Yingxue Zhang, an assistant professor in Binghamton University’s School of Computing, has received a five-year, $584,649 National Science Foundation CAREER Award to investigate how offline reinforcement learning can be applied to urban environments. The prestigious award supports early-career researchers whose work has the potential to shape their fields. Zhang’s project is among the first efforts to use this particular form of machine learning to model decision-making across the spatial and temporal dimensions of a living city.

Reinforcement learning is an AI technique in which a system learns how to act by observing an environment, taking actions, and receiving feedback. A robotic vacuum, for example, can gradually learn the layout of a room by exploring it, encountering obstacles, and adjusting its behavior. Over time, these experiences allow the system to develop a “policy,” or a strategy for choosing actions under different conditions. In a city, a similar policy might determine how a taxi should search for passengers, how a traveler should select a route, or how transportation resources should respond to changing demand.

Traditional reinforcement learning, however, depends on trial and error. That approach may work for a robot operating in a controlled setting, but it is risky when applied directly to urban systems. An AI model that experiments with traffic signals, transit recommendations, or vehicle routes could cause congestion, delays, or safety problems while it learns. Offline reinforcement learning avoids this danger by training models on historical data rather than allowing them to test unproven actions in the real world.

The challenge is that urban data is not clean or uniform. GPS signals can be inaccurate, vehicles and mobile devices may report information at different frequencies, and transportation records come from many independent sources. Each person may contribute only a small amount of information, while interactions among millions of people create complex patterns. A commuter’s decision affects congestion, congestion changes travel times, and those travel times influence the decisions of other commuters. Zhang’s research will examine how AI models can extract reliable decision-making patterns from this fragmented and highly diverse data.

Urban systems are also spatial-temporal systems, meaning that location and time are inseparable. A traffic pattern on one street can affect neighboring roads, while a delay during the morning rush hour may create consequences throughout the day. This combination makes city data substantially more complicated than information collected from many conventional robotics applications. Zhang’s models will need to account for spatial correlations, such as nearby vehicles moving together, as well as temporal correlations, such as recurring congestion or sudden disruptions caused by accidents and weather.

Another major obstacle is known as distribution shift. The data used to train an AI system may describe one environment, while the model is eventually deployed in another. A system trained using transportation patterns from one city may perform poorly in a different city with unfamiliar roads, travel habits, infrastructure, or demographics. Even within the same city, behavior can change over time. Zhang plans to develop methods that make offline reinforcement-learning policies more adaptable, reliable, and useful when real-world conditions differ from the training data.

The project will include safeguards before any resulting system is used in practical urban settings. Zhang and collaborators at the University of Maryland, College Park, the University of Pittsburgh, and institutions in Hong Kong will test their methods in simulated environments. Simulations allow researchers to expose models to unusual conditions and evaluate their decisions without placing people or transportation networks at risk. The research will also explore how natural-language information can improve machine-learning systems. Descriptions of an environment, a task, or an object could give an AI model additional context that raw sensor data alone might miss.

Zhang’s work is intended to influence more than transportation technology. The project is expected to support new courses and training in offline reinforcement learning, while outreach to schools and educators could introduce younger students to artificial intelligence and machine learning. Industry collaborations may eventually lead to workshops and broader workforce-development programs. At the end of the five-year award, Zhang plans to make the project’s results open source, allowing researchers, developers, and communities to examine, reproduce, and improve the models. By transforming everyday movement into a resource for safer and more responsive urban planning, the research could help cities learn from their own inhabitants without asking the real world to serve as an uncontrolled laboratory.

Subject of Research: Offline reinforcement learning for urban decision-making, transportation systems, spatial-temporal data, and smart-city applications.

Web References:
Binghamton University faculty profile: https://www.binghamton.edu/computer-science/people/profile.html?id=yzhang42
NSF CAREER Award: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2539629
Thomas J. Watson College of Engineering and Applied Science: https://www.binghamton.edu/watson
Binghamton University School of Computing: https://www.binghamton.edu/computer-science

Image Credits: Binghamton University, State University of New York

Keywords

Machine learning, artificial intelligence, offline reinforcement learning, computer modeling, urban planning, urban studies, traffic flow, cities, transportation engineering, smart cities

Hosted by HKU’s Urban Systems Institute (USI) in collaboration with the Institute for Climate and Carbon Neutrality, the State Key Laboratory of Subtropical Building and Urban Science Hong Kong Base, the Otto Poon Charitable Foundation Smart Cities Research Institute, The Hong Kong Polytechnic University and the IEEE Geoscience and Remote Sensing Society Hong Kong Chapter, the forum brought together researchers working across geography, ecology, economics, sociology, engineering and computer science. Its central message was that sustainable cities will require models capable of connecting neighborhood-scale changes with regional and global forces, including migration, capital flows, climate risk and resource consumption.

In his opening address, HKU Vice-President for Academic Development Professor Peng Gong said conventional urban research often depends too heavily on qualitative descriptions and narrowly defined case studies. Such approaches, he argued, are poorly suited to cities whose systems are nonlinear, interconnected and constantly changing. He called for an urban research framework built on interdisciplinary integration, quantitative modeling and a global perspective. Mathematical and computational models should not replace human judgment, he said, but should provide empirical evidence for it, allowing researchers to evaluate multiple objectives at once, including economic prosperity, social inclusion, public health, environmental quality and carbon reduction.

Gong also introduced a proposed “Exquisite City” framework for sustainable Chinese urban development. Rather than measuring progress mainly through gross domestic product, the framework evaluates urban performance through inputs such as labor, capital, land and carbon footprint, together with 48 indicators covering health, inclusivity, prosperity, ecology and livability. Drawing on ten years of data from 148 Chinese cities, the approach is designed to identify whether development is becoming more efficient, equitable and environmentally sustainable. Its proponents say such multidimensional assessment could help cities move from broad growth targets toward more precise, low-carbon and quality-oriented planning.

Demographic change emerged as another major force reshaping urban systems. Professor Cindy Fan of the University of California, Los Angeles, examined China’s sustained population and fertility decline, linking it to structural changes in families, education, employment and housing costs. The one-child era produced a “4-2-1” family structure in which one child may eventually support two parents and four grandparents, increasing the cost of care and weakening incentives to have larger families. Fan emphasized that higher female educational attainment, workforce participation, urbanization and expensive living conditions are deeply connected to fertility decisions. Financial incentives alone, she suggested, are unlikely to reverse the trend without broader reforms in social welfare, family support and care institutions.

A parallel transformation is taking place in the field of urban informatics, described by Professor John Wenzhong Shi of The Hong Kong Polytechnic University as a discipline with five interconnected dimensions: urban science, urban sensing, urban big-data infrastructure, urban computing, and urban systems and applications. Technologies such as GeoAI, mobile three-dimensional mapping and digital twins can now integrate observations from satellites, sensors, vehicles and human activity. These tools make it possible to simulate urban conditions and test interventions before they are implemented. Examples presented at the forum included spatiotemporal prediction of pandemic risk and detailed real-world modeling of streets, buildings and infrastructure.

Mobility data was identified as a critical missing variable in many traditional urban studies. Professor Bo Huang of HKU argued that static population counts and land-use statistics cannot adequately describe cities in which people move continuously between homes, workplaces, schools and services. His Geographically and Temporally Weighted Regression model, known as GTWR, incorporates changing relationships across both location and time. Applications involving pandemic transmission and heatwave exposure showed how mobility data can identify vulnerable populations and reveal gaps in urban governance. The approach could support more targeted public-health responses, transport planning and low-carbon strategies in densely populated cities.

Artificial intelligence was presented not merely as a tool for processing information, but as a potential research partner. Professor Yong Li of Tsinghua University described an “AI Urban Scientist” system that combines tens of thousands of research papers with multiple urban datasets. The system can generate research hypotheses, fuse spatiotemporal information, conduct simulations and empirical calculations, and produce evidence-based conclusions. Other researchers applied AI to long-term street monitoring, tree-canopy analysis and health-risk mapping. Five years of street-view imagery from Xining, totaling more than 225 gigabytes, revealed how annual monitoring can expose overlooked renewal needs and environmental decline. Satellite imagery analyzed with the U-Net deep-learning architecture enabled individual tree crowns to be segmented and measured across large landscapes, providing detailed evidence of ecological disturbance and urban forest resilience.

The forum also highlighted the value of linking urban models to physical materials, ecosystems and human health. Researchers presented studies showing that mixed urban forests can provide greater ecological benefits than single-species stands, while multi-sector models can connect land expansion with carbon budgets using satellite observations, road networks and industrial statistics. Other projects examined how the built environment influences shared-mobility behavior across macro-, meso- and micro-scales; how Bayesian inference can detect hidden structures in noisy transportation networks; and how real-time geographic information systems can map environmental exposure and health risks hour by hour. Low-carbon building materials, including bio-based products, low-carbon concrete and stabilized earth blocks, were assessed for their potential to meet future housing demand without undermining climate goals.

During two roundtable discussions, participants addressed both the promise and the risks of increasingly automated urban science. Artificial intelligence, large-scale models, network inference and real-time GIS could accelerate discovery and improve planning, but they also raise concerns about algorithmic bias, energy consumption, computing costs and the interpretability of model outputs. Speakers argued that human expertise remains essential, particularly when models influence decisions about vulnerable communities, housing, transport or public health. Closing the forum, Gong stressed that accurate information about population distribution, demographic structure and human mobility is the foundation of effective policy. He added that robotics and AI may help societies respond to aging populations and labor shortages, but insisted that the ultimate purpose of urban systems research is to convert scientific knowledge into practical governance that respects Earth’s ecological limits and supports sustainable human settlements.

Subject of Research: Advanced modeling of urban systems for sustainable cities, including artificial intelligence, urban informatics, remote sensing, mobility analysis, demographic change, ecological monitoring and low-carbon planning.

Article Title: AI, Mobility Data and Digital Twins Point to a New Science of Sustainable Cities

News Publication Date: 19 July 2026

Image Credits: The University of Hong Kong

Keywords: Human geography; demography; remote sensing; urban systems; artificial intelligence; digital twins; urban informatics; sustainable cities; mobility data; climate neutrality

This ambitious five-year project, underpinned by a significant investment of US$9.4 million from the Grundfos Foundation—the foundation’s largest research grant to date—aims to revolutionize urban cooling by developing intelligent, water-based sustainable systems tailored for megacities. The initiative, titled Sustainable Water-based Cooling in Megacities (SWiM), leverages the complementary strengths of Danish and Singaporean urban infrastructure innovation. Through integrated research spanning engineering, artificial intelligence, and urban planning, SWiM seeks to break the entrenched cycle of high energy consumption and carbon emissions caused by conventional cooling technologies.

District cooling and heating technologies form the foundational expertise upon which this project builds. Denmark, a global pioneer in district heating systems, has long demonstrated the efficiency advantages of centralized thermal energy distribution. Facilities such as the Avedøre Power Station and the Amager Bakke waste-to-energy plant epitomize cutting-edge combined heat and power technologies, providing sustainable, large-scale thermal solutions. Meanwhile, Singapore has adeptly adapted these concepts into district cooling networks optimized for tropical urban conditions. The Marina Bay district’s extensive underground chilled water pipeline system exemplifies this, significantly reducing carbon emissions citywide.

Despite these successes, current district cooling installations in megacities are typically limited in their geographical coverage and scalability. Business districts and housing estates can be served effectively, but extending these benefits to entire cities requires overcoming substantial technical challenges. The SWiM project directly addresses these barriers by focusing on scalable, modular cooling architectures enabled by advanced control systems. These systems are designed to respond dynamically to varying urban environments, demand fluctuations, and operational anomalies.

Central to the SWiM initiative is the development of autonomous control mechanisms that can ensure reliable, fault-tolerant operation without the need for constant expert supervision. This autonomy is critical for deployment in complex urban settings where human error, potential cyber-attacks, and equipment failures could otherwise compromise system integrity. Aarhus University’s expertise in electrical and computer engineering drives this domain, utilizing digital twin technologies that model physical cooling infrastructure and support adaptive control strategies. Such digital replicas provide real-time operational insights, enabling predictive maintenance and optimal system adjustments.

Artificial intelligence plays a transformative role in the SWiM framework. By integrating machine learning algorithms, the system can monitor performance continuously, detect inefficiencies or faults early, and employ predictive analytics to prevent downtime. Notably, the project incorporates smart algorithms that balance the competing demands of cooling load, energy efficiency, and grid stability. This ensures that cooling systems contribute positively to the broader urban energy ecosystem rather than destabilizing it.

A distinctive aspect of this research is its focus on applicability under real-world conditions. SWiM’s approach transcends laboratory testing by constructing physical testbeds at multiple scales — room, floor, and building levels — within Singapore’s urban fabric. These physical environments will be complemented by comprehensive digital twin simulations, enabling scalable replication of system behavior across various city scenarios. Such rigorous validation is essential for transitioning innovations into practical, large-scale solutions that city planners and policymakers can adopt confidently.

The collaborative nature of SWiM, uniting Singaporean and Danish academic and industrial stakeholders, embodies a model for global scientific partnership. With Grundfos Foundation’s funding strategically underpinning the initiative, industry knowledge will be deeply integrated into research outcomes to ensure feasibility and immediate applicability. This collaboration is particularly timely as both Singapore and Denmark pursue ambitious climate objectives—Singapore targeting net-zero emissions by 2050 and Denmark aiming for climate neutrality by 2045.

Professor Madhavi Srinivasan of NTU Singapore highlights the convergence of interdisciplinary expertise in this project, noting how the blend of sustainability science, engineering, and artificial intelligence can yield cutting-edge urban cooling solutions. Similarly, Professor Rafael Wisniewski of Aalborg University underscores the importance of developing systems that are not only theoretically sound but also resilient and user-friendly, capable of deployment without reliance on specialist intervention.

The envisioned integration of digital tools such as Building Information Models (BIM) with real-time monitoring systems promises unprecedented precision in managing energy flow and cooling demands. Professor Peter Gorm Larsen of Aarhus University elaborates on how digital twins will facilitate seamless transitions between operational states, ensuring that cooling resources are allocated efficiently under varying conditions.

SWiM’s innovations aim to disrupt the current paradigm, making cooling systems vital components of sustainable urban infrastructure rather than significant contributors to environmental degradation. By combining low-energy water-based cooling methods with intelligent control architectures and comprehensive urban planning tools, the project charts a visionary pathway for megacities grappling with the twin crises of urban heat and climate change.

As urban populations continue to expand, particularly in tropical megacities, the stakes for sustainable cooling solutions have never been higher. SWiM represents a bold leap forward, promising to reduce city-wide energy consumption for cooling by up to 30 percent—a transformative achievement with profound implications for global carbon emissions and urban liveability.

In the coming years, the success of SWiM will be measured not only by technological milestones but also by its ability to influence policy, shape standards, and catalyze widespread adoption of sustainable cooling infrastructures worldwide. This initiative underscores the critical role of cross-border collaboration and innovation in addressing one of the defining environmental challenges of our time.

Subject of Research: Sustainable urban cooling systems for megacities involving water-based, intelligent district cooling technologies.

Article Title: (Not provided)

News Publication Date: (Not provided)

Web References: (Not provided)

References: (Not provided)

Image Credits: Rasmus Reimer Larsen

Keywords: Applied sciences and engineering, Systems engineering, Mechanical engineering, Electrical engineering, Civil engineering, Computational science

Published in the prestigious journal Landscape Ecology, this pioneering research introduces a multi-temporal vegetation visualization methodology, combining cutting-edge deep learning algorithms with 3D reconstruction techniques derived from street view imagery. The innovative integration of artificial intelligence (AI) technologies not only enhances the granularity of urban vegetation assessment but also allows dynamic modeling of plant color shifts, structural variations, and phenological stages across different seasons. This level of temporal and species-specific analysis stands as a paradigm shift from traditional green view metrics, which tend to be static and less detailed.

The core advancement lies in the creation of the Seasonal Species-Specific Plant View Index, an AI-powered system capable of distinguishing among 51 different urban plant species with impressive average identification accuracy of over 82%. This index harnesses convolutional neural networks trained on vast datasets of street-level photographic information, paired with sophisticated 3D spatial reconstruction to correct for distortions and gaps inherent in panoramic imagery. Consequently, the system can isolate plants displaying striking seasonal visual phenomena—such as the ephemeral cherry blossoms of spring and the fiery hues of maple foliage in autumn—enabling a nuanced appreciation of urban botanical diversity.

The team, led by Anqi Hu and senior researcher Tomohiro Fukuda, selected Suita City in Osaka Prefecture as their urban laboratory to validate the framework. By deploying sensors and computational processing pipelines, they generated standardized viewpoints that can automatically adjust over temporal sequences and spatial coordinates. This approach surmounts the issues of inconsistent coverage and perspective distortion in street imagery, ensuring robust and uniform data for plant species identification and seasonal visualization. Such precision is critical for urban designers who aim to orchestrate greenery arrangements that reinforce biodiversity while maximizing aesthetic pleasure year-round.

One of the remarkable implications of this technological breakthrough lies in its applicability to the restoration of brownfield sites, often characterized by heterogeneous and neglected landscapes. The framework’s capacity to virtually simulate and visualize diverse plant species over time equips planners with a powerful tool to redesign and revitalize marginalized urban areas. By strategically introducing species with complementary growth patterns, colors, and ecological functions, it is possible to foster greener, more resilient, and visually vibrant environments that contribute to urban sustainability and inhabitant well-being.

Furthermore, the approach elevates the concept of 4D urban design, where the dimension of time is embedded into spatial planning processes. This multi-temporal perspective allows the anticipation of phenological transitions and their impact on urban aesthetics and ecological services. In practice, the methodology facilitates scenario modeling that can predict how different vegetation assemblages will evolve across seasons, informing maintenance schedules, planting strategies, and public engagement efforts. Beyond ecological modeling, it reinforces a symbiotic relationship between nature and human experience in the urban milieu.

The integration of AI-driven deep learning with 3D reconstruction also signals a broader shift in how urban ecological data is collected and interpreted. Traditional remote sensing tools often lack the spatial resolution or temporal frequency required for species-specific monitoring in complex urban contexts. In contrast, street view imagery, ubiquitously available and frequently updated, provides rich, high-resolution visuals that, when analyzed with sophisticated AI tools, reveal intricate vegetation patterns that were previously hidden. This democratization of urban ecological data leverages existing technologies for scalable, cost-effective green space management.

Hu highlights that color diversity and species richness significantly enhance the psychological benefits derived from urban green spaces, contributing to what she terms the ‘feel-good’ factor. By enabling planners to visualize these variables dynamically, the new framework paves the way for creating multifunctional urban landscapes that resonate with human emotions and ecological imperatives. This aligns with growing evidence that well-designed green spaces mitigate stress, foster social interaction, and promote physical activity, thereby improving overall quality of life in dense metropolitan areas.

Another technical triumph of this research is its ability to automate the generation of consistent viewpoints in both spatial and temporal domains. This is particularly valuable for longitudinal studies aiming to track urban vegetation changes over months or years, permitting the systematic evaluation of urban greening initiatives and natural phenological cycles. Such automation minimizes the labor-intensive nature of manual data collection and increases the reliability of monitoring programs, providing urban ecologists and planners with actionable intelligence.

The study’s findings and tools hold the promise of revolutionizing urban landscaping, creating greener cities that are not only ecologically healthier but also aesthetically richer and emotionally rewarding. The fusion of AI, 3D spatial analytics, and seasonal dynamics sets a new benchmark for urban environmental research and practice. It also opens up pathways for integrating citizen science and participatory planning, where community members can visualize and contribute to the greening process, fostering shared stewardship of urban natural assets.

In an era where urban expansion often threatens ecological integrity, this research underscores the importance of innovative technological solutions to sustain and amplify urban biodiversity. By meticulously mapping how plant species distribute, evolve, and influence cityscapes through time, the work from Osaka University offers a scalable, precise, and dynamic toolkit that meets both scientific and societal needs. Its implications extend to public health, urban policy, and sustainable development, making it a landmark contribution to contemporary urban ecology.

As cities worldwide grapple with climate change challenges, environmental degradation, and social inequities, the ability to design, monitor, and adapt green spaces with such forward-looking precision is invaluable. This new AI-driven framework enriches the urban planning arsenal, ensuring that green infrastructure can be optimized to deliver ecosystem services, economic benefits, and human happiness simultaneously. Ultimately, it exemplifies how the intersection of technology, ecology, and design can inspire more harmonious and resilient urban futures.

Subject of Research: Not applicable

Article Title: Multi-temporal Analysis of Urban Vegetation Using Deep Learning and 3D Reconstruction

News Publication Date: 1-Jul-2025

Web References:
https://doi.org/10.1007/s10980-025-02090-4

Image Credits: 2025 Anqi Hu et al., Landscape Ecology

Keywords:
Social sciences, Urban planning, Urbanization, Land management, Land use, Cities, Species diversity, Species richness, Quantitative modeling, Sustainability