Image generated by Flux Dev 

starting point: 

How can computational methods, machine learning, and artificial intelligence be used to enhance sustainability outcomes and optimise value creation in the Architecture, Engineering, and Construction (AEC) sector?

This topic may include consideration of: computational methods, machine learning, and artificial intelligence insights to help the AEC sector overcome challenges that consider key decision factors around resource use, carbon emissions, waste production, project timelines, project budgets, and project benefits; how manufacturing information databases and systems can be leveraged to inform decision-making across key aspects of sustainable built assets. 

project summary: 

Wayfinding in the built environment is not only concerned with users finding their targeted destinations efficiently, but it also saves lives in critical cases of emergency evacuation. Although studies show that layout geometry directly affects ease of user navigation in space, it is rarely addressed in the early planning stages. Existing computational approaches for evaluating wayfinding in spatial layouts require developed floor plans and only give an insight into their performance, but no guidance on what to change. 

AI approaches for space planning provide an opportunity for providing rapid feedback and, furthermore refining floor plans in development stages according to evaluation results. Most used AI generative approaches in floor planning automation are generative adversarial networks (GANs), graph convolutional networks (GCNs) and diffusion models. Several AI applications addressed planning design requirements such as adjacency, daylighting and accessibility with success.  Diffusion models in particular show promise in terms of controllability and flexibility. 

This research aims to identify appropriate approaches for controlling diffusion models through dataset curation and preparation, and input data type to generate floor plans of large complex buildings. Furtherly, evaluating generated floor plans for wayfinding requirements both quantitatively and qualitatively and refining them for optimal user navigation experience.