Overview

Many rapidly urbanizing regions are shifting from outward expansion to quality-oriented renewal within existing built-up areas. In practice, the most challenging spaces are often those with high building height, high density, and high development intensity—areas that can deliver strong economic and social benefits, yet may also accumulate risks related to comfort, safety, environmental stress, and uneven performance. This project develops a scalable, data-driven technical pathway to identify, diagnose, and improve such high-intensity built environments across diverse city contexts, while keeping the workflow efficient, explainable, and suitable for planning and design decision support.

Why this matters

Cities increasingly shift from outward expansion to quality-oriented renewal within existing built-up areas. High-intensity districts can deliver strong economic and social value, but may also concentrate risks such as environmental stress, reduced comfort, safety concerns, and uneven performance. Conventional indicator-based controls and heavy simulations are often too coarse or too slow for iterative renewal. This project bridges the gap with an interpretable, scalable workflow that supports fast screening, prioritisation, and design feedback.

Key research questions

• How can high-intensity built-up patterns be consistently identified across cities and spatial scales?
• What is the minimal, transferable set of indicators that preserves recognition accuracy and interpretability?
• How can we map multi-dimensional urban performance and prioritise interventions under trade-offs?
• How can we translate data insights into planning-ready guidance ranges and rapid design iteration tools?

Technical approach

1. Multi-source data integration: fuse built form, street networks, land cover, activity proxies, and remote-sensing signals into comparable spatial units.
2. Multi-scale typology discovery: compute morphological indicators at multiple resolutions; use clustering to reveal recurring high-intensity types; match types across scales for robustness.
3. Explainable feature distillation: apply interpretable ML to learn a minimal indicator set and produce reusable, rule-like screening logic.
4. Performance atlas & prioritisation: build multi-dimensional performance layers (e.g., environmental/social/economic proxies); use multi-objective reasoning to identify low-performing yet high-potential areas.
5. Actionable guidance ranges: train surrogate models to link form indicators to performance outcomes; extract reasonable and target ranges from nonlinear response patterns.
6. Generative rapid simulation (optional): learn conditional image-to-image mappings between built-form layers and performance heatmaps to enable near-instant scenario comparison and early-stage ideation.

Keywords

high-intensity built-up areas; urban morphology; multi-scale analysis; explainable machine learning; multi-objective optimisation; performance atlas; generative AI; planning decision support