ANN x Generative Design

research | a generative tower study in London exploring how evolutionary form finding and machine learning (ML) can connect programme, environmental performance and structural behaviour. developed through personal research and the Emergent Technologies and Design (EmTech) programme at the Architectural Association (AA), the study uses artificial neural networks (ANN) to predict spatial functions.

Street-level visualisation of a planted multi-tower urban development. — 2024

the project develops a multi-conventional tower complex through a generative workflow in which form, environmental performance, circulation and structural behaviour are evaluated across multiple design options. evolutionary processes are used to generate and compare tower morphologies.

Evolutionary tower design process showing alternative geometries and performance comparisons.

a second workflow integrates machine learning into spatial programming. artificial neural networks are trained on spatial relationships and then used to predict program distribution, allowing generated layouts to be evaluated and refined as part of the design process.

Generated tower forms compared through environmental and structural evaluations.

Neural-network workflow linking training data with predictions of spatial functions.

Colour-coded tower diagrams showing the distribution of different programmes.

the final study brings form generation and program prediction together: commercial, office, residential, hotel and public uses are distributed across the tower system while geometry continues to respond to environmental, structural and circulation objectives.