when specialised vehicle supply exceeds demand. Spatial analysis within the simulation showed that response and survival disparities are most severe in peripheral geographic areas.

Why it matters — It demonstrates that the clinical benefits of specialized ambulances depend heavily on dispatch flexibility, showing that overly restrictive policies backfire by reducing overall system efficiency and worsening outcomes for general patients.

Caveat: The findings are based on a simulated agent-based model rather than empirical dispatch data from an active, real-world deployment.

Transport Policy · doi

The effects of autonomous truck adoption on industry: insights from a grounded theory analysis of interviews with industry professionals

This study presents six propositions on how autonomous semi-trucks will impact the United States trucking sector and its supporting businesses, using a grounded theory analysis of interviews with industry professionals. The findings map expected changes across trucking safety, labor dynamics, roadway infrastructure, market competitiveness, operating costs, and business models.

Why it matters — It provides a structured, qualitative framework of industry expectations directly from sector professionals, mapping out how autonomous vehicle adoption will reshape the broader logistics ecosystem beyond just the technology itself.

Caveat: The findings are based on qualitative interview propositions rather than empirical operational data or post-adoption measurements.

Transportation Research Interdisciplinary Perspectives · pdf · doi

Improving combined transport coordination through cooperative allocation mechanisms in a physical internet platform: Evidence from a German corridor

This study evaluates cooperative allocation mechanisms for a Physical Internet-inspired platform to coordinate road-rail freight transport, using empirical corridor data from Germany. It identifies overlapping services and underutilized rail capacity, demonstrating through stylized profit calculations that consolidating parallel services generates a collective surplus. The paper compares how different allocation rules distribute these gains and outlines a conceptual platform design that protects data confidentiality while scheduling shipments.

Why it matters — It demonstrates a practical governance framework for horizontal logistics cooperation, showing how digital platforms can distribute financial gains transparently to incentivize operators to shift freight from road to rail.

Caveat: The findings are based on stylized profit calculations and a conceptual platform design rather than a real-world deployment or a formal game-theoretic equilibrium model.

Transportation Research Interdisciplinary Perspectives · doi

CommuniWave:A Machine Learning Model for Quantifying the Degree of Temporary Informal Behavior in Urban Communities

The researchers developed CommuniWave, a machine learning model that quantifies the Degree of Informal Behavior (DIB) in urban communities using street video footage. The system integrates a Behavior Capture Net built on mmaction2, a custom YOLOv10 object detection model, and a random forest-based Behavior Eval Model to generate dynamic DIB fluctuation charts.

Why it matters — It provides urban planners and managers with a quantitative, automated tool to monitor temporary informal resident behaviors, offering an empirical basis to align top-down community planning with actual spatial usage.

Caveat: The model's performance metrics, training dataset size, and specific deployment locations are not quantified in the abstract.

arXiv · pdf

Self-similarity of mobility networks

The study introduces a Neighbor-Limited Box Covering method to renormalize undirected weighted mobility networks by iteratively merging high-strength nodes with their strongest neighbors. This method was applied to analyze inter-city human mobility and freight trip networks in China, evaluating how topological structures, weighted features, and dynamic processes behave across scales.

Why it matters — It demonstrates that both human and freight mobility networks exhibit self-similarity across scales, and reveals that the resulting renormalized nodes naturally align with political and socio-economic boundaries without using any explicit spatial data during the clustering process.

arXiv · pdf

MobiDiff: Semantic-Aware Multi-Channel Discrete Diffusion for Human Mobility Data Generation

The researchers developed MobiDiff, an end-to-end discrete diffusion framework that generates synthetic human mobility data by directly denoising multi-channel semantic skeletons across spatial, activity, and temporal channels. Evaluated on three large-scale datasets from Atlanta, Boston, and Seattle, the model preserves trajectory length and temporal interval distributions while running 5.3 times faster during inference than the state-of-the-art GeoGen model.

Why it matters — It provides a highly efficient and interpretable method for generating privacy-preserving synthetic mobility data, bypassing the computationally expensive interpolation and latent trace construction steps required by continuous diffusion models.

arXiv · pdf

Does online sustainability communication shape public discourse? Insights from six years of tenant-housing provider interactions

The study analyzed 792 Facebook posts and 3,197 tenant comments from 92 Dutch public housing providers between 2018 and 2023 using a machine-learning pipeline and multinomial logistic regression. It classified tenant responses across three dimensions—communicative intent, sentiment, and semantic relatedness—to identify six distinct discourse types. The analysis revealed that tenant comments are semantically aligned with the original posts, and that organizational traits, such as larger size and higher rent levels, are the primary drivers of substantive and evaluative engagement rather than specific post-design features.

Why it matters — It demonstrates that public social media interactions are structured and responsive to organizational context rather than being merely random or superficial, providing a scalable framework to measure the actual quality and alignment of civic discourse beyond simple engagement metrics like likes.

Caveat: The findings are based on Facebook interactions within the specific context of Dutch public housing providers, which may not fully represent discourse dynamics on other social platforms or in different public service sectors.

arXiv · pdf

Temporal Boundaries of Newell's Car-Following Model: Insights from Lane-Free Traffic

This study evaluates Newell's car-following model using high-resolution UAV trajectory data from a lane-free highway corridor in Chennai, India. It compares aggregate and pair-specific linear regressions against a trajectory-shifting optimization method, and introduces a boundary-corrected variant to isolate transitional regimes at interaction endpoints. Pair-specific regression outperformed the aggregate model with an R-squared of 0.84 versus 0.54, while the boundary-corrected method increased the mean adjusted R-squared from 0.66 to 0.95, proving optimal for 81% of vehicle pairs under the Bayesian Information Criterion.

Why it matters — It demonstrates that the primary source of estimation error in Newell-type models is not a failure of their core behavioral assumptions, but rather boundary effects at the start and end of vehicle interactions. Accounting for these temporal boundaries dramatically improves parameter calibration accuracy without adding model complexity, which is essential for simulating unstructured, lane-free traffic.

Caveat: The evaluation is restricted to a single highway corridor in Chennai, meaning the performance of the boundary-correction method under different geometric layouts or highly congested urban intersections remains untested.

arXiv · pdf

INTENT: An LSTM Framework for Vehicle Intention Prediction in Intersection Scenarios with Comprehensive Ablation Analysis

The INTENT framework uses a Long Short-Term Memory (LSTM) model to predict whether a vehicle at an intersection will go straight, turn left, or turn right. Evaluated on the inD dataset of German intersection traffic, the model predicts these maneuvers 2 seconds before they occur with an accuracy of 99.71%.

Why it matters — It demonstrates that vehicle intent can be classified with near-perfect accuracy well ahead of the maneuver, providing a reliable input that can be used to improve real-time trajectory prediction and collision avoidance systems in autonomous vehicles.

Caveat: The framework's high accuracy was evaluated on a single specialized dataset of German intersections, which may not reflect the driving behaviors, intersection geometries, or sensor noise profiles of other regions.

arXiv · pdf

A Comparative Review of Methods to Create a Composite Index for Sustainable and Inclusive Wellbeing

The study evaluates 13 different aggregation methods for constructing a composite index of Sustainable and Inclusive Wellbeing (SIW) against nine theoretical conditions derived from needs theory and strong sustainability. Using an illustrative country-level dataset, the analysis demonstrates how different mathematical approaches—ranging from simple arithmetic means to machine learning, data envelopment analysis, and penalty-based indices—alter final rankings.

Why it matters — It demonstrates that conventional compensatory aggregation methods produce highly similar, potentially misleading rankings that fail to respect environmental ceilings or lower limits. By mapping which mathematical techniques satisfy specific sustainability criteria, it provides a concrete framework for designing non-compensatory, multi-level indicators required for 'Beyond GDP' policy frameworks.

Caveat: The comparison relies on an illustrative example rather than a newly deployed, fully realized global index.

arXiv · pdf

Large Multimodal Model-Based Environment-Aware Mobility Management

This paper introduces a mobility management scheme for wireless networks that uses Large Multimodal Models (LMMs) to process RGB-D camera images of the surrounding environment. The model extracts contextual information about static reflectors and dynamic obstacles to map user equipment and small base station positions to channel capacity, predicting future capacities along user trajectories to make proactive handover decisions. Simulation results show that this environment-aware approach improves cumulative channel capacity compared to traditional deep learning methods.

Why it matters — By integrating visual sensing data directly into wireless handover decisions, the system can anticipate signal blockages and reflections before they affect connection quality, a capability missing from conventional mobility management that relies solely on radio frequency measurements.

Caveat: The performance improvements and model feasibility are demonstrated exclusively through simulations rather than real-world physical deployments.

arXiv · pdf