Contraction theory provides a dynamically intrinsic framework for characterizing and controlling how real-world nonlinear systems should behave to achieve their operational ideals. While rooted in classical model-based control theory, the concept is increasingly recognized as a broader principle that might extend to learning-based, data-driven, or model-free settings central to intelligent control architectures equipped with modern AI tools. This talk provides a brief mathematical overview of contraction theory, along with some of our recent efforts to generalize the ideas to broader problem settings. Examples from collaborations with NASA-JPL and DARPA illustrate applications in aerospace and robotic autonomy. 

Dr. Hiroyasu Tsukamoto is an Assistant Professor of Aerospace Engineering at the University of Illinois Urbana-Champaign and the director of the N-ACXIS Laboratory (Nonlinear Autonomous Control, Exploration, Intelligence, and Systems). Prior to joining Illinois, he was a Postdoctoral Research Affiliate in Robotics at the NASA Jet Propulsion Laboratory, where he contributed to the Science-Infused Spacecraft Autonomy for Interstellar Object Exploration and Multi-Spacecraft Autonomy Technology Demonstration projects. He received his Ph.D. and M.S. in Space Engineering from the California Institute of Technology in 2018 and 2023, respectively, advised by Prof. Soon-Jo Chung, and his B.S. degree in Aeronautics and Astronautics from Kyoto University, Japan, in 2017. He is the recipient of several awards, including the William F. Ballhaus Prize for the Outstanding Doctoral Dissertation in Space Engineering at Caltech, the Innovators Under 35 Japan Award from MIT Technology Review, the Forbes 30 Under 30 Asia – Class of 2025, and the AIAA Journal of Guidance, Control, and Dynamics Editor’s Choice Award. 

Spatial optimization has long served as a foundational framework in urban planning and geography, enabling decision-makers to efficiently allocate resources, design service networks, and locate critical facilities. At the heart of these mathematical locational models lies the concept of accessibility—the measure of how effectively demands, such as populations, can reach facilities and services. However, despite its critical importance, accurately measuring accessibility remains a profound challenge. Traditional metrics often rely on oversimplified assumptions, such as straight-line distances or static travel times, which fail to capture the complex, dynamic, and multifaceted realities of human mobility, continuous spatial variations, and complex urban networks.

This colloquium introduces the core principles of spatial optimization and explores how the emergence of Urban AI can address these fundamental limitations. By leveraging massive geospatial datasets, predictive modeling, and sophisticated machine learning algorithms, Urban AI can capture the nuanced, real-world dynamics of urban movement that traditional models miss. Also, AI can offer new opportunities and broaden the perspectives of spatial optimization to our society. Ultimately, the talk will discuss how refining the measurement of accessibility through AI not only enhances the accuracy and realism of spatial optimization models but also paves the way for smarter, more equitable locational decision-making in future cities.

Dr. Seonga Cho is a researcher and educator currently affiliated with the School of Digital Humanities and Computational Social Sciences at KAIST. He earned his Ph.D. in Geography from the University of California at Santa Barbara in 2024, alongside an M.A. in Statistics in 2022, and holds both an M.A. and B.A. in Geography from Seoul National University. Prior to joining KAIST, Dr. Cho served as a Postdoctoral Researcher for the Blue City Project at the École Polytechnique Fédérale de Lausanne from 2024 to 2025. His interdisciplinary research focuses on multi-objective spatial optimization, locational decision-making models, and geographic information systems, with publications in leading journals such as The Professional Geographer and the International Journal of Geographical Information Science. Throughout his career, Dr. Cho's academic contributions have been recognized with numerous honors, including the 2023 Jack and Laura Dangermond Fellowship and multiple competitive awards from the American Association of Geographers.

Cities are the substrate of modern life: they house us, move us, and mediate nearly every exchange we make. In doing so, they consume an outsized share of global resources (energy, land, materials, and time) while also shaping how their citizens work, travel, collaborate, and develop the human, social, and physical capital that defines a society. To steer this exchange—transforming raw resources into societal value—we need to understand how the day-to-day operation of urban systems influences these outcomes, and we need tools that let us reason about interventions before we make them. Sensors and data analytics have transformed how much of this exchange we can observe, but what we really need are better models. In particular, we need custom, continuously updated, uncertainty-aware, and flexible digital twins of both the physical *and* social systems in a city—from buildings and traffic to curbside activity and human-infrastructure interactions— and we need to be able to calibrate them and deploy them at scale. I'll draw on recent work from my lab and other colleagues at CMU—spanning buildings, transportation, and curbside management—to illustrate the promises and pitfalls of these research directions.

Mario Bergés is a professor in the Department of Civil and Environmental Engineering at Carnegie Mellon University (CMU). He is interested in making our built environment more operationally efficient and robust through the use of information and communication technologies, so that it can better deal with future resource constraints and a changing environment. Currently his work largely focuses on developing approximate inference techniques to extract useful information from sensor data coming from civil infrastructure systems, with a particular focus on buildings and energy efficiency. Bergés is the faculty co-director of the Smart Infrastructure Institute at CMU, as well as the director of the Intelligent Infrastructure Research Lab (INFERLab). Among recent awards, he received the Professor of the Year Award by the ASCE Pittsburgh Chapter in 2018, Outstanding Early Career Researcher award from FIATECH in 2010, and the Dean's Early Career Fellowship from CMU in 2015. Bergés received his B.Sc. in 2004 from the Instituto Tecnológico de Santo Domingo, in the Dominican Republic; and his M.Sc. and Ph.D. in Civil and Environmental Engineering in 2007 and 2010, respectively, both from Carnegie Mellon University.

World Model is an ambiguous term. I will discuss how people define the World Model differently, what they are building, and how I define it myself. And then I will discuss its significance, especially in the context of Robotics.

Minjoon Seo is Co-Founder & CEO at Config and an Associate Professor at KAIST AI. He works on the data infrastructure and technology for robotics. His research background is in foundation models, especially language models and vision-language models. He received the NAACL 2025 Best Paper Award and is recognized as a Forbes Korea Y30s Rising AI Leader

What if a world simulation model could render not an imagined environment but a city that actually exists? Prior generative world models synthesize visually plausible yet artificial environments by imagining all content. We present Seoul World Model (SWM), a cityscale world model grounded in the real city of Seoul. SWM anchors autoregressive video generation through retrieval-augmented conditioning on nearby street-view images.

However, this design introduces several challenges, including temporal misalignment between retrieved references and the dynamic target scene, limited trajectory diversity and data sparsity from vehicle-mounted captures at sparse intervals. We address these challenges through cross-temporal pairing, a large-scale synthetic dataset enabling diverse camera trajectories, and a view interpolation pipeline that synthesizes coherent training videos from sparse street-view images. We further introduce a Virtual Lookahead Sink to stabilize long-horizon generation by continuously re-grounding each chunk to a retrieved image at a future location.

We evaluate SWM against recent video world models across three cities: Seoul, Busan, and Ann Arbor. SWM outperforms existing methods in generating spatially faithful, temporally consistent, long-horizon videos grounded in actual urban environments over trajectories reaching hundreds of meters, while supporting diverse camera movements and text-prompted scenario variations.

Seungryong Kim is an Associate Professor at Kim Jaechul Graduate School of AI, KAIST, Seoul, Korea. Before joining KAIST, he was an assistant professor at Korea University, a postdoctoral researcher in the School of Computer and Communication Science at École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, and a postdoctoral researcher in the School of Electrical and Electronic Engineering at Yonsei University, Seoul, Korea. 

He was a research intern at Microsoft Research Asia (MSRA), Beijing, China. He received the B.S. and Ph.D. degrees in the School of Electrical and Electronic Engineering at Yonsei University, Seoul, Korea, in 2012 and 2018, respectively. His research interests include Computer Vision, Computational Photography, Machine Learning, and Deep Learning, particularly representation learning, visual scene reconstruction and understanding.

Maintaining consistent levels of vigilance over time is critical for workers in high-risk work systems. The decline in performance that often occurs when monitoring and screening for occasional and unpredictable signals, known as vigilance decrement, endangers workers’ health and safety and threatens efficiency. Vigilance decrement is especially concerning in clinical environments, where shift work and long working hours impair clinicians’ ability to sustain attention over extended period. A clinical environment is a place where monitoring vigilance decrement is essential given its impact on clinicians’ job satisfaction productivity, patients’ safety, and healthcare quality. Despite its importance, efforts to monitor and predict vigilance decrement have largely relied on behavioral models that depend on retrospective or laboratory-based data, neither of which is feasible for continuously tracking individuals’ vigilance levels in real-time. To date, no human-technology interfaces that provide feedback to mitigate vigilance decrement in real-world operations exist. This talk presents recent work from the Human and Systems Lab on measuring, modeling, and mitigating vigilance decrement using artificial intelligence and features derived from neural and physiological sensors. The findings demonstrate the potential for integrating these approaches into assessment, training, and intervention tools to enhance human performance and ultimately support safer work environments.

Ji-Eun Kim is an Associate Professor in the Department of Industrial and Systems Engineering at the University of Washington (UW) and the Director of the Human and Systems Lab. Her research centers on designing adaptive interventions that accommodate diverse groups of users. To achieve this, she uses neurophysiological sensors to model and predict human performance. She holds a Ph.D. in Industrial Engineering from the Pennsylvania State University. She is a recipient of the 2023 National Science Foundation (NSF) CAREER Award and the 2020 UW Faculty Appreciation for Career Education & Training Award. Her advisees have received numerous honors, including the Human Factors and Ergonomics Society (HFES) Best Student Paper Award (Human Performance Modeling Technical Group), HFES Council of Technical Groups Student Presenter Award, and the Institute of Industrial and Systems Engineers (IISE) Doctoral Colloquium Dissertation Award. She serves as the Chair of the Human Performance Modeling Technical Group at the HFES.

Climate startups are critical to driving mitigation and adaptation innovation, yet they face distinctive ecosystem demands that generic support conditions alone cannot meet. Despite their growing importance, how climate startup ecosystems actually function and why they differ across cities remains poorly understood. Drawing on a multi-city comparative study of Korea, six key enablers (policy, finance, talent, market access, entrepreneurial culture, and support organizations) are examined to understand how they combine across urban contexts to shape ecosystem outcomes. No single enabler is sufficient, and effective ecosystems can emerge through distinct configurations shaped by each city’s industrial legacy, governance, and adaptive capacity. A climate-specific, configurational lens reveals the underlying logic of urban startup ecosystems offering practical implications for cities and policymakers navigating uneven regional development.

Esther Choi is Research Lead for Nature-Based Solutions and the Private Climate Sector at World Resources Institute. She leads research and engagement initiatives focused on scaling private-sector contributions to climate and nature action by aligning incentives, mobilizing capital and knowledge, and strengthening the enabling environments in which innovative solutions can emerge and scale.

Her current work centers on three intersecting areas: identifying ecosystem enablers for early-stage climate startups - particularly in emerging and developing economies; guiding companies and investors on nature-positive strategies; and advancing de-risking, blended finance, and institutional approaches that catalyze private investment for climate and nature action. Esther is also a Lead Author for the Finance Chapter of the Intergovernmental Panel on Climate Change (IPCC)’s Seventh Assessment Report.

From February to July 2026, Esther serves as a Visiting Professor at the Korea Advanced Institute of Science and Technology (KAIST), where her research and teaching focuses on sustainable finance, climate startup ecosystems, and the institutional conditions that enable climate innovation to scale.

With over a decade of experience spanning climate finance, sustainable development, and international governance, Esther has held research and policy roles at Stanford University's Sustainable Finance Initiative, the Green Climate Fund, the World Bank, and the Global Green Growth Institute. Her work has involved close collaboration with governments, investors, and multilateral institutions across Asia, Africa, and Latin America to design policies, financing mechanisms, and partnerships that support climate-aligned growth.

Esther holds a Ph.D. in Environmental Science, Policy, and Management from the University of California, Berkeley, and a Master’s degree in Environmental Management from the Yale School of the Environment. She is based in the San Francisco Bay Area.

In this seminar, we introduce SimMobility, an agent-based simulation platform that designed to help urban planners and policymakers predict the impacts of mobility services and technologies at an urban scale. We present key applications that evaluate the impacts of introducing Automated Mobility-on-Demand (AMOD) services, designing demand-responsive adaptive transit systems, and integrating passenger-freight mobility across user, operator, and network performance. Furthermore, a recent case study in Korea demonstrates the potential of combining this simulation platform with deep generative models to accurately predict population distribution and travel demand over the large-scale metropolitan region.

Simon Oh is an Associate Professor of Transportation and Mobility in the Department of Mobility Science and Engineering at Korea University, Principal Investigator of the Smart Mobility Lab, and Vice President of the Research and Business Foundation Sejong Campus. He earned PhD degree in Civil and Environmental Engineering from KAIST in 2015 and worked as a Senior Postdoctoral Associate at the Future Urban Mobility IRG in Singapore-MIT Alliance for Research and Technology (SMART) from 2016 to 2021, where he developed behavioral models and integrated demand and supply models into an agent-based urban mobility simulation platform, SimMobility. His expertise includes traffic simulation modeling, calibration and optimization, automated mobility-on-demand (AMOD) systems, and big data for predictive analytics.

In this seminar, Professor Hayeon Choi presented recent research on the relationship between daily summer temperature and violent crime across U.S. cities. Drawing on violent crime records from 44 cities in 33 states between 2005 and 2022, the study examined how short-term heat exposure is associated with changes in violent crime risk. Rather than assuming a simple linear relationship, the research applied non-linear generalized additive models to identify diverse temperature–crime response patterns across cities, including J-shaped, inverted J-shaped, and linear associations. The findings showed that violent crime risk generally increased on hotter days compared to moderate-temperature days, while the strength and shape of this relationship varied substantially by local climate, urban greenspace, air-conditioning prevalence, and temperature anomalies relative to historical baselines. The seminar highlighted how climate change may intensify heat-related social risks and emphasized the importance of considering local environmental and social conditions when developing climate adaptation and violence-prevention strategies.

Professor Hayeon Choi is an Assistant Professor at the KAIST Graduate School of Green Growth and Sustainability and is jointly affiliated with the Department of Civil and Environmental Engineering. Her research focuses on the health impacts of climate change and environmental exposures, with particular attention to climate resilience, environmental health, and environmental justice. She received her B.A. in Education and Statistics from Korea University, earned her master’s degree in Biostatistics and Epidemiology from Seoul National University, and completed her Ph.D. in Environmental Studies at Yale University. Prior to joining KAIST, she was a postdoctoral research fellow at the Harvard T.H. Chan School of Public Health, where she studied the associations between environmental hazards and child neurodevelopmental health. Her current work uses data-driven approaches to examine how climate and environmental factors, such as extreme heat, air pollution, and urban green space, affect vulnerable populations and community health, with the goal of informing strategies to reduce climate-related health inequalities.

AI and machine learning have drawn significant interests in recent years and have found many applications in engineering. The purpose of this presentation is to discuss the potential uses of machine learning models for a variety of applications.  The presentation covers a broad review of AI research and development in engineering and introduces selected examples to illustrate the broad applications of the AI from natural language interface, regulation management, machine diagnostic, and control problems to smart manufacturing and mobility technologies for the built environment. 

Kincho H. Law received his B.Sc. in Civil Engineering and B.A. in Mathematics from the University of Hawaii in 1976, and M.S. and Ph.D. in Civil Engineering from Carnegie Mellon University in 1979 and 1981, respectively.  After serving as Assistant Professor at Rensselaer Polytechnic Institute from 1982 to 1988, he joined Stanford University in 1988 and is currently Professor of Civil and Environmental Engineering.  Prof. Law’s professional and research interests focus on computational and information science in engineering.  His research has dealt with various aspects of high performance computing; sensing, monitoring and control of engineering systems; legal and engineering informatics; enterprise integration; smart manufacturing; web services and cyber-physical systems.  He has authored and co-authored about 500 articles in journals and conference proceedings.   

Prof. Law was the recipient of the ASCE Computing in Civil Engineering Award in 2011 and the Excellence in Research Award by ASME’s Division of Computers and Information in Engineering in 2023.  He has received a number of best paper awards from the American Society of Civil Engineers (ASCE), American Society of Mechanical Engineers (ASME), the Institute of Electrical and Electronics Engineers (IEEE), Digital Government Society and others. He has been on the advisory boards for a number of start-up companies on Data Analytics, IoT Platform for Manufacturing, Autonomous Vehicles, and others.  Prof. Law was elected Distinguished Member of ASCE in 2017, Fellow of ASME in 2017, Life Member of ASCE in 2018, and Senior Member of IEEE in 2019.

Fire and rescue services in the UK have faced challenges of downgraded services and delayed responses to fire incidents, and the service quality varies significantly within a region. Understanding the pattern and factors of fire service quality is essential for enhancing these services. In this presentation, I will discuss metrics and methods for understanding the barriers of place-based fire service quality, using geospatial machine learning techniques. Then, I will present the preliminary results from 15-year records of fire incidents from West Midlands Fire Services.

Huanfa Chen is an Associate Professor in Spatial Data Science at the UCL Centre for Advanced Spatial Analysis (CASA). He holds a PhD degree in GeoInformatics from UCL SpaceTimeLab, and BSc and MSc from Peking University. His research interests span spatial optimisation and GeoAI, with applications in public health, fire services, and transport research. He currently serves as associate editor of Annals of GIS and reviewers for multiple academic journals. He has published over 15 articles on international journals, including IJGIS and CEUS.

Access inequity extends beyond our neighborhoods, shaping the activity spaces we navigate daily such that we either come together or remain apart. Achieving equitable access to essential services and promoting active transportation are key goals of contemporary city and mobility planning. This seminar will introduce the latest breakthroughs in accessibility measurement developed at TU Delft’s Urban Analytics Lab. A cutting-edge software tool designed to assess access equity in X-minute neighborhoods will be presented, integrating measures of both individual and collective access to essential destinations and capturing dimensions of perceived accessibility. 

Dr. Achilleas Psyllidis is an Assistant Professor of Urban Mobility and the Director of the Urban Analytics Lab at TU Delft. His research interests include sustainable urban mobility, accessibility measurement, and spatial analytics. He specializes in how the design, structure, and perceived qualities of the urban environment affect the mobility and accessibility opportunities of different population groups. He has developed novel measurement methods and software tools for human mobility and accessibility analysis and planning. He leads several European and International research projects on urban mobility, city planning, environmental exposures, public health, and well-being. 

As layers of networks and digital information blanket urban space, new approaches to the study of the built environment are emerging. Combining data analytics, technology, and design, the mission of the Senseable City Laboratory—a research initiative at the Massachusetts Institute of Technology—is to anticipate these changes and study them from a critical point of view 

Fabio Duarte is the Associate Director of MIT Senseable City Lab, and principal research scientist leading research on the intersection of data analytics, technology, and design. His research has been published in Nature Water, Science Robotics, Plos One, and his most recent book argues that technology is powerful when it is playful (Urban Play, MIT Press).

Traditional methods to understand diverse changes of urban state such as traffic estimation, population change prediction, etc are usually done by analyzing spatial temporal changes of their corresponding physical attribute. However, they lack in terms of causal effect analysis and prediction accuracy. We propose a new noble approach where we analyze how urban dwellers exploit a target urban space - human flow analysis. This allows us to understand various spatial temporal changes in urban space in more accurate and explanable manner.

Dongman Lee is a provost and executive vice president at Korea Advanced Institute of Science and Technology, Daejeon, South Korea, and also with school of computing at KAIST. His research interests include smart space middleware, edge IoT virtualization, social media analysis, and trust management. He is a member of KISS and IEEE, and a Senior Member of ACM. 

In this talk, we investigate two fundamental questions related to safety and insurance in the age of automation. First, we touch upon the question of safety and liability under infrastructure-assisted automated driving. In such an environment, automakers provide vehicle automation technology while infrastructure service providers (ISSPs) provide smart infrastructure services. Additionally, customers can receive coverage for accidents from either of these actors but also from legacy auto insurers. We investigate the effect of market structure on safety and accident coverage and show that an integrated monopoly provides full coverage and fully accounts for accident costs when choosing safety levels. However, in the Nash setting, even though full coverage obtains, lack of coordination leads to partial internalization of accident costs by the automaker. Moreover, multiple equilibria might exist, some of them undesirable. We show that, both in the presence and absence of legacy insurance, an appropriate liability rule can induce optimal safety levels under the Nash setting. Our second question concerns itself with the role of legacy auto insurance in the age of infrastructure-assisted automated driving. Our analysis shows that the industry is not necessary for optimal coverage when the cost of accidents is known in advance and all possible accident scenarios are contractible. In fact, their presence can even harm safety, even though it ensures full coverage for accidents. However, when only insurance contracts with capped liability for automakers and ISSPs are available, a window of opportunity opens up for legacy insurers to enter the market while improving coverage.

Daniel Vignon’s research seeks to inform the design, regulation and operation of emerging mobility services and of smart infrastructure systems. Drawing from his background in both engineering and economics, he models and analyzes the interactions of these systems with different markets, studies their performance and their impact on social welfare, and designs policies to optimally and parsimoniously regulate them. He holds a BSc in Mechanical Engineering from MIT, as well as an MA in Economics and a PhD in Civil Engineering from the University of Michigan.

Carefully designed urban and architectural spaces can reduce stress, promote physical activities, reduce crime rates, prevent infection, and even save lives. I will introduce studies that highlight how the design of cities and buildings could improve the health and well-being of individuals. More specifically, our studies regarding the relationships between the design of cities and the health of older adults will be shared. Using GPS data of older adults in South Korea, we will illustrate the walking behaviors of older adults in relation to the design of cities and streets.

Lisa Lim, Ph.D. is a researcher, designer, and educator with her primary focus on improving the health and wellness of users through design. She has an academic and practical background in architectural design and majored in Evidence-based design for her PhD. She joined KAIST in 2021 and prior to joining KAIST, she was an assistant professor at Texas Tech University. She is interested in how spatial layouts can support individual experience and organizational outcomes and how designers can provide such environments to users.

Data analytics and data-driven processes have been used to make urban planning decisions and to improve related city service operations. With the proliferation of digital data, new opportunities are being availed to measure, understand and propose changes to the communities in which we live, work, and play. This has led to a host of new terms and disciplines – urban science, big data, smarter cities, urban informatics, civic analytics – that seeks to understand the intersection of digital technologies and the human environment. The most benefit of those disciplines is not only an in-depth understanding of urban phenomena but also predicting and preparing for future scenarios in cities composed of complex systems. There are immense opportunities with big data and analytic capacities to support responsive and effective urban systems, ultimately aiming at sustainable and livable cities through a problem-driven analytic approach. This presentation focuses on the introduction to the next phase of urban planning based on analytics and introduce a research project sample using different scale of urban data.

Boyeong Hong is a Researcher in the Civic Analytics Program at the NYU Marron Institute of Urban Management. Her research interests focus on how to apply urban informatics to real world problems in urban planning and operations. Boyeong’s current work deals with predictive city analytics using Big Data and Machine Learning techniques to deliver better city services allocation. Additionally, she is working on the human mobility project associated with the disaster management and the urban resilience planning. Boyeong is currently an affiliated adjunct faculty at Columbia University, the Graduate School of Architecture, Planning, and Preservation. Boyeong earned a Ph.D. in Civil and Urban Engineering, majoring in Urban Informatics from New York University, and she holds a M.S. in Applied Urban Science and Informatics from NYU Center for Urban Science and Progress (CUSP). She holds a B.Arch from Yonsei University and a Master of City Planning degree from Seoul National University.

North Korea has long been a black box with no official data for outsiders to assess its economic development. The lack of ground truth labels also makes it difficult to apply existing inference models with remote sensing data for the country’s economic measurement. To overcome these constraints, we develop a human-machine collaborative algorithm that leverages satellite imagery and lightweight human annotations in the machine-learning process. When applied to North Korean satellite images for the period from 2016 to 2019 to generate grid-level estimates of the country’s economic development, our human-machine collaborative algorithm outperforms machine-only learning approaches based on nightlight intensity or land cover classification. Using our measure as a proxy of economic development indicates that amid rising pressure from economic sanctions, the centrally planned economy has been directing more resources towards its capital and regions with highly publicized state-led development projects. Our model can be applied to other developing countries with insufficient data and provide reliable and inexpensive indicators on a granular level

Jihee Kim is an associate professor in the School of Business and Technology Management, College of Business at KAIST. She is an economist interested in how economic outcomes are distributed across individuals and regions. Her primary focus is on the study of income distribution, specifically at the top. She has explored how the creative destruction of entrepreneurs, tax policy, and CEO pays have contributed to increases in top income inequality. She also has expanded her research by applying machine learning algorithms to satellite images in collaboration with computer scientists, providing detailed economic insights into regions like North Korea. Jihee holds a Ph.D. in Management Science and Engineering, an M.A. in Economics, both from Stanford University, and a B.S. in Computer Science from KAIST.

Diversity of physical encounters in urban environments is a key feature of cities that foster economic productivity and social capital. How and where do we have the most and least diversity? How did the diversity of our social encounters change due to the pandemic? I will introduce our work that answers these questions using large-scale, privacy-enhanced mobility dataset of more than one million anonymized mobile phone users in Boston, Dallas, Los Angeles, and Seattle, across three years spanning before and during the pandemic.

Taka is currently a Postdoctoral Associate at the MIT Media Lab, working on the intersection of computational social science and urban science with Alex 'Sandy' Pentland and Esteban Moro. His research develops tools and models for analyzing large-scale human behavior data to better understand collective social dynamics during disruptions, and to improve the resilience of communities and cities to shocks (e.g., disasters, pandemics, and disruptive technology). He will join New York University Center for Urban Science and Progress (CUSP) as an Assistant Professor in January 2024.

Mobility service route design requires potential demand information to well accommodate travel demand within the service region. Transit planners and operators can access various data sources including household travel survey data and mobile device location logs. However, when implementing a mobility system with emerging technologies, estimating demand level becomes harder because of more uncertainties with user behaviors. Therefore, this study proposes an artificial intelligence-driven algorithm that combines sequential transit network design with optimal learning. An operator gradually expands its route system to avoid risks from inconsistency between designed routes and actual travel demand. At the same time, observed information is archived to update the knowledge that the operator currently uses. Three learning policies are compared within the algorithm: multi-armed bandit, knowledge gradient, and knowledge gradient with correlated beliefs. For validation, a new route system is designed on an artificial network based on public use microdata areas in New York City. Prior knowledge is reproduced from the regional household travel survey data. The results suggest that exploration considering correlations can achieve better performance compared to greedy choices in general. In future work, the problem may incorporate more complexities such as demand elasticity to travel time, no limitations to the number of transfers, and costs for expansion.

Joseph Chow is an Institute Associate Professor at the NYU Tandon School of Engineering’s Civil and Urban Engineering Department with affiliations at CUSP, Rudin Center for Transportation Policy & Management, and Sustainable Engineering Initiative. Chow is an NSF CAREER award recipient, a former Canada Research Chair, and the co-founding Deputy Director of the C2SMART(ER) University Transportation Center at NYU. He is a co-chair of the Subcommittee on Route Choice & Spatiotemporal Behavior at TRB and former TSL Cluster Chair and elected Urban Transportation SIG Chair at INFORMS. He has published almost 90 journal articles since 2010 and is an editor for three transportation journals including Transportation Research Part B. Dr. Chow received his PhD ('10) at UC Irvine and his MEng (’01) and BS (’00) at Cornell University.

The traffic forecasting problem is a challenging task that requires spatial-temporal modeling and gathers research interests from various domains. In recent years, spatial-temporal deep learning models have improved the accuracy and scale of traffic forecasting. While hundreds of models have been suggested, they share similar modules, or building blocks, which can be categorized into three temporal feature extraction methods of recurrent neural networks, convolution, and self-attention and two spatial feature extraction methods of convolutional graph neural networks (GNN) and attentional GNN. More importantly, the models have been mostly evaluated for their entire architectures with limited efforts to characterize and understand the performance of each category of building blocks. In this study, we design an extensive, multi-faceted experiment to relate the choice of building blocks on traffic forecasting accuracy considering environmental characteristics and distributions of datasets including outliers. Specifically, we implement six traffic forecasting models using three building blocks for temporal modeling and two for spatial modeling. When we evaluate the models on four datasets with diverse characteristics, the results show each building block demonstrates distinguishable characteristics depending on study sites, prediction horizons, and traffic categories. The results of this study can enhance the utility of existing models and suggest guidelines for researchers building traffic forecasting model architectures and for practitioners implementing these state-of-the-art techniques in real-world applications.

Yuyol Shin is currently a postdoctoral researcher in the Department of Civil and Environmental Engineering at Korea Advanced Institute of Science and Technology (KAIST). He received B.S. (2016) and Ph.D. (2022) in Civil and Environmental Engineering at KAIST, and worked as a visiting scholar in Department of Civil and Environmental Engineering at University of California, Berkeley from October 2022 to June 2023. His research interests are spatial-temporal data mining, graph neural networks, artificial intelligence in the field of transportation engineering, and transportation network analysis. His recent works have focuses on core technologies of intelligent transportation systems such as traffic forecasting in urban areas using Graph Neural Networks and time-series models such as causal convolution and self-attention. Recently, Yuyol Shin is investigating the field of AI-based maritime transportation including topics such as vessel trajectory prediction, weather routing, and smart port operation.

This talk will present recent advancements in computational methods and large-scale, high-resolution urban data to address issues of social justice, public health, and climate action in cities. I will demonstrate how data-driven approaches can be applied to understand urban dynamics, support evidence-based policy and planning decisions, and empower residents through the democratization of data. Specific attention will be given to algorithmic bias and algorithm-in-the-loop decision-making. 

Constantine E. Kontokosta, PhD, is an Associate Professor of Urban Science and Planning and Director of the Civic Analytics Program at the Marron Institute of Urban Management at New York University (NYU). He also directs the Urban Intelligence Lab and holds faculty appointments at the NYU Center for Urban Science and Progress (CUSP) and the NYU Tandon School of Engineering. He previously served as the founding Deputy Director/Academic Director of CUSP. His work has been published in leading peer-reviewed journals, including PNAS, Nature Communications, and Nature Energy, and he is the recipient of research awards and grants from the National Science Foundation, IBM, Amazon, and the MacArthur Foundation, among others, and best paper awards from the Journal of the American Planning Association, ICLR, and the Bloomberg Data for Good conference. He holds degrees from the University of Pennsylvania, New York University, and Columbia University. 

Flooding is one of the most dangerous and costly natural hazards, and has a large impact on infrastructure, mobility, public health, and safety. Despite the disruptive impacts of flooding and predictions of increased flooding due to climate change, municipalities have little quantitative data available on the occurrence, frequency, or extent of urban floods. To address this, FloodNet has been designing, building, and deploying low-cost, ultrasonic sensors to systematically collect data on the presence, depth, and duration of street-level floods in New York City (NYC). FloodNet is a partnership between academic researchers and NYC municipal agencies, working in consultation with residents and community organizations. FloodNet sensors are designed to be compact, rugged, low-cost, and deployed in a manner that is independent of existing urban power and network infrastructure. These requirements were implemented to allow deployment of a hyperlocal, city-wide sensor network, given that urban floods often occur in a distributed manner due to local variations in land development, population density, sewer design, and topology. Thus far, 70 FloodNet sensors have been installed across the five boroughs of NYC. These sensors have recorded flood events caused by high tides, stormwater runoff, storm surge, and extreme precipitation events, illustrating the feasibility of collecting data that can be used by multiple stakeholders for flood resiliency planning and emergency response.

Charlie Mydlarz is a Research Associate Professor at NYU CUSP and the Music and Audio Research Laboratory. He is an acoustician/engineer who designs, develops, and deploys IoT devices to tackle different challenges, including: urban noise sensing, acoustic condition monitoring, urban flood detection, soundscape perception, building/classroom efficiency, and urban mobility. His PhD research at The University of Salford's Acoustic Research Centre enabled public engagement in a large-scale mass participation soundscape study using smart phones for global subjective and objective data collection and analysis.