Ongoing
Projects
The projects focuses on artificial intelligence in urban environments, emphasizing the latest innovations in GeoAI, OpenAI, location intelligence, and big data for urban planning. We are now conducting research in collaboration with Harvard University's Spatial Data Lab and several globally renowned universities in the USA, Japan, and Europe.
AI and Compact Cities
This project investigates how artificial intelligence can support the planning, assessment, and management of compact, accessible, inclusive, and environmentally sustainable cities. It combines GeoAI, machine learning, deep learning, GIS, mobility data, population information, land-use patterns, and urban-service indicators to measure compactness and livability at city, neighborhood, and transit-station levels. The research examines access to public transport, employment, healthcare, education, recreation, green spaces, retail facilities, and other essential services within walkable distances.
AI models are developed to identify spatial inequalities, classify urban areas, predict compactness scores, and compare present conditions with future development scenarios. The project also explores population-density projections and changing service requirements to understand how urban neighborhoods may perform over time. Interactive dashboards and decision-support applications translate model outputs into accessible information for planners, policymakers, researchers, and communities.
The central objective is to develop an evidence-based planning framework that can recommend targeted interventions for different urban contexts. These may include mixed-use development, improved public transport, additional social infrastructure, pedestrian improvements, and better distribution of public facilities. Ultimately, the project seeks to promote efficient land use, shorter travel distances, lower emissions, improved accessibility, and a higher quality of urban life.
Quantum AI and Cities
This exploratory project investigates the future potential of quantum machine learning for analyzing, modeling, and optimizing complex urban systems. Cities generate large and highly interconnected datasets involving transportation, land use, buildings, population, infrastructure, energy, environment, public services, economic activity, and human behavior. Processing these relationships can become computationally demanding, particularly when planners must compare many possible development scenarios.
The research examines quantum and hybrid quantum-classical algorithms for urban classification, spatial optimization, pattern recognition, prediction, clustering, and scenario evaluation. Possible applications include transport routing, traffic management, facility location, land-use allocation, energy-network optimization, emergency-response planning, and the analysis of high-dimensional geospatial datasets. Quantum optimization techniques may also help evaluate large numbers of planning alternatives involving competing social, economic, spatial, and environmental objectives.
As quantum computing remains an emerging field, the project does not assume that quantum models will automatically outperform established machine-learning approaches. Instead, conventional, hybrid, and quantum models are systematically compared using selected urban-planning problems. Their accuracy, computational efficiency, scalability, interpretability, and hardware limitations are assessed. The research aims to identify realistic areas where quantum methods may eventually provide an advantage while avoiding unsupported claims. Ultimately, the project establishes a foundation for responsible experimentation with quantum computing in urban research and long-term planning.
"Image of the City" and AI
This project reinterprets Kevin Lynch’s influential concept of the “Image of the City” through artificial intelligence, computer vision, natural language processing, and geospatial analysis. Lynch explained how people understand and remember cities through five elements: paths, edges, districts, nodes, and landmarks. The project develops computational methods for identifying, mapping, and evaluating these elements across different urban environments.
Street-view photographs, satellite images, social-media content, volunteered geographic information, mobility patterns, and public surveys are analysed to understand how urban places are visually perceived and mentally organised. Computer-vision models can detect physical features such as building façades, greenery, street furniture, signs, boundaries, landmarks, and public spaces. Natural language processing can examine how residents and visitors describe particular locations, while spatial analysis connects these perceptions to their geographic context.
The research evaluates urban legibility, visual character, identity, safety, memorability, orientation, and sense of place. It also considers how perceptions may differ among age groups, genders, communities, visitors, and people with different mobility needs. By combining physical evidence with human experience, the project seeks to develop measurable indicators of urban imageability. The resulting tools can support wayfinding strategies, public-space improvement, heritage conservation, place branding, inclusive design, and the creation of distinctive, understandable, and human-centred cities.
AI and Volumetric Urbanism
This project applies Bayesian statistical modelling to understand cities as complex three-dimensional systems shaped by building height, built volume, floor-area distribution, population density, land use, infrastructure, open space, and environmental conditions. Conventional urban models often treat these variables as fixed or two-dimensional. In contrast, Bayesian models incorporate existing planning knowledge, observed spatial data, and uncertainty within a single analytical framework.
The research examines how vertical and horizontal development patterns influence accessibility, infrastructure demand, daylight availability, ventilation, energy consumption, public-space quality, and overall urban performance. Prior knowledge derived from planning standards or previous case studies can be combined with new evidence collected from GIS databases, remote sensing, building models, and field observations. Model estimates can then be updated continuously when additional information becomes available.
Different volumetric development scenarios are tested to understand the likely consequences of increasing building height, changing floor-area ratios, modifying land uses, or redistributing density around transport nodes. Instead of producing a single deterministic result, the model estimates a range of possible outcomes and indicates the level of confidence associated with each prediction. The project ultimately aims to provide planners with a transparent decision-support system for evaluating development alternatives, identifying risks, and managing uncertainty in high-density urban environments.