This work proposes discrete style transfer (DST) as an alternative to conventional neural style transfer methods, reframing the operation from seamless blending to part-based substitution. Neural models such as VGG-based style transfer, GANs, and diffusion models privilege continuity, producing visually coherent but cognitively estranging images that erase seams essential to human perception, critique, and fabrication. Building on theoretical insights, the paper argues that seams and discontinuities should be treated as design data rather than flaws. The method developed here subdivides images into adaptive quadtree tiles, encodes them with an autoencoder into latent embeddings, and substitutes parts via nearest-neighbor search. Results demonstrate significant gains in speed and stylistic fidelity compared to alternative image comparison methods and mosaic baselines, while remaining competitive with neural style transfer in visual similarity. Beyond efficiency, discrete style transfer produces modular, interpretable outputs aligned with architectural logics of repetition, assembly, and material agency.
Urban sensing systems increasingly rely on vibration-acoustic and kinetic signals that move through bodies, buildings, and terrain-to detect and respond to activity. Platforms such as gunshot-detection arrays in US cities, seismic sensors along the southern border, and piezoelectric monitoring systems in elder-care settings convert conditions shaped by segregation, migration control, and demographic precarity into problems of signal detection and rapid response. This article examines that vibrational regime through a close analysis of ShotSpotter, the most widespread urban acoustic-detection platform in the United States. It traces ShotSpotter's seismological lineage in Cold War sensing infrastructures, its operational life in Chicago, and the temporal form it produces through realtime detection. The article then turns to Mendi + Keith Obadike's Free/Phase: Node 1 Beacon to develop a counter-practice grounded in historical memory and collective listening. Together, these cases show how vibration organizes contemporary urban governance by structuring what can be registered and acted on, while alternative sonic practices sustain other relations to place, history, and shared experience. The article argues that any recalibration of these systems requires confronting their imperial lineage and developing forms of sensing that remain accountable to the conditions they engage.
Current approaches to coastal resilience often fail because they rely on prediction and control in an era of deep climatic uncertainty. This research proposes an environmental architecture that serves as an adaptive mediator, facilitating co-assembly between human intervention and natural systems. This design methodology shifts the paradigm from computed solutions to computing systems that continuously adapt via environmental feedback. Using a high-throughput computational pipeline, we explore a parameter space of geometric configurations, designing submerged structures that modulate hydrodynamic energy to direct sediment transport and grow a nascent island. These structures were deployed as a Real-World Lab in the Maldives, serving as a full-scale test of this Co-assembly approach to resilience. Over a 12-month period, the structures facilitated the accretion of nascent landforms, demonstrating an adaptive alternative to dredging. The results illustrate the potential for computational design to instrumentalize environmental science, creating architectural systems that do not merely inhabit a site but actively compute with its physical forces to build resilience over time.
The rup-rup technique is a traditional incision-based method for bending bamboo culms that relies on tacit craftsmanship and material feedback rather than formalized geometric rules. Although recent computational studies have modeled rup-rup using curvature-based segmentation, most approaches assume uniform culm geometry and do not account for longitudinal variability in internode spacing and diameter fluctuation. This study develops a material-informed parametric framework that translates tacit craft heuristics into rule-based computational logic while integrating culm-specific geometric data.An experimental-computational methodology was employed, combining craft observation, manual geometric measurement of twenty-one Gigantochloa apus specimens, parametric scripting in Rhinoceros/Grasshopper, and full-scale bending experiments. Two computational models were evaluated against traditional practice: a curvature-driven Tangent-Curve Method and a craft-informed Length-Differential Method. Each model was tested with and without incorporation of literature-based elastic parameters. Results indicate that generalized morphological datasets were insufficient for accurate curvature prediction, whereas element-specific geometric input significantly improved bending accuracy and reproducibility. Furthermore, stabilizing cut width within the parametric logic reduced operator-dependent variability while preserving adaptive material response.The findings demonstrate that computational formalization does not substitute craftsmanship but redistributes decision-making by encoding geometric consistency while maintaining material sensitivity. The proposed framework advances material-aware parametric modeling for non-uniform natural materials and contributes to hybrid digital-craft design methodologies in architectural computation.
Traditional apprenticeship models struggle to scale as the construction industry faces a growing shortage of skilled workers and an aging workforce. This study evaluates the potential and strategies of Large Language Models (LLMs) to support apprentices in learning hands-on construction tasks through real-time, conversational instruction. Drawing on prior research in conversational AI and intelligent tutoring systems, we conduct a comparative analysis of LLM-based guidance versus traditional video demonstrations in controlled masonry tasks. Through a mixed-methods approach, we assess task performance, interaction patterns, and participants' self-reported confidence and understanding. Findings from our exploratory comparative study suggest that LLMs can deliver relevant, adaptive, and context-aware procedural guidance. However, limitations emerged in conveying tacit knowledge and adapting tool use to the specific task context. The results underscore the importance of interface design and instructional modality in sustaining engagement. This work offers early insights into the design of scalable, AI-assisted learning systems for skilled trades.
Participatory design in informal urban contexts often relies on low-threshold analog media to surface situated knowledge, while computational tools enable rapid iteration and auditable records. Yet, when analog outputs are digitized in a one-way pipeline, cultural meaning and participant authorship can be stripped away. This paper examines how representational media shape design cognition and participation by comparing two workshop conditions in a Bangkok neighborhood: on-site analog illustration and a browser-based voxel editor built with Three.js. Across 16 youth and young-adult participants (ages 18-28), analog sessions elicited narrative, symbolic, and affective reasoning that articulated place-based memories, rituals, and care practices. Voxel sessions supported modular configuration, circulation testing, and visible revision histories through rapid branching and interaction logs. Because neither modality alone was sufficient, we frame reciprocity cautiously. The study establishes a comparative foundation and provides a worked proof-of-concept for round-trip annotation that traces one analog narrative into voxel translation and back while preserving authorship and context. We contribute comparative evidence linking media to cognition and participation dynamics, a lightweight specification for a community-ready voxel workflow including export and logging primitives, and an operational protocol for narrative round-tripping to support design justice and accountable participation in understudied Southeast Asian contexts.
Evaluating the quality of generative design machine learning tools is a critical challenge. Existing methods range from human-based assessments to performance-based metrics and statistical comparisons. We focus on generative urban design and critically review the evaluation methods employed in recent literature. We experimentally test and comprehensively analyze these methods. We find that existing approaches favor disrupted designs over well-designed ones, have inherent limitations, and fail to capture the tool performance quality. To address this critical gap, we develop two strategies: (1) modifying the Fr & eacute;chet Inception Distance (FID) score to align with specific design principles, and (2) leveraging visual language models to assess design outputs. Our experiments show that these approaches provide more robust and comprehensive evaluations. The findings underscore the need for practical and reliable evaluation frameworks for AI in design fields to advance the research in this field.
Urban Digital Twins have been criticized for their reductive modeling of social processes and lack of contextual embedding. In response, the proposed unified Geospatial Digital Twin (GDT) framework not only combines diverse modeling techniques into a standardized environment, the cyber focus that enables technical interoperability but, critically, contextualizes these geospatial data within local social and physical realities, the social focus of the GDT. Specifically, the GDT integrates computational modeling with participatory processes through four continuums: abstracting, twinning, experiencing, and gaming. This modular architecture allows implementations to emphasize dimensions according to project-specific needs. We demonstrate GDTs' flexibility through a neighborhood-scale application named Campustwin and a city-scale application named JaxTwin. By integrating cyber-focused technical capabilities with social-focused contextual understanding, GDTs bridge computational modeling with place-based social realities, offering opportunities to address environmental challenges and advance human-centered decision-making for resilient urban futures.
Miniatures do not simply reduce scale-they generate alternate temporal and perceptual regimes that anticipate how we now inhabit digital space. Their scenographic logic flattens distinctions between architecture and entourage, compressing duration into private time bubbles and collapses haptic intimacy with optic distance. This condition operates through the same logics as windowed interfaces, smartphone screens, and augmented reality. What architectural theory once dismissed has become the fundamental condition of contemporary spatial production. Tracing miniaturization from full-scale composite urbanism through scaled models to life-sized reconstructions, this essay reveals how architecture has absorbed the logic of zooming, resolution, and ontological montage-operations that now define how we navigate environments assembled from appropriated fragments through incompatible scales and temporalities.
AI technologies have been widely explored in the architectural design process over the last decade. This paper addresses the limitations of the existing reviews on AI applications in the architectural design process, including a lack of focus on the design process, limited coverage of AI technologies, and under-exploration of human-AI interaction environments. The paper systematically reviews and comparatively analyses 63 articles filtered through 1138 publications from 8 databases. The findings include comparative analysis charts and tables of the different AI-enhanced design processes, the human-AI interaction environment, as well as their evaluation. They highlight the expectation that AI will function as a conversational design partner. Lastly, the paper presents a novel framework for the AI-enhanced conversational architectural design process.
The current advancements of deep learning models offer potential applications for computational design through sets of generated images controlled by parametric inputs, yet they remain disconnected from geometry-driven parametric tools. For this reason, we study the implications of text and image-based generation methods to be used in traditional parametric design procedures. We implement this study by integrating Stable Diffusion and ControlNet to Rhino Grasshopper through a Python-based remote-API plug-in. This API allows a direct connection to the diffusion-based image generation methods without any middleware. Our main contribution is to enable architects and designers to interactively generate and investigate new design ideas in their native parametric design environment. We evaluate potential impact on parametric design education with 15 architecture students using a single GPU server running Stable Diffusion v1.5 across three exercises: Text-to-Image, Image-to-Image using Rhinoceros view captures, and Parametric-Model-to-Image with ControlNet. Quantitative results showed that the API-enabled image generation averaged 4-15 seconds per image, allowing seamless integration with parametric workflows for all 15 students in a classroom setting. Performance evaluations show that our approach offers significantly improved efficiency and responsiveness compared to existing diffusion-based tools, highlighting its suitability for seamless integration within parametric design environments. Qualitative feedback indicated improved design ideation, greater fluency in prompt engineering, and enhanced understanding of parametric logic through iterative visual experimentation. These findings demonstrate the potential of real-time AI integration to augment both conceptual design and parametric design education.
This study proposes a studio pedagogy that integrates knitting techniques with digital design and fabrication in architectural education. Grounded in Kolb’s Experiential Learning Theory, the workflow spans hands-on material trials, AI-assisted variation, parametric modeling, and computer-aided manufacturing (3D printing). Using a rubric-based assessment with third-year students ( N = 10), we observed the largest gains in material experimentation (mean 3.1 → 4.5/5) and 3D printing optimization (3.0 → 4.3/5); 80% of projects improved after structured feedback and iteration. These results indicate that knitting, coupled with CAD/CAM, supports a transparent, measurable learning environment that links craft-based exploration to computational reasoning. Findings are preliminary due to the small elective cohort and warrant replication with larger samples and controlled variables.
A computational framework is introduced to map aesthetic categories, highlighting distinctions between human-created competition entries and AI-generated designs. Assessment of architectural aesthetics has traditionally been subjective; however, computational methods now enable the systematic analysis and visualization of design variations. Using 14 aesthetic parameters, the framework applies Principal Component Analysis (PCA) to create visual maps of aesthetic relationships. Two hypotheses are discussed: (1) There is a hegemony of the aesthetic category of the beautiful in mainstream architectural competitions, and (2) that generative models like Stable Diffusion can produce visuals belonging to other aesthetic categories. The computational aesthetic framework is applied to verify both. The first hypothesis is disproven by revealing that human-created designs demonstrate significant aesthetic variation and unpredictability, while the second is confirmed by demonstrating the capacity to produce images in 16 other categories beyond the beautiful. The maps of visual relationships group images by aesthetic categories, encouraging designers to explore and enhance underpopulated areas. The computational aesthetics framework is used to analytically and quantitatively support arguments regarding architectural aesthetics.
This paper argues for the pedagogical reframing of architectural theory as a constructive endeavor within an undergraduate architecture curriculum. Defining theory as a construct empowers students to engage in their own theory-making through a series of analytic and synthetic studies that activate understanding of the principles, practices, and procedures embedded in architectural precedents and their representations. The pedagogical research outlined here presents an approach in which a second-year undergraduate course introduces architectural theory by engaging a corpus of Paul Rudolph's early Florida houses, completed between 1946 and 1962. A series of four exercises focused on decoding, transforming, blending, and curating guides the conversations in the course. Uniquely, these exercises rely on the shape grammar formalism, a discourse that bridges the gap between analysis and synthesis, paving the way for students to connect architectural ideas and their numerous interpretations to the rule-based, constructive thinking that is fundamental to computation and design.
Applications of Artificial Intelligence AI are increasingly significant for designers across various fields, with a particular emphasis on architectural design. These applications offer support by providing relevant data and suggesting diverse design ideas. This study explores the applications of AI in the architectural computational design process, employing a mixed-review approach that combines bibliometric analysis and systematic review. The objective is to study the uncovered research area figured from the review by clarifying the primary functions of AI in all architectural design stages, as well as the associated opportunities and challenges. This study examines the application scale, methodologies, and tools of AI in architectural design. Then, it conducts a comprehensive survey of commonly used AI tools, analyzing and comparing them based on phase classification, deployment classification, scale of application, and their integration with BIM, VR, and parametric design Additionally, the study proposes an AI-powered Architectural Computational Design Process (AI-ACD), a workflow designed to aid architects in effectively incorporating AI technologies into various stages of computational architectural design processes. Afterwords the study also introduces a classification of commonly used AI tools and maps them to specific design tasks within the (AI-ACD) workflow. Multiple design scenarios and a set of core integration principles are proposed to demonstrate how AI engagement can be tailored to different levels of use and project contexts. Finally, the study presents a matrix that maps each core integration principle to the five key design stages The study demonstrates how AI-ACD, an AI-assisted workflow, enhances the architectural design process through visualization, data analysis, and optimization tools, adaptable to architecture, urban design, and heritage, ultimately boosting creativity, efficiency, and design quality.
This paper proposes a computational framework that integrates vernacular architecture knowledge (VAK) into genetic algorithms (GA) to enhance architectural design optimization. First, in addition to the parameters required for the design optimization process, constants derived from VAK are introduced. Secondly, an algorithmic model is presented in which these extracted parameters and constants are incorporated into GA processes. The integration of VAK-based constants has the potential to improve architectural design optimization while preserving the local structural characteristics of the design. This approach emphasizes the designer's expertise by reducing the number of meaningless variations in GA processes and increasing efficiency. The proposed method is demonstrated through a case study that generates design variations for Kara & ccedil;ad & imath;r, a traditional structure used by the Y & ouml;r & uuml;ks, a nomadic culture in Anatolia. The study incorporates key design elements: cover modules, load-bearings, and connectors to produce variations that preserve the traditional structure of Kara & ccedil;ad & imath;r.
To enhance the deployment potential and architectural integration of photovoltaic technologies in open public spaces, the design and development of an actuated, lightweight and unitized prototype canopy structure is presented. The system incorporates thin-film photovoltaic modules mounted on aluminum substrates, configured to track the solar trajectory for optimized energy generation. The structural assembly comprises a primary cable net coupled with a secondary system of struts and control cables, anchored to a perimetric frame. The design emphasizes structural and technical simplicity, low self-weight and a minimal number of actuation elements. The integrated interdisciplinary process followed encompasses iterative phases of geometrical system development, and analysis of the load-deformation behavior, associated kinematics and energy performance, as well as the fabrication and experimental verification of a scaled 1:5 prototype. The process is anticipated to contribute to technological innovation in future real case applications, addressing key architectural, structural, kinematic, control and energy performance parameters.
The existing multi-objective scheduling optimization methods for prefabricated construction projects do not consider scheduling problems under uncertain project structures, and the efficiency of the scheduled projects is not ideal. To address this issue, a multi-objective scheduling optimization method grounded on an improved Nondominated Sorting Genetic Algorithm II (NSGA-II) is proposed. The study first establishes a multi-objective optimization mathematical model considering the uncertainty of project structure from the perspective of changes in project structure. This model aims to minimize project cost and duration, considering various execution modes and uncertainties of activity processes in prefabricated construction projects. Afterwards, by introducing the NSGA-II and taboo search algorithm, the scheduling problem of the construction project is optimized and solved to obtain the best scheduling solution. The results showed that after using this optimization method, the resource utilization rate of the project was 92.36%, the material cost consumption was 158.60 yuan/m2, and the completion rate of the construction period was 98.77%. Its progress deviation rate was only 2.41%, and the unit area cost was 1850.32 yuan/m2, which has a higher cost-effectiveness compared to other methods. The engineering qualification rate of the application project was 99.78%. The research designs a multi-objective scheduling optimization method for prefabricated construction projects, which can effectively improve project construction efficiency and reduce costs while ensuring construction quality.
