An AI-driven framework for rapid and localized optimizations of urban open spaces

AI can make urban planning faster and more practical: The framework optimized urban spaces in 1 minute compared to 15-30 minutes for traditional methods, making it feasible for cities to improve multiple neighborhoods rather than just a few flagship projects.

Explainability matters for public trust: By using SHAP and Counterfactual Explanations, the framework shows why specific design changes are recommended (e.g., "adjust building width by X meters"), allowing city planners and residents to understand and debate AI suggestions rather than accepting a "black box" recommendation.

Smaller, targeted changes can be as effective as major redesigns: The AI identified minimal design modifications needed to improve thermal comfort and safety, suggesting cities don't always need expensive overhauls—incremental improvements to existing spaces may deliver comparable benefits.

The framework prioritizes computational efficiency for scalability: By requiring significantly less computing power than global optimization methods, this approach could be deployed across many cities and neighborhoods simultaneously, democratizing access to AI-driven urban planning tools beyond wealthy municipalities.

Specific design factors drive real outcomes: The research identified that building width, park area, and proximity to surrounding buildings are the key factors influencing both thermal comfort and perceived safety, giving planners concrete, data-backed priorities for where to focus improvement efforts.