AI for Architecture: Proceedings of the 5th International Conference on Computational Design (CCD 2024)
Jiayan Fu
About this book
This open access book series includes compilations of selected papers from the International Conference on Computational Design and Robotic Fabrication. The books focus on novel techniques for computational design and robotic fabrication. It not only aims at the most recent research results from the key scholars in the computational design and robotic fabrication, but also offers an in-depth examination of intelligence in design and construction industry, referring to the invention and application of machine intelligence in architecture.The proceedings in this series are related with Sustainable Development Goals 11 (Sustainable Cities & Communities) and 9 (Industry Innovation & Infrastructure). The contents make valuable contributions to academic researchers, designers, and engineers in the industry. As well, readers will encounter new ideas about understanding intelligence in architecture.
Questions & Answers from this book
Questions and answers are connected to the referenced book and its available source material.
Chapter 1: Computational Design and Intelligent Construction
How do House-GAN++ and the method by Zheng and Petzold differ in their approach to generating floor plan layouts?
House-GAN++ directly represents a floor plan as a graph, where rooms are nodes and their relationships are edges, and it generates the layout conditioned on room count, types, and areas. Zheng and Petzold differ by separating the prediction of room topology from the prediction of room geometry, using subgraph neural networks for topology and neural-guided plan sketching for geometry. In short, House-GAN++ uses a unified graph-based generative process, while Zheng and Petzold decouple topological reasoning from geometric layout generation.
What is the role of graph neural networks (GNN) in the cross-modal spatial layout generation method proposed in this chapter?
In the proposed method, the GNN acts as the generative model that receives feature embeddings created by AI agents from text, graph, and boundary inputs, and outputs node and edge features that are converted into a detailed 3D spatial layout. Nodes represent spatial grids and edges represent wall categories, so the GNN learns to infer layout details such as wall types and spatial divisions from the fused cross-modal inputs.
How does the RoomGNN approach convert bubble diagrams and building boundaries into a format suitable for AI generative models?
RoomGNN first uses AI agents to convert bubble diagrams and building boundaries into structured text. Those structured data are then turned into room feature embeddings by mapping room coordinates inside the boundary, applying a Signed Distance Function to color potential diffusion areas, and overlaying a grid to compute influence probabilities. These embeddings become node features for the GNN, whose edges represent wall categories, enabling the model to generate spatial layouts.
Chapter 2: Computational Urban Science
What are the limitations of location-based analysis methods (WiFi, mobile signaling, GPS) for studying human behavior, and how do computer vision methods address these limitations?
Location-based methods like WiFi, mobile signaling, and GPS can track spatiotemporal trajectories at large scales with minimal interference, but they only capture location information. They cannot provide deeper insights into micro-level behavioral details, such as gait or body posture, leaving fine-grained behavioral studies difficult to quantify. Computer vision methods address this by adding object detection, tracking, and pose estimation to extract detailed key-point data and enable automated, visual, real-time analysis of micro-behaviors.
How does the pix2pixHD model improve upon the traditional conditional GAN loss, and what is the total loss function used in this study?
The pix2pixHD model improves the traditional conditional GAN loss by adding feature matching loss and perceptual (VGG) loss to the standard GAN loss. The total loss used in the study is the sum over multi-scale discriminators of the GAN loss plus weighted sums of feature matching and VGG losses, with both weights set to 10.
What are the limitations of the proposed spatial combination prediction method for community-embedded retirement buildings as discussed in the conclusions?
The conclusions identify two main limitations: the dataset was selected and produced based on the researcher's experience due to resource and time constraints, introducing bias and contingency, and despite data enhancement its size and diversity are lacking, affecting model generalization and accuracy. In addition, the method is still a pre-design spatial layout approach that has not been developed into an end-user application for designers or community managers, so further optimization and experimentation are needed for practical use.
Chapter 3: Computational Urban Design
According to the study by Jiang and Zou, which morphological indicators were analyzed for their correlation with building energy consumption?
The study analyzed twelve morphological indicators; the named ones include building floor area ratio, building density, average height of individual buildings, and weighted average height of building area. Other indicators that appear in the reported analyses include unit area, average building perimeter, standard deviation of building height, average distance between buildings, average shortest distance of buildings, and area.
What are the four key dimensions proposed in the chapter for the digital analysis of historical and cultural districts from a spatiotemporal vitality perspective?
The four key dimensions are functional diversity, supply–demand coupling, accessibility, and spatial permeability.
What is the formula for calculating the technical solar potential of an office block, and what do the variables H_A, A_pv, η_sys, K, R_1, and R_d represent?
The technical solar potential of an office block is calculated as E_p = H_A × A_pv × η_sys × K × Σ_{i=1}^{N} [1 − R_1 − (N − 1) × R_d]. H_A is the cumulative solar radiation received by the building surface, A_pv is the mountable area of the PV modules, η_sys is the photovoltaic conversion rate of the PV modules (taken as 21.86%), K is the combined efficiency factor (taken as 85%), and R_1 and R_d are the attenuation rates of the PV module in the first year and in subsequent years (1.5% and 0.48%, respectively).
Chapter 4: Computational Landscape
According to the study, what is the relationship between wind speed, solar radiation, and PET in courtyard thermal comfort?
Wind speed is negatively correlated with PET, so higher wind speed lowers PET and improves thermal comfort, while solar radiation is positively correlated with PET, raising PET and reducing thermal comfort. Solar radiation generally had a more significant overall impact on PET than wind speed, although wind speed showed greater variability in its effect.
What are the four passive renovation strategies proposed for optimizing thermal comfort in Lingnan school courtyards, and how are they evaluated?
The four passive renovation strategies are opening, shading, inner facades, and underlying surfaces. They are evaluated by simulating each strategy on a Lingnan school courtyard prototype with ENVI-met and using the mean Physiologically Equivalent Temperature (mPET) of the courtyard observation area as the thermal comfort indicator. The simulations are run for the summer solstice period, and the results are compared qualitatively and quantitatively to identify the most effective renovation approach.
What are the recommended ranges for opening size, shading density, and grass coverage to optimize thermal comfort in courtyards?
The recommended ranges are: opening size from 1/4 to 3/4 of the courtyard's one-side length; west-facing vertical shading density from 0.25 to 0.50; horizontal shading covering 1/4 to 1/2 of the courtyard area; and grass coverage from 1/4 to 1/2 of the courtyard area.
Chapter 5: Computational Media and Art
How does the study use semantic analysis of historical texts to reconstruct the urban space of Luoyang in the Northern Wei Dynasty?
The study applies semantic analysis to historical texts about Northern Wei Luoyang using HanLP for part-of-speech tagging and syntactic parsing. This produces semantic trees and abstract meaning representations that identify buildings and spatial relations such as which temple lies north or south of another. By cross-referencing texts like A Record of Buddhist Monasteries in Luoyang, the researchers verified the relative positions of 48 temples to map the road network, water system, and temple distribution. Semantic analysis at the block scale then used quantitative architectural descriptions to infer building sizes, typologies, courtyard layouts, and street layouts.
What is the 'shifting lens effect' in axonometric drawings and how is it achieved?
The shifting lens effect is the visual effect in axonometric drawings where moving the viewpoint or frame across the scene produces a sense of camera movement while all lines remain undistorted, because axonometric drawings have no perspective distortion. It is achieved by translating the frame over the axonometric image, and it can be enhanced by moving graphical blocks or other elements within the scene to advance the animated plot.
You may also be interested in
FastAPI: Modern Python Web Development
Bill Lubanovic;
37 questions
Generative Deep Learning: Teaching Machines to Paint, Write, Compose, and Play
David Foster;
33 questions
AI for the Ordinary_ A Non-technical Playbook for Citizens, Students, and Manage
Unknown
30 questions
AI for Qualitative Research: A Hands-On Guide for Management Scholars
Diana Garcia Quevedo
29 questions