New RESTOR Paper Maps the Future of AI-Driven Design Automation in the Built Environment

The RESTOR project team has published a new review paper in the Journal of Building Engineering exploring how building and structural design automation is evolving from traditional generative design towards the integration of artificial intelligence, deep generative models, and topology optimisation.

The paper, “Trajectory of building and structural design automation from generative design towards the integration of deep generative models and optimization: A review,” was authored by Soheila Kookalani, Erika Parn, Ioannis Brilakis, Samir Dirar, Marios Theofanous, Asaad Faramarzi, Mohammad Ali Mahdavipour, and Qixian Feng. Supported by industrial partners Chetwoods Architects and Cleveland Steel.

The review addresses a central challenge for the built environment: design processes remain highly manual, time-consuming, and difficult to optimise across competing goals such as cost, structural performance, material efficiency, and sustainability. The paper explains that design automation can help designers rapidly generate and assess multiple design options, while topology optimisation can reduce unnecessary material use by placing material only where it is structurally needed.

For Project RESTOR, this work is important because the reuse of structural steel requires more than material testing and recovery. It also requires smarter design workflows that can match available reclaimed components with new design requirements. Generative design, AI, and optimisation methods provide a pathway for making reused steel easier to specify, compare, and integrate into real projects.

Key findings from the paper

1. Design automation can transform slow, manual design workflows.The paper shows that generative design allows computers to automatically explore many design alternatives based on defined objectives and constraints. This can help architects and engineers move beyond one-at-a-time manual design iterations and instead assess a much wider design space.

2. Generative design supports more sustainable construction.By optimising for material use, energy performance, structural efficiency, and cost, generative design can support lower-impact design decisions. This is particularly relevant for RESTOR, where the aim is to make structural steel reuse more practical and environmentally beneficial.

3. AI is expanding what generative design can do.The review highlights the growing role of deep generative models, including generative adversarial networks, variational autoencoders, and reinforcement learning. These methods can learn from existing designs, generate new options, and support faster, more intelligent design exploration.

4. Topology optimisation remains a key method for material efficiency.Topology optimisation helps identify where material is actually needed in a structure. The paper shows that when topology optimisation is combined with generative design and AI, it can support more efficient, innovative, and structurally informed design solutions.

5. Hybrid methods are the future.A major finding is that no single method solves every design problem. The paper points towards hybrid approaches that combine generative design, deep learning, and optimisation. These methods could help designers produce solutions that are not only structurally sound, but also more resource-efficient and easier to evaluate.

6. Practical adoption still faces barriers.The paper identifies several challenges that need to be addressed before these technologies can be widely adopted in practice. These include computational complexity, data availability, interpretability of AI-generated designs, integration with BIM workflows, and the need for user-friendly tools that designers and engineers can trust.

Why this matters for RESTOR

RESTOR is focused on enabling the reuse of structural steel in construction. One of the biggest barriers to reuse is that reclaimed steel does not arrive as a clean, unlimited catalogue of standard new products. Instead, designers must work with available sections, variable dimensions, uncertain stock, and project-specific constraints.

This is where design automation becomes critical. AI-driven generative design and optimisation can help match reclaimed steel members to new structural designs, explore alternative layouts, minimise waste, and support more informed decisions at early design stages.

The paper therefore provides an important methodological foundation for RESTOR’s wider ambition: to move structural reuse from a manual, case-by-case process towards a more intelligent, data-driven, and scalable design workflow.

Ultimately, the study shows that the future of sustainable construction will depend not only on reusing materials, but also on redesigning the way we design. By combining generative design, deep learning, topology optimisation, and BIM-based workflows, the built environment sector can move towards faster, smarter, and more circular forms of structural design.

Published paper URL: https://doi.org/10.1016/j.jobe.2024.110972

ScienceDirect: https://www.sciencedirect.com/science/article/pii/S2352710224025403