Urban Currents

Smart routes: a system for development and comparison of algorithms for solving vehicle routing problems with realistic constraints

The researchers developed the Smart Routes platform to compare exact, heuristic, and deep learning models for solving the Capacitated Vehicle Routing Problem with Time Windows (CVRPTW). Benchmarks comparing the SCIP exact solver against heuristic algorithms (LKH, 2-OPT, 3-OPT, ORTools) and a deep learning model (JAMPR) show that for problem sizes of 50, heuristic and neural solutions match the exact solver's quality in less time. For problem sizes of 100, the exact solver is approximately 13 times slower than neural and classical heuristics for the same route cost, and yields 50% worse route costs when restricted to the same execution time.

Why it matters — It establishes a benchmarked platform that allows researchers to directly compare traditional optimization solvers with emerging deep learning routing models under realistic urban constraints, demonstrating where neural models achieve parity in solution quality at a fraction of the computational cost.

Caveat: The performance comparisons are limited to problem sizes of up to 100 nodes.

arXiv · pdf

Local capacity building and policy evaluation in climate and environmental governance

The study develops and applies the Local Climate-Environmental Institutional Capacity Index (ICI-CLIMA), a multicriteria evaluation framework with seven valuation components, to assess local institutional capacities in Mexico City. The index was calculated and compared for the years 2019 and 2022 to track changes in climate and environmental governance capabilities.

Why it matters — It provides a structured, quantitative index to measure and monitor local institutional capacity, shifting the focus of urban climate governance from theoretical policy design to the empirical evaluation of administrative readiness and learning systems.

Caveat: The index and its components are demonstrated using only a single case study of Mexico City, which may limit its immediate generalizability to smaller municipalities with different administrative structures.

Discover Cities · pdf · doi

Physical activity from bus travel: objective estimates linking accelerometer and travel card data

The study linked accelerometer data with smartcard-recorded bus boardings for 298 participants in Tasmania over three timepoints of up to 7 days each. It identified 1,396 active travel bouts to bus stops, finding that these bouts averaged 6.8 minutes and 473 steps, with 93% of the activity classified as moderate intensity.

Why it matters — It provides objective, empirical measurements of the physical activity generated by routine bus use, demonstrating that a round-trip bus commute four days a week can fulfill approximately 75% of the recommended weekly physical activity for adults.

Caveat: The calculations assume that the unmeasured egress walk from the bus to the final destination is of equal duration and intensity to the measured access walk to the bus stop.

Journal of Transport & Health · doi

Unveiling nature-based tourism destinations on the Qinghai-Tibet plateau: A journey-based semantic embedding framework

This study developed a journey-based semantic embedding framework to identify nature-based tourism destinations using GPS trajectory data from an outdoor recreation platform on the Qinghai-Tibet Plateau. The framework maps destinations based on collective travel flows rather than static point-density, revealing that 63.2% of identified destinations overlapping national nature reserves lack formal management systems. Temporal analysis from 2017 onward shows a decline in newly emerging destinations and faster decline rates for newer sites compared to established ones.

Why it matters — It shifts the paradigm of tourism destination mapping from static spatial clusters to flow-based journeys, exposing linear destinations like highways and showing that administrative closures of nature reserves often redirect visitor flows to adjacent, unregulated areas rather than stopping them.

Caveat: The findings rely on GPS trajectory data from a single outdoor recreation platform, which may reflect the specific demographics and behaviors of its user base rather than all tourists.

Journal of Transport Geography · doi

Data-driven rebalancing support for station-based bike sharing: A hybrid centralized-distributed deployment in Chiba city

This study developed and tested a hybrid centralized-distributed rebalancing framework for the HELLO CYCLING electric bike-sharing network in Chiba City, Japan. Instead of complex optimization models, the system uses historical demand and historical empty or full events to generate a station-level decision table of target inventory ranges and priorities for field staff. The framework was evaluated during a live two-week deployment and compared against both a pre-intervention baseline and a randomized relocation baseline.

Why it matters — It demonstrates that a low-cost, rule-based decision support tool can outperform both routine heuristic operations and randomized relocations in avoiding demand losses. This provides smaller or regional bike-share operators with a computationally light, easily integrated alternative to rigid, centrally optimized routing systems that are difficult to maintain during real-time traffic disruptions.

Caveat: The evaluation was limited to a single two-week live deployment in one city, and the execution relied on contractor-based field staff rather than crowdsourced user-side rebalancing.

Transportation Research Part A Policy and Practice · doi

Digital twin as regulatory sandbox for intersection-level traffic management and infrastructure decision-making: A case study in Leipzig, Germany

The researchers developed a high-fidelity digital-twin-based regulatory sandbox that evaluates detector deployment schemes and signal control algorithms under real-world operational constraints. Tested at a single signalized intersection in Leipzig, Germany, the system uses imitation learning to calibrate driving behavior and a controller-compatible middleware to run actual signal control files. The simulation demonstrated that camera-based detection yielded greater operational benefits than loop or LiDAR systems under the current control logic, and successfully quantified performance gains from a candidate control algorithm.

Why it matters — It provides a risk-free, virtual environment for municipal traffic authorities to test and compare alternative signal programs and sensor configurations in parallel before investing in physical infrastructure or deploying new algorithms.

Caveat: The sandbox was demonstrated and validated at only a single intersection in Leipzig, Germany.

Transportation Research Part C Emerging Technologies · doi

Predicting moose (Alces alces) activity at roadside mineral licks: Implications for moose-vehicle collisions

Using camera trap data collected over a 12-year period at four roadside mineral licks in north-central British Columbia, Canada, this study analyzed the environmental drivers of moose visits. Generalized linear mixed models revealed that annual variations in the timing of peak spring and summer lick visits are best explained by snowpack levels and the timing of spring green-up.

Why it matters — By identifying the specific environmental cues that trigger peak moose activity near highways, road safety planners can transition from static seasonal warnings to dynamic, weather-responsive mitigation measures to prevent moose-vehicle collisions.

Caveat: The predictive models are based on a small sample of only four roadside mineral licks within a single region of British Columbia.

Transportation Research Part D Transport and Environment · doi

Macroscopic fundamental diagram-based stability analysis of emerging low-altitude transport networks with multi-dimensional corridor configurations

This study develops a macroscopic fundamental diagram (MFD) framework to analyze the stability and capacity of a two-region low-altitude air transport network. It models safety-constrained aircraft movements using a channelized distributed collision-avoidance model across linear, planar, and three-dimensional corridor configurations under cargo-only, passenger-only, and mixed operations. Upgrading from linear to planar and three-dimensional configurations improved the mean end-point relative change of the stable region area from -4.1039% to -1.0768% and -0.9204% respectively for mixed operations, and from -2.3025% to -0.8071% and -0.3381% for cargo-only operations.

Why it matters — It establishes a quantitative method to evaluate how physical airspace corridor dimensions and flight operation modes interact to prevent congestion collapse in emerging urban air mobility networks, proving that three-dimensional corridors offer statistically significant stability benefits for cargo and mixed flights but not for passenger-only flights.

Caveat: The findings are based on a simulated two-region airport-centered network model rather than empirical flight data from an active urban air transport system.

Transportation Research Interdisciplinary Perspectives · pdf · doi

Behavioral Dynamics at Signalized Intersections during a Green Phase: A Data-Driven Model for Speed Prediction in Mixed-Traffic Conditions

This study analyzed vehicular speed and acceleration profiles at four signalized intersections in Nashik and Nagpur, India, using UAV video data collected 100 meters upstream and 40 meters downstream of the stop line. Tracking four vehicle classes across 20-meter intervals, the researchers found that cars and two-wheelers exhibited the most predictable speed behaviors, with polynomial fits reaching R-squared values above 0.90 and 0.58 to 0.74 respectively, while three-wheelers and heavy vehicles behaved more erratically. These empirical profiles were used to construct a universal polynomial model framework that predicts mean vehicle speeds during the green phase based on distance from the stop line.

Why it matters — It provides a standardized mathematical framework to model and calibrate traffic microsimulations in highly heterogeneous, non-lane-disciplined environments, a capability previously hindered by ground-level line-of-sight obstructions.

Caveat: The predictive model is calibrated specifically for green-phase dynamics and may not capture behaviors during transition intervals or red phases.

Journal of Transportation Engineering Part A Systems · doi

Urbanization, Land Use Change, and Land Surface Temperature: A Remote Sensing-Based Assessment for Sustainable City Planning and Development

This study analyzed the spatial changes in land use, land cover, and land surface temperature in Bhopal, India, over a 20-year period (2000, 2010, and 2020) using Landsat 5-TM and Landsat 8-OLI satellite imagery. By calculating six remote sensing indices, including NDVI, NDBI, and NDWI, the research quantified how the expansion of built-up areas and bare soil, alongside the loss of vegetation and water bodies, correlated with rising urban temperatures.

Why it matters — It provides empirical, long-term spatial evidence of how specific land cover transitions drive local temperature increases, offering urban planners concrete localized data to support climate-resilient zoning and green infrastructure policies.

Caveat: The study relies on decadal interval satellite observations, which may overlook short-term seasonal variations or rapid transitional phases in land surface temperature dynamics.

Journal of Urban Planning and Development · doi

Research on the Impact of FinTech Development on Urban Innovation in China: An Empirical Analysis Based on 282 Cities

An empirical analysis of panel data from 282 Chinese cities between 2007 and 2020 shows that a 1% increase in the FinTech index is associated with a 4.947% increase in urban innovation. The positive effect is strongest in eastern, coastal, densely populated, and resource-based cities. Additionally, the relationship is nonlinear, requiring a development threshold of 4.890 to become significant and accelerating further after passing a second threshold of 5.727.

Why it matters — It establishes that FinTech's capacity to stimulate urban innovation is not uniform, demonstrating that its benefits require a baseline level of financial technology development to materialize and are highly dependent on existing market maturity and economic openness.

Caveat: The study relies on a composite FinTech index and urban innovation proxies to establish a correlation, which may not fully capture the specific mechanisms of local financing.

Journal of Urban Planning and Development · doi

Human-Centered Planning Through Design Thinking: Patterns, Possibilities, and Perceptions

An online survey of 86 U.S. planning practitioners, conducted in partnership with the American Planning Association, analyzed how design thinking is applied in professional practice. The study found that 87% of respondents use design thinking in their work, frequently employing prototyping and testing across diverse project scales, including comprehensive, long-range plans.

Why it matters — It provides the first national-level evidence of how U.S. planners use design thinking, challenging previous assumptions that the methodology is rarely prototyped in planning or that it is ill-suited for complex, large-scale projects.

Caveat: The findings rely on a self-selected sample of 86 planners recruited through a design thinking interest group, which likely skews the sample toward active practitioners of the method.

Journal of the American Planning Association · pdf · doi

Advancing Climate and Equity Goals in Regional Transportation Planning: Learning from California

Through a thematic analysis of interviews and focus groups with 11 board members, 8 community advocates, and 13 senior staff across seven California metropolitan planning organizations (MPOs), this study examines how regional bodies integrate greenhouse gas reduction and socioeconomic equity into transportation plans. The findings show that while state mandates like Senate Bill 375 have successfully shifted planning discussions and project centering, MPOs struggle to resolve conflicting definitions of equity and frequently fail to abandon legacy projects that increase emissions or worsen inequality.

Why it matters — It demonstrates that legislative mandates alone are insufficient to displace status quo infrastructure commitments without explicit state guidance on equity metrics, structural reforms to MPO governance, and dedicated consensus-building processes.

Caveat: The findings are based on qualitative self-reports from a small sample of stakeholders within a single state's regulatory framework, which may limit generalizability to states without equivalent climate mandates.

Journal of the American Planning Association · doi

From crown candidates to neighborhood screening: integrating optical GeoAI and spatial modeling for urban-canopy assessment in Davis, California

The study developed an optical GeoAI workflow using 0.6-meter resolution 2022 National Agriculture Imagery Program imagery to map the urban canopy in Davis, California. By combining DeepForest, NDVI thresholding, and the Segment Anything Model, the workflow identified 11,741 candidate tree crowns and mapped 2.43 square kilometers of canopy, which represents 9.37% of the city. When validated against a 2022 LiDAR-assisted reference product, the optical method achieved 87.8% pixel agreement and 97.4% center agreement, capturing 34.2% of the reference canopy area.

Why it matters — It demonstrates a reproducible, updateable screening framework that can estimate urban canopy and its cooling effects using widely available optical imagery, reducing the reliance on expensive and less frequent LiDAR flights for municipal planning.

Caveat: The optical workflow significantly underestimated the total canopy area, capturing only about one-third of the canopy area identified by the reference LiDAR product.

arXiv · pdf · code

Deep Reinforcement Learning solution for pickup and delivery routing problems with time window and capacity constraints

The study adapted the JAMPR deep reinforcement learning model to solve the Capacitated Pickup and Delivery Problem with Time Windows (CPDPTW). The modified model generates fast, optimal routing solutions for small and medium-sized problems, and fast, suboptimal solutions for larger problems containing more than 200 nodes.

Why it matters — It demonstrates the first successful application of deep reinforcement learning to this specific, highly constrained routing problem, offering a real-time optimization capability for medium-to-large logistics operations where classical exact methods fail to scale.

arXiv · pdf