Architecture

Architecture is a multifaceted discipline that shapes the built environment to meet aesthetic, functional, environmental and social objectives. At its core lie processes of conceptualisation, design development and performance assessment, which draw on structural engineering, environmental science, digital technologies and humanistic inquiry. Computational tools—parametric modelling, generative algorithms and data-driven simulation—now accelerate the exploration of form, daylighting, energy performance and spatial organisation. Building Information Modelling (BIM) provides a centralised digital twin of geometry, materials, systems and lifecycle metadata, enabling stakeholders to coordinate design, construction and operation. Concurrently, participatory and evidence-based methods integrate user needs, cultural context and post-occupancy evaluation to inform inclusive, healthy and resilient solutions. Sustainable strategies span passive orientation, high-performance façades and renewable systems, paired with circular-economy material choices and green infrastructure at urban scale to mitigate climate change and foster biodiversity. Heritage conservation employs non-invasive diagnostics and computational reconstruction to preserve authenticity while adapting historic fabric for contemporary use. Across scales—from individual rooms to whole cities—architecture research advances by bridging theory, practice and technological innovation, with global significance in decarbonisation, community cohesion and ecological regeneration.

Research from Nature Portfolio

Recent studies have applied active learning to automate floorplan element detection for rapid energy assessment. A model trained on a small annotated set iteratively selects the most uncertain unlabelled plans for human review, achieving precision and recall above 0.95 in detecting walls, windows and doors and drastically reducing manual labelling burden. Another investigation employs text-conditioned diffusion models to generate interior design schemes, producing visually coherent layouts that satisfy functional criteria such as circulation and spatial zoning. Designers can refine style or programme through simple prompt adjustments, speeding early-stage ideation. In the context of building stock decarbonisation, a life-cycle assessment coupled with cost analysis identifies robust retrofit strategies for Swiss residential typologies. By incorporating bio-based insulation and electrified heating, the study balances embodied and operational impacts under uncertain future climates, offering a transferable framework for low-carbon renovation across temperate regions.

Research from all publishers

A deep reinforcement learning framework recasts space-layout design as a sequential decision problem in a cell-based simulation. Agents trained via Proximal Policy Optimisation explore insertion and adjustment moves within a CAD environment to maximise rewards for adjacency, circulation efficiency and geometric regularity, outperforming genetic algorithms in both quality and convergence speed. Elsewhere, conditional generative adversarial networks trained on a dataset of energy-efficient residential floor plans generate new schemes that reduce annual energy consumption by double-digit percentages while preserving logical programme adjacencies. A complementary materials-focused study develops ambient-cured geopolymer mortars from mixed construction and demolition wastes, yielding 28-day compressive strengths above 30 MPa and up to 60 % lower embodied CO₂ compared with Portland cement mortars. The research demonstrates scalable routes to greener masonry elements via waste valorisation and mixture optimisation for durability and hygrothermal performance.

Architecture publication trend

The graph below shows the total number of articles in architecture across all publications each year (not limited to Nature Index journals).

Technical terms

Building Information Modelling (BIM): A coherent digital representation of a facility’s physical and functional properties, supporting design, construction and lifecycle management.

Generative Adversarial Network (GAN): An adversarial pair of neural networks (generator and discriminator) trained to produce realistic synthetic data, such as floor-plan layouts.

Diffusion model: A probabilistic generative approach that iteratively denoises random noise according to a learned process to synthesise images or layouts.

Active learning: A machine-learning strategy in which a model selects the most informative unlabelled examples for annotation, reducing labelling requirements.

Deep reinforcement learning: A paradigm where an agent interacts with an environment to learn sequential actions that maximise cumulative rewards, applied to tasks like space planning.

Life cycle assessment (LCA): A method to quantify environmental impacts associated with all life stages of a product or building, from material extraction through end-of-life.

References

1. The power of progressive active learning in floorplan images for energy assessment. Scientific Reports (2023).
2. Integrating aesthetics and efficiency: AI-driven diffusion models for visually pleasing interior design generation. Scientific Reports (2024).
3. Strategies for robust renovation of residential buildings in Switzerland. Nature Communications (2024).
4. Reimagining space layout design through deep reinforcement learning. Journal of Computational Design and Engineering (2024).
5. A Deep Learning Approach toward Energy-Effective Residential Building Floor Plan Generation. Sustainability (2022).
6. A comprehensive study on the compressive strength, durability-related parameters and microstructure of geopolymer mortars based on mixed construction and demolition waste. Journal of Cleaner Production (2023).