AI Architecture Research: Floor Plans, BIM & Building Performance
Read 116 author-sourced research records through specific architectural questions, dates and evaluation limits.
These records summarize author-reported research, including preprints and supporting datasets. They do not establish available product features or independently reproduced performance. Publication and revision dates are distinct.
116 records · Page 5 / 5
Artificial Intelligence-Assisted Energy and Thermal Comfort Control for Sustainable Buildings: An Extended Representation of the Systematic Review
RESEARCH /
This extended review considers AI-assisted energy and comfort control within a wider indoor-environment context. It reminds readers that temperature is only one part of acceptability. Architectural interpretation should ask which conditions each control study measures and which occupant needs remain outside its objective.
Architectural question
Which indoor-environment needs remain outside a comfort-control objective?
Ghezlane Halhoul Merabet, Mohamed Essaaidi, Mohamed Ben-Haddou, Basheer Qolomany, Junaid Qadir, Muhammad Anan, Ala Al-Fuqaha, Riduan Mohamed Abid, Driss Benhaddou
Transfer Learning for Thermal Comfort Prediction in Multiple Cities
RESEARCH /
This research explores transferring thermal-comfort models when local occupant feedback is limited. It addresses a data problem rather than guaranteeing universal comfort predictions. Building teams should examine climate, population and operating differences, and preserve a local feedback sample for testing the transferred model.
Architectural question
What local evidence is needed before transferring a comfort model?
Nan Gao, Wei Shao, Mohammad Saiedur Rahaman, Jun Zhai, Klaus David, Flora D. Salim
Comfort-as-a-Service: Designing a User-Oriented Thermal Comfort Artifact for Office Buildings
RESEARCH /
The research proposes a user-oriented approach to individual comfort in offices, emphasizing feedback rather than energy costs alone. It is useful for examining how comfort services fit existing infrastructure. Readers should distinguish the proposed artifact and its study setting from an independently verified, generally available workplace service.
Architectural question
How can office comfort feedback fit existing systems?
Building Footprint Generation by IntegratingConvolution Neural Network with Feature PairwiseConditional Random Field (FPCRF)
RESEARCH /
The proposed pipeline couples image features from a convolutional network with a graph model that represents spatial relationships. Its focus is retaining building edges in footprint maps. Architectural use would require checking narrow gaps, attached buildings and the original image resolution.
Architectural question
Can adjacent buildings remain distinct in a footprint map?
MAP-Net: Multi Attending Path Neural Network for Building Footprint Extraction from Remote Sensed Imagery
RESEARCH /
MAP-Net addresses the loss of small buildings and unstable edges on larger footprints in convolutional extraction. The study is useful for questioning scale sensitivity in urban maps. A comparison should include fine-grained blocks and large buildings rather than reporting only one overall accuracy number.
Architectural question
What does a footprint model miss at the smallest and largest scales?
Qing Zhu, Cheng Liao, Han Hu, Xiaoming Mei, Haifeng Li
Building Information Modeling and Classification by Visual Learning At A City Scale
RESEARCH /
The paper presents visual-learning approaches for extracting building information from satellite and street imagery and examining spatial patterns. Its architectural relevance is assembling urban inventories. It should be read as inferred attributes and statistical analysis, with uncertainty retained for properties that imagery cannot directly establish.
Architectural question
Which building attributes can imagery support reliably?
Qian Yu, Chaofeng Wang, Barbaros Cetiner, Stella X. Yu, Frank Mckenna, Ertugrul Taciroglu, Kincho H. Law
Deep Floor Plan Recognition Using a Multi-Task Network with Room-Boundary-Guided Attention
RESEARCH /
This paper separates boundary recognition from room-type recognition and connects them through attention. It aims to extract more than walls alone, including doors, windows and room functions. Design-document interpretation should therefore check object identity and adjacency together rather than accepting a segmented image uncritically.
Architectural question
Can room-boundary evidence improve room-type interpretation?
Zhiliang Zeng, Xianzhi Li, Ying Kin Yu, Chi-Wing Fu
Uncovering Dominant Social Class in Neighborhoods through Building Footprints: A Case Study of Residential Zones in Massachusetts using Computer Vision
RESEARCH /
The authors study neighborhood-level associations between building morphology and income using figure-ground maps. The architectural relevance is urban-form analysis, with substantial contextual limits. Aggregate correlations should not be used to label individual residents or treated as universal explanations of social conditions.
Architectural question
Which urban-form associations remain meaningful after local context is considered?
A Review of Reinforcement Learning for Autonomous Building Energy Management
RESEARCH /
The review surveys reinforcement learning in autonomous energy management rather than presenting a single ready-to-install controller. It helps frame how sensors, controls and energy objectives interact. Architectural teams can use it to question comfort constraints and the difference between simulated and occupied buildings.
Architectural question
Which comfort constraints belong in an energy-control experiment?
Energy-Efficient Thermal Comfort Control in Smart Buildings via Deep Reinforcement Learning
RESEARCH /
The proposed framework learns thermal-control policies while considering HVAC energy use and comfort. Its relevance is the tension between energy reduction and occupant conditions. Evaluation should retain the building's operating constraints and compare the learned policy with a clearly specified conventional-control baseline.
Architectural question
How should a thermal-control policy be compared fairly?
Non-invasive thermal comfort perception based on subtleness magnification and deep learning for energy efficiency
RESEARCH /
This study relates subtle skin motion and texture features to thermal measurements using deep learning. It examines non-contact sensing for comfort feedback. Architectural implementation would require a separate assessment of measurement validity, consent and data handling, especially when model output is used to change occupied-space conditions.
Architectural question
What evidence justifies non-contact personal comfort sensing?
Xiaogang Cheng, Bin Yang, Anders Hedman, Thomas Olofsson, Haibo Li, Luc Van Gool
Building Footprint Generation Using Improved Generative Adversarial Networks
RESEARCH /
The authors adapt conditional adversarial generation for satellite-derived footprints, using a different loss formulation. This is a study of image-to-map reconstruction, not building design generation. Its outputs need careful comparison with source imagery before informing coverage or context diagrams.
Architectural question
When does a generated footprint remain faithful to source imagery?
Overcoming Missing and Incomplete Modalities with Generative Adversarial Networks for Building Footprint Segmentation
RESEARCH /
This research uses generative methods to address missing or incomplete modalities in remote-sensing fusion. It matters when a site dataset lacks one expected input. Imputed information should remain visible in the record, since a plausible replacement is not equivalent to an observed measurement.
Architectural question
How should a map distinguish fused measurements from imputed data?
Benjamin Bischke, Patrick Helber, Florian König, Damian Borth, Andreas Dengel
Classification of Building Information Model (BIM) Structures with Deep Learning
RESEARCH /
The authors classify images exported from BIM software into broad building categories and compare learning methods. The study uses model images rather than site photographs. Its relevance is organizing design archives, with a clear limit: visual category recognition cannot establish a building's actual program, performance or regulatory use.
Architectural question
How useful is BIM-image classification for a design archive?
Francesco Lomio, Ricardo Farinha, Mauri Laasonen, Heikki Huttunen
Machine Learning for Building Energy and Indoor Environment: A Perspective
RESEARCH /
This perspective discusses machine learning for energy and indoor-environment problems, including fungal-concentration prediction. It is useful for examining the difference between a fitted model and an actionable building decision. Model structure and available training data remain central limitations.
Architectural question
What makes an indoor-environment prediction actionable?
Multi-Task Learning for Segmentation of Building Footprints with Deep Neural Networks
RESEARCH /
This paper examines boundary preservation in high-resolution satellite segmentation through multiple learning tasks. It is relevant to urban figure-ground preparation, where outlines carry spatial meaning. Readers should inspect whether sharper edges also preserve small buildings and narrow intervening spaces.
Architectural question
Does sharper segmentation preserve small-scale urban grain?
Benjamin Bischke, Patrick Helber, Joachim Folz, Damian Borth, Andreas Dengel
On-line Building Energy Optimization using Deep Reinforcement Learning
RESEARCH /
The authors explore deep reinforcement learning for adjusting energy-management schedules as data arrive. The research concerns operating policy, including alternative learning approaches. A real-building comparison would need to account for exploration costs, equipment limits and comfort disruptions during learning, not merely the eventual optimized schedule.
Architectural question
What costs arise while an energy controller learns online?
Elena Mocanu, Decebal Constantin Mocanu, Phuong H. Nguyen, Antonio Liotta, Michael E. Webber, Madeleine Gibescu, J. G. Slootweg
Building Energy Load Forecasting using Deep Neural Networks
RESEARCH /
This paper uses recurrent neural-network methods to forecast energy demand at building scale. The architectural relevance is planning for changing loads rather than describing a fixed consumption profile. Readers should compare time horizons, weather inputs and error patterns during peaks as well as average periods.
Architectural question
Where do load forecasts fail during unusual operating conditions?
Online survey for collective clustering of computer generated architectural floor plans
RESEARCH /
The study surveys how practitioners and students group computer-generated plans. It investigates perceived similarity and its relevance to retrieval and clustering. This is useful for questioning whether an algorithm's distance measure matches architectural judgment about organization, proportion and use rather than superficial visual resemblance.
Architectural question
What makes two floor plans meaningfully similar to architects?
David Sousa-Rodrigues, Mafalda Teixeira de Sampayo, Eugénio Rodrigues, Adélio Rodrigues Gaspar, Álvaro Gomes, Carlos Henggeler Antunes
Social Game for Building Energy Efficiency: Utility Learning, Simulation, and Analysis
RESEARCH /
The authors model an occupant voting game for lighting preferences, linking comfort with incentives. The study offers a behavioral lens on building energy decisions rather than an architectural performance guarantee. Its assumptions about utility and participation need scrutiny before transferring the game to another workplace.
Architectural question
How can lighting incentives account for different occupant preferences?
Ioannis C. Konstantakopoulos, Lillian J. Ratliff, Ming Jin, S. Shankar Sastry, Costas Spanos
