Mixed reality construction expands the possibilities of traditional construction methods across multiple dimensions. In exploring these possibilities, various research trends have emerged, including the use of natural materials, the construction of complex geometries. Building upon these existing studies, this paper pro-poses a workflow for additive construction using natural materials within a mixed reality environment, and validates its feasibility through a practical project. By combining discrete design rule-setting with human interaction in mixed reality, the workflow simplifies the processing, fabrication, and assembly of irregular materials. In this process, discrete design—as a form of additive manufacturing algorithm—enables natural materials to be aggregated in a structured manner, forming construction systems that are both random in appearance and operable in practice. This approach takes advantage of the small volume of offcut materials, and the high mobility of discarded natural components, while also leveraging mixed reality technology to reduce the assembly complexity typically associated with discrete design.

This paper proposes a deep learning-based intelligent modeling framework for generating 3D architectural models from manual sketches, addressing the domain gap in 2D-to-3D transformation. By integrating architectural domain knowledge—specifically the phased, selective, and cyclic characteristics of the design process—the framework ensures a structured and iterative generative approach. The framework consists of a 2D design phase, where image retrieval, Stable Diffusion, and CycleGAN facilitate conceptual exploration, multi-scheme generation, and depth map extraction, and a 3D design phase, where Pixel2Mesh generates 3D forms, refined through Grasshopper-based parametric optimization. Empirical evaluation demonstrates that the framework effectively preserves structural fidelity while allowing for generative variations. Structural similarity and geometric accuracy metrics validate its performance, confirming its ability to balance AI-driven massing generation with architectural precision. A Mars habitat case study, conducted in an academic research setting, serves as a controlled experiment to assess adaptability. While demonstrating the framework's potential for AI-assisted architectural generation, the study also highlights the need for broader validation across diverse architectural typologies. This research bridges traditional and AI-driven design methodologies by integrating computer vision and generative models into architectural workflows. The proposed framework contributes to architectural design by introducing a cross-disciplinary approach that enhances the efficiency, quality, and innovation of design processes.

The increasing reliance on digital tools in architectural education has transformed design workflows, offering new opportunities for creativity while posing challenges to students’ logical reasoning and structured problem-solving abilities. While digital tools facilitate automation and generative design, over-reliance on them can limit students’ ability to navigate design complexity independently. Addressing this issue, this study develops the Logic-Driven and Technology-Supported Creativity Development Model to examine the roles of logical frameworks, digital tools, and open-ended design tasks in fostering structured creativity. The findings reveal that logical frameworks provide essential cognitive scaffolding, helping students balance creative exploration with structured decision-making. Digital tools enhance form generation but introduce challenges such as automation bias and steep learning curves. Open-ended tasks promote design flexibility, yet their effectiveness depends on logical structures to maintain coherence. This study highlights the importance of curriculum design in supporting structured creativity, emphasizing the integration of technical training, interdisciplinary methods, and reflective learning. The findings contribute to design education theory and provide practical insights for improving course structures and pedagogical approaches in digital design environments.

Accurate prediction of workplane daylight illuminance and eye-height glare is crucial for lighting control. Existing studies used machine learning to predict illuminance at predetermined locations based on indoor sensors, but they may encounter challenges in scenarios 1) with flexible seating arrangements, 2) with dynamic shading devices, and 3) requiring the prediction of glare. To address these challenges, we proposed a novel method fusing Transformer and Diffusion models, with the input being data collected from sparse ceiling-mounted illuminance sensors, and the outputs being high-resolution workplane illuminance and glare. The model works well for rooms without and with dynamic roller shades. For the former, the mean absolute errors for illuminance below 3000 lx and Daylight Glare Index (DGI) are only 20.77 lx and 0.20, and the error rates in detecting illuminance <500 lx and DGI>22 are only 0.85 % and 5.55 %. For the more complicated latter case, the aforementioned four numbers are 34.78 lx, 0.59, 2.47 % and 23.13 %. The model significantly outperforms the linear and the ANN models, particularly in glare prediction. The influence of sensor number and placement strategy on model performance was also revealed. The model can potentially enhance lighting control, especially in cases with dynamic shading, with flexible seating arrangements, and where glare is of interest.

Electroencephalography (EEG) can help understand user feedback in the built environment, as a real-time and less intrusive method than questionnaires, and is more comprehensive and informative than other physiological measurements. However, existing studies either used simple rhythm-specific metrics, or utilized machine learning (ML) but excluded the more challenging cross-subject scenarios, restricting the result reliability and application scope of EEG. This study integrated EEG and ML to predict comprehensive occupant feedback in the built environment, including scene identification, spatial perception and emotion, environmental quality perception, and work efficiency, totalling 12 items. The EEG data was acquired while subjects were immersed in three VR scenes, with questionnaires collecting perceptual feedback and cognitive tasks reflecting work efficiency. Within-, all- and cross-subject scenarios were all tested, and four common ML algorithms and five EEG feature types were evaluated. Results suggested good prediction capabilities of EEG coupled with ML for various feedback items, achieving accuracies ranging from 0.85 to 0.96 for within-subject scenarios and from 0.88 to 0.94 for all-subject scenarios. For cross-subject scenarios, most items achieved accuracies near or higher than 0.7, suggesting potential applications in public spaces with new subjects. Random Forest is generally the best-performing and most stable algorithm, and Differential Entropy and Geometrical Features are the recommended EEG feature types, especially for cross-subject scenarios. These findings may enhance intelligent building control, improve interaction with the immersive virtual environment, and support human-centred design, while methods to improve cross-subject prediction accuracies such as deep learning and multi-modal approaches need to be further investigated.

Visual reinforcement learning (RL) has shown promise in continuous control tasks. Despite its progress, current algorithms are still unsatisfactory in virtually every aspect of the performance such as sample efficiency, asymptotic performance, and their robustness to the choice of random seeds. In this paper, we identify a major shortcoming in existing visual RL methods that is the agents often exhibit sustained inactivity during early training, thereby limiting their ability to explore effectively. Expanding upon this crucial observation, we additionally unveil a significant correlation between the agents' inclination towards motorically inactive exploration and the absence of neuronal activity within their policy networks. To quantify this inactivity, we adopt dormant ratio as a metric to measure inactivity in the RL agent's network. Empirically, we also recognize that the dormant ratio can act as a standalone indicator of an agent's activity level, regardless of the received reward signals. Leveraging the aforementioned insights, we introduce DrM, a method that uses three core mechanisms to guide agents' exploration-exploitation trade-offs by actively minimizing the dormant ratio. Experiments demonstrate that DrM achieves significant improvements in sample efficiency and asymptotic performance with no broken seeds (76 seeds in total) across three continuous control benchmark environments, including DeepMind Control Suite, MetaWorld, and Adroit. Most importantly, DrM is the first model-free algorithm that consistently solves tasks in both the Dog and Manipulator domains from the DeepMind Control Suite as well as three dexterous hand manipulation tasks without demonstrations in Adroit, all based on pixel observations.

Enhancing accessibility for older adults, particularly addressing challenges related to seniors navigating stairs, constitutes a significant theme in current human-computer interaction research. This paper details the design and development of an economical, small-scale walking assistive device, StairMate, to substantially increase the balance in older adults during stair use. The device can facilitate the convenient carriage of daily necessities for older adults and support ascending and descending stairs through its front handle. The preliminary verification of StairMate’s effectiveness in assisting elderly individuals in navigating stairs is conducted by measuring surface electromyography activity in muscles. The results demonstrate over 65 % increased lower limb stability for elderly individuals ascending stairs and over 55 % during descent. The device offers a practical approach to accessible design, bringing great convenience to daily life and enhancing the living environment for older adults.

Interactive art education is an important part of aesthetic education, and a number of AI-related technologies, especially robotics, should be integrated into the creativity and practice of interactive art, and in turn, art should also permeate the whole process of engineering or technology teaching. The curriculum reform takes “great aesthetic education” as the teaching philosophy, emphasizing the integration of doing, learning, reading and thinking; The teaching content of the original Single discipline field of engineering technology application was sorted out and transformed into the teaching content positioned in the interdisciplinary field between art and other technical disciplines. Adjust the content of the design and production process; Adopt various forms including school-enterprise cooperation; Highlight the practical features of the course. By the comprehensive feedback results after implementation, the teaching reform model was applied to achieve Better teaching effect. It has laid a certain foundation for promoting the combination of aesthetic education in colleges and universities with some technical disciplines education.

The varying significance of distinct primitive behaviors during the policy learning process has been overlooked by prior model-free RL algorithms. Leveraging this insight, we explore the causal relationship between different action dimensions and rewards to evaluate the significance of various primitive behaviors during training. We introduce a causality-aware entropy term that effectively identifies and prioritizes actions with high potential impacts for efficient exploration. Furthermore, to prevent excessive focus on specific primitive behaviors, we analyze the gradient dormancy phenomenon and introduce a dormancy-guided reset mechanism to further enhance the efficacy of our method. Our proposed algorithm, ACE: Off-policy Actor-critic with Causality-aware Entropy regularization, demonstrates a substantial performance advantage across 29 diverse continuous control tasks spanning 7 domains compared to model-free RL baselines, which underscores the effectiveness, versatility, and efficient sample efficiency of our approach. Benchmark results and videos are available at https://ace-rl.github.io/.

The remarkable performance of large-scale Text-to-Image generation(TI) models is evident in their ability to produce high-quality and diverse images. However, despite advancements, the field of image editing still faces challenges. Current methods struggle to strike a balance between fidelity and powerful editing capabilities. Moreover, approaches that do not involve fine-tuning fail to produce diverse editing results. We introduce Noise Weighting Phased Prompt Image Editing (NWPP), a method that excels in powerful editing, high fidelity, and diverse results without fine-tuning. Our approach involves a two-phase generation process. The first phase employs the original prompt to guide initial image editing, ensuring a layout resembling the original image. In the second phase, a noise-weighting technique based on the Cross-Attention map minimizes the impact of the target text on non-editing regions. Further enhancement is achieved through the integration of the KV injection module, expanding the editing capabilities and enabling diverse result generation. Experimental evaluations, conducted on both generated images and the COCO dataset, affirm the efficacy of our method.

The Internet of Things (IoT) has the potential to significantly ameliorate living environments, notably by integrating smart architecture and smart cities. To actualize this potential, however, addressing a range of technical, social, and economic challenges is imperative. This study systematically explores these challenges and delineates the future prospects of IoT in augmenting living environments. The technical obstacles, including issues of data privacy and security, network and device compatibility, and interoperability and standardization, are examined in conjunction with social challenges such as user acceptance and adoption, accessibility, and inclusiveness, as well as ethical and legal considerations. The economic challenges, such as cost and funding, return on investment, and implementation and deployment processes, are also scrutinized. This work underscores the importance of smart architecture and smart cities in the context of IoT's future scope. It advocates for incorporating IoT within an intelligent ecosystem, which includes technical elements like data collection and integration, advanced data analytics, and real-time monitoring, as well as social and economic aspects like enhanced user experience, increased efficiency, and improved security. The insights presented in this work are of immense value to researchers, practitioners, and policymakers striving to devise and implement robust, inclusive, and ethical IoT solutions. Hence, this study stands as an essential resource, shedding light on the journey towards a more interconnected, efficient, and enhanced living environment.

3D Concrete Printing (3DCP) is considered a solution to confront both labor shortages and material waste in the construction industry. However, to realize non-standard, large-scale buildings, substantial amounts of supporting material and manual assembly are still required to print structures with significant overhangs, including the inclined structure and the cantilever structure. In response, the study draws inspiration from mushroom lamellae and proposes a geometry-based approach to adjusting the large-overhang shape into a self-supported one. This approach imitates lamella morphology by creating folds on the given geometry based on constraints established through empirical studies, thereby reducing overhangs and improving printability. Specifically, the proposed approach includes two shape variation methods. The first method generates lamellae perpendicular to the inclined shape, thereby providing lateral support to the filament. The second method creates lamellae under the cantilever, facilitating a seamless horizontal-to-vertical transition. The effectiveness of these methods is demonstrated through the generation and printing of four column capital samples. Upon evaluating the overhang mechanisms of these geometries, an approachable optimization method is also presented to further minimize the amplitude of lamellae while maintaining self-supporting ability. Compared with control groups that focus on either maximizing printability or minimizing material usage, the optimal result for the inclined structure prototype showcases an over 75 % overhang reduction on the critical layer with just about 15 % use of additional material. As for the cantilever prototype, an over 75 % overhang reduction is achieved on the critical layer with about 35 % additional material usage. Various large-scale structure examples are presented at the end of the research, illustrating the applicability and scalability of the lamella-inspired method in a wider range of building types.

Research and development of a set of integrated robot sorting inspection experimental system, the system in industrial automation, quality control, safety monitoring and other fields has a wide range of application prospects. The technical method uses a GPU-enabled TensorFlow Docker image, based on the Mask R-CNN (Mask Region-based Convolutional Neural Network) network architecture, uses the COCO (Common Objects in Context) weight file as the pre-trained deep learning model, and uses VIA (VGG Image Annotator) to label the dataset of the part of end cap of the reducer. The actual experimental results show that: the UR10 Robot is driven to pick up and handle the a single end cap part in the stacked parts to the designated location, the UR10e Robot, with on the MetraSCAN 3D scanner, is equipped with a C-Track tracker and realizes automatic inspection experimental through a programmed program. The experimental teaching application effect has also been well feedback. The research results not only provide a good research foundation for the development and application of system-related technologies, especially measurement or robotics, but also provide a reference for the integration of software and hardware resources in the laboratory.

This paper explores the potential of digital stereotomy in combination with robotic fabrication to increase the precision and complexity of stone processing. To enable the application of these techniques in outdoor environments, modular joints designed for robotic assembly are necessary. Additionally, the cutting process must be efficient and minimize material waste. To address these challenges, this research proposes a parametric wave joint design that enables rapid cutting and straightforward assembly by a robotic system. The joint contains motion space allowing it to slide into accurate assembly position, enabling the robot to complete the assembly without requiring highly precise vision or gripper in outdoor situations. Furthermore, the wave joint design eliminates the need for milling, reducing the processing time. The paper presents a robotic arm-cutting method for this joint and conducts experiments using foam and robotic arm hot-wire cutting to simulate stone cutting. The feasibility of the joint is tested through the assembly of a bent column, and finite element analysis is used to compare the stresses on two joint parts under shear force with different control parameters. The study confirms the feasibility of the wave joint design for robotic assembly and the efficiency of robotic arm cutting. The findings may inform the development of modular assemblies for robotic systems in stone processing applications.

3D concrete printing (3DCP) technology is a construction method that offers a unique combination of automation and customization. However, when the printing area goes large, generating the print path becomes a sophisticated work. That’s because the customized print path should not only be expandable but also printable, such rules are hard to follow as both the printing area and construction requirements increase. In this paper, the Shenzhen Baoan 3D Printing Park project serves as a case study to introduce space-filling and print path generation methods for three types of large-area concrete pavement. The space-filling methods utilize geometry-based rules to generate complex and expandable paving patterns, while the print path generation methods utilize construction-oriented rules to convert these patterns into print paths. The research provides easy-to-operate design and programming workflows to achieve a pavement printing area of 836 sqm, which significantly increases the construction scale of large-format additive manufacturing (LFAM) and shows the potential of 3D printing technology to reach non-standard results by using standard workflows.

Sketch is a way for architects to communicate with others. Architects record their own ideas through rapid drawing. However, sketches are abstract, vague, and even ambiguous. To this end, architects need to spend a lot of time, through modeling and other means, to present the architectural plan that can be understood by people. However, this method is time-consuming and laborious. Due to the development of deep learning technology, especially convolutional neural networks (CNN) and generative adversarial networks (GAN), they have shown great advantages in the field of image recognition and generation. With the help of these technologies, ambiguous architectural sketches can be directly transformed into architectural scheme drawings, and architects’ creative intentions can be continuously improved and developed, It will be very convenient and efficient. Therefore, based on the image-to-image translation, this paper realizes the mapping from architectural sketches to architectural scheme drawings with the help of CycleGAN. Through the analysis of the architectural generation design results of Frank Gehry's and Alberto Campo Baeza's architectural sketches, firstly, the feasibility of this method is verified. Secondly, it is found that this method can well complete the identification of sketch boundaries. In the generated scheme drawings, it can not only reflect the volume and lighting changes of the building, but also reflect the architect's creative intention and style to a large extent, The side reflects the cognitive ability of this method to architectural design.

The labour shortage is impacting the construction industry globally, and 3D concrete printing (3DCP) technology is considered a solution for such an issue by using robots instead of manpower in building construction.However, due to the unstable mechanical properties of the fresh-state concrete, current 3DCP building forms exclude a massive number of options, one of them being the overhang shape.Therefore, this paper introduces bionic wrinkles to improve the overhang of 3DCP shell prototypes and demonstrates the feasibility of this idea through three groups of buckling simulations.This study proves that surface wrinkles can improve the in-process mechanical properties as well as the overhang capacity of shell structures.In the best scenario, the original design which only reached 30° overhang can be promoted to 77.5°.Besides, the applications of all wrinkle types are robust.

本文面向机器人时代的建筑环境发展需求,建立"机器人建筑综合环境"的概念,体现为"增强机器人适应性的建筑综合环境"和"以人为本的机器人建筑综合环境"两方面内容.本文从移动适应、信息交互、空间安全、策略节能及场景设计与场所营造等方面出发,提出系统性的设计框架,以期为未来机器人建筑的设计提供理论支持与设计参考.