Dr. Hao Zheng graduated from the Ph.D. program at the University of Pennsylvania, specializing in machine learning, digital fabrication, mixed reality, and generative design. He holds a Master of Architecture degree from the University of California, Berkeley, and Bachelor of Architecture and Arts degrees from Shanghai Jiao Tong University. Phone: 2133784454 Address: G19 Meyerson Hall, 210 South 34th Street, Philadelphia, PA, USA, 19104
Proceedings of the 41st Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), 2021
This research aims to explore the quantitative relationship between urban planning decisions and ... more This research aims to explore the quantitative relationship between urban planning decisions and the health status of residents. By modeling the Point of Interest (POI) data and the geographic distribution of health-related outcomes, the research explores the critical factors in urban planning that could influence the health status of residents. It also informs decision-making regarding a healthier built environment and opens up possibilities for other data-driven methods. The data source constitutes two data sets, the POI data from OpenStreetMap, and the CDC dataset PLACES: Local Data for Better Health. After the data is collected and joined spatially, a machine learning method is used to select the most critical urban features in predicting the health outcomes of residents. Several machine learning models are trained and compared. With the chosen model, the prediction is evaluated on the test dataset and mapped geographically. The relations between factors are explored and interpreted. Finally, to understand the implications for urban design, the impact of modified POI data on the prediction of residents' health status is calculated and compared. This research proves the possibility of predicting residents' health from urban conditions with machine learning methods. The result verifies existing healthy urban design theories from a different perspective. This approach shows vast potential that data could in future assist decision-making to achieve a healthier built environment.
Proceedings of the 27th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2022
The regeneration of the industrial waterfront is a global issue, and its significance lies in tra... more The regeneration of the industrial waterfront is a global issue, and its significance lies in transforming the waterfront brownfield into an eco-friendly, hospitable, and vibrant urban space. However, the industrial waterfront naturally has comparatively unmanageable morphological features, including linear shape, irregular waterfront boundary, and separation with urban networks. Therefore, how to subdivide the vacant land and determine the land-use type for each subdivision becomes a challenging problem. Accordingly, this study proposes an application of machine learning models. It allows the generation of morphological elements of the vacant industrial waterfront by comparing the before-and-after scenarios of successful regeneration projects. The data collected from New York City is used as a showcase of this method.
Proceedings of the 3rd International Conference on Computational Design and Robotic Fabrication (CDRF), 2021
Health environment is a key factor in public health. Since people's health depends largely on the... more Health environment is a key factor in public health. Since people's health depends largely on their lifestyle, the built environment which supports a healthy living style is becoming more important. With the right urban planning decisions, it's possible to encourage healthier living and save healthcare expenditures for the society. However, there is not yet a quantitative relationship established between urban planning decisions and the health status of the residents. With the abundance of data and computing resources, this research aims to explore this relationship with a machine learning method. The data source is from both the OpenStreetMap and American Center for Decease Control and Prevention (CDC). By modeling the Point of Interest data and the geographic distribution of health-related outcome, the research explores the key factors in urban planning that could influence the health status of the residents quantitatively. It informs how to create a built environment that supports health and opens up possibilities for other data-driven methods in this field.
Proceedings of International Association for Shell and Spatial Structures Annual Symposia (IASS), 2021
This research investigates the use of graphic statics in analyzing the structural geometry of a n... more This research investigates the use of graphic statics in analyzing the structural geometry of a natural phenomenon to understand its performance and its relevant design parameters. Nature has always been the source of inspiration for designers, engineers, and scientists. Structural systems in nature are constantly evolving to optimize themselves with their boundary conditions. This optimization follows certain design rules that are quite challenging for a human to formulate or even comprehend. A dragonfly wing is an instance of a high-performance, lightweight structure that has intrigued many researchers to investigate its geometry and its performance as one of the most light-weight structures designed by nature. There are extensive geometrical and analytical studies on the pattern of the wing, but the underlying design logic is not clear. The geometry of the internal members of the dragonfly wings mainly consists of convex cell which may represent a compression-only network on a 2D plane. However, this property has not been geometrically analyzed from this perspective to confirm the hypothesis. In this research, we use the methods of 2D graphic statics to construct the force diagram from the given structural geometry of the wing. We use algebraic and iterative methods to report the topological and geometric properties of the form and force diagrams such as the degrees of indeterminacies of the network. For sample wings, we separate the internal and the boundary edges, construct the force diagram, and finally reconstruct the structural forms. Comparing the magnitude of the forces of the reconstructed network with the actual structure of the wing using the edge lengths of the force diagram will shed light on the performance of the structure. Multiple analytical studies will be provided to compare the results in both synthetic and natural networks and drive solid conclusions. The success in predicting the force flow in the natural structural pattern using graphic statics will expand the use of these powerful methods in reproducing the similar geometry of the natural structural system for the use in many engineering and scientific problems. It will also ultimately help us understand the design parameters and boundary conditions for which nature produces its master-pieces.
Proceedings of the 40th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), 2020
In this paper, we propose a geometry-based generative design method to generate and optimize a fl... more In this paper, we propose a geometry-based generative design method to generate and optimize a floor structure with funicular building members. This method challenges the antiquated column system, which has been used for more than a century. By inputting the floor plan with the positions of columns, designers can generate a variety of funicular supporting structures, expanding the choice of floor structure designs beyond simply columns and beams and encouraging the creation of architectural spaces with more diverse design elements. We further apply machine learning techniques (artificial neural networks) to evaluate and optimize the structural performance and constructability of the funicular structure, thus finding the optimal solutions within the almost infinite solution space. To achieve this, a machine learning model is trained and used as a fast evaluator to help the evolutionary algorithm find the optimal designs. This interdisciplinary method combines computer science and structural design, providing flexible design choices for generating floor structures.
Proceedings of the 40th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), 2020
The growing popularity of machine learning has provided new opportunities to predict certain beha... more The growing popularity of machine learning has provided new opportunities to predict certain behaviors precisely by utilizing big data. In this research, we use an image-based neural network to explore the relationship between the built environment and the activity of bicyclists in that environment. The generative model can produce heat maps that can be used to predict quantitatively the cycling and running activity in a given area, and then use urban design to enhance urban vitality in that area. In the machine learning model, the input image is a plan view of the built environment, and the output image is a heat map showing certain activities in the corresponding area. After it is trained, the model yields output (the predicted heat map) at an acceptable level of accuracy. The heat map shows the levels and conditions of the subject activity in different sections of the built environment. Thus, the predicted results can help identify where regional vitality can be improved. Using this method, designers can not only predict the behavioral heat distribution but also examine the different interactions between behaviors and aspects of the environment. The extent to which factors might influence behaviors also is studied by generating a heat map of the modified plan. Besides the potential applications of this approach, its limitations and areas for improvement also are proposed.
Proceedings of the 26th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2021
The free-form ice shell is the most challenging type in the design and construction of free-form ... more The free-form ice shell is the most challenging type in the design and construction of free-form buildings due to the regional and temporary nature of its materials. This paper presents a case study of the integration of design and fabrication of free-form ice shell. Taking the computational design and robotic fabrication of the ice shell as the main object, we discuss that combines the form-finding of the shell structure of graphic static with the tessellation technology of stereotomy, and propose a new method and workflow for the integration of discrete free-form ice shell design and construction. In the end, we built a free-form ice shell consisting of 116 discrete ice blocks. Practice has proved the feasibility of the integrated method of discrete free-form ice shell design and construction in the article.
Proceedings of the 26th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2021
Street vending is a recent policy advocated by city governments to support small and intermediate... more Street vending is a recent policy advocated by city governments to support small and intermediate businesses in the post-pandemic period in China. Street vendors select their locations primarily based on their intuitions about the surrounding environment; they temporarily occupy popular locations that benefit their business. Taking the city of Chengdu as an example, this study aims to formulate the rules governing vendors' location selection using machine learning and big data analysis techniques, thus identifying streets likely to become vital street markets. We propose a semantic segmentation method to construct heat maps that visualize and quantify the distribution of street vendors and pedestrians on public urban streets. The image-based generative adversarial network (GAN) is then trained to predict the vendors' heat maps from the pedestrians' heat map, finding the relationship between the locations of the vendors and the pedestrians. Our successful prediction of the vendors' locations highlights machine learning techniques' ability to quantify experience-based decision strategies. Moreover, suggesting potential marketing locations to vendors could help increase cities' vitality.
Style transfer is a design technique that is based on Artificial Intelligence and Machine Learnin... more Style transfer is a design technique that is based on Artificial Intelligence and Machine Learning, which is an innovative way to generate new images with the intervention of style images. The output image will carry the characteristic of style image and maintain the content of the input image. However, the design technique is employed in generating 2D images, which has a limited range in practical use. Thus, the goal of the project is to utilize style transfer as a toolset for architectural design and find out the possibility for a 3D modeling design. To implement style transfer into the research, floor plans of different heights are selected from a given design boundary and set as the content images, while a framework of a truss structure is set as the style image. Transferred images are obtained after processing the style transfer neural network, then the geometric images are translated into floor plans for new structure design. After the selection of the tilt angle and the degree of density, vertical components that connecting two adjacent layers are generated to be the pillars of the structure. At this stage, 2D style transferred images are successfully transformed into 3D geometries, which can be applied to the architectural design processes. Generally speaking, style transfer is an intelligent design tool that provides architects with a variety of choices of idea-generating. It has the potential to inspire architects at an early stage of design with not only 2D but also 3D format.
Proceedings of the 38th International Conference on Education and research in Computer Aided Architectural Design in Europe (eCAADe), 2020
Can a virtual city game be built by both the public and computer-based on real-site data? In the ... more Can a virtual city game be built by both the public and computer-based on real-site data? In the current process of deepening global connectivity, requirements for an effective urban design are no longer limited to functions or aesthetics, but a smart, dynamic complex with multi-interactions of data, group behaviours, and physical space. This paper introduces the logic of swarm intelligence and particle system for proposing a new urban design methodology. The platforms range from simulations that quantify the impact of the disruptive interventions of city activities to communicable collaboration between different users in a UI system, which creates virtual connections between optimized urbanscape and users. In the design system, based on the context data, the computer firstly simulates and optimizes the existing 2D activity joints between the people and analyzed the current spatial connection nodes into certain design rules. Through optimal programming for spatial connection and data iterations, the activity connection structures in the second simulation are abstracted into a set of interactive 3D topographic. The final data-visualization results are presented as a co-building megacity in a virtual construction game. Users can choose the virtual building unit types and intuitively influence the future urbanscape decision through virtual construction.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
Machine Learning, a recently prevalent research domain in data prediction and analysis, has been ... more Machine Learning, a recently prevalent research domain in data prediction and analysis, has been widely used in a variety of fields. In the design field, especially for architectural design, a machine learning method to learn and generate design data as pixelized images has been developed in previous researches. However, proceeding pixelized image data will cause the problems of precision loss and calculation waste, since the geometric architectural design data is efficiently stored and presented as vectorized CAD files. Thus, in this article, the author developed a specific machine learning neural network to learn and predict design drawings as vectorized data, speeding up the learning and predicting process, while improving the accuracy. First, two necessary geometric tests have been successfully done, which shows the central concept of neural network construct. Then, a design rule prediction model was built to demonstrate the methods to optimize the neural network and data structure. Lastly, a generation model based on human-made design data was constructed, which can be used to predict and generate the bedroom furniture positions by inputting the boundary data of the room, door, and window.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
When drawing architectural plans, designers should always define every detail, so the images can ... more When drawing architectural plans, designers should always define every detail, so the images can contain enough information to support design. This process usually costs much time in the early design stage when the design boundary has not been finally determined. Thus the designers spend a lot of time working forward and backward drawing sketches for different site conditions. Meanwhile, Machine Learning, as a decision-making tool, has been widely used in many fields. Generative Adversarial Network (GAN) is a model frame in machine learning, specially designed to learn and generate image data. Therefore, this research aims to apply GAN in creating architectural plan drawings, helping designers automatically generate the predicted details of apartment floor plans with given boundaries. Through the machine learning of image pairs that show the boundary and the details of plan drawings, the learning program will build a model to learn the connections between two given images, and then the evaluation program will generate architectural drawings according to the inputted boundary images. This automatic design tool can help release the heavy load of architects in the early design stage, quickly providing a preview of design solutions for architectural plans.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
Simulated Annealing is an artificial intelligence algorithm for finding the optimal solution of a... more Simulated Annealing is an artificial intelligence algorithm for finding the optimal solution of a proposition in an ample search space, which is based on the similarity between the physical annealing process of solid materials and the combinatorial optimization problem. In architectural layout design, although architects usually rely on their subjective design concepts to arrange buildings in a site, the judging criteria hidden in their design concepts are understandable. They can be summarized and parameterized as a combination of penalty and reward functions. By defining the functions to evaluate a design plan, then using the simulated annealing algorithm to search the optimal solution, the plan can be optimized and generated automatically. Six penalty and reward functions are proposed with different parameter weights in this article, which become a guideline for architectural layout design, especially for residential area planning. Then the results of several tests are shown, in which the parameter weights are adjusted, and the importance of each function is integrated. Lastly, a recommended weight and "temperature" setting are proposed, and a system of generating architectural layout is invented, which releases architects from building arranging work in an early stage.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
In the architecture field, humans have mastered various skills for creating unique spatial experi... more In the architecture field, humans have mastered various skills for creating unique spatial experiences with unknown interplays between known contents and styles. Meanwhile, machine learning, as a popular tool for mapping different input factors and generating unpredictable outputs, links the similarity of the machine intelligence with the typical form-finding process. Style Transfer, therefore, is widely used in 2D visuals for mixing styles while inspiring the architecture field with new form-finding possibilities. Researchers have applied the algorithm in generating 2D renderings of buildings, limiting the results in 2D pixels rather than real full volume forms. Therefore, this paper aims to develop a voxel-based form generation methodology to extend the 3D architectural application of Style Transfer. Briefly, through cutting the original 3D model into multiple plans and apply them to the 2D style image, the stylized 2D results generated by Style Transfer are then abstracted and filtered as groups of pixel points in space. By adjusting the feature parameters with user customization and replacing pixel points with basic voxelization units, designers can easily recreate the original 3D geometries into different design styles, which proposes an intelligent way of finding new and inspiring 3D forms.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
When drawing urban scale plans, designers should always define the position and the shape of each... more When drawing urban scale plans, designers should always define the position and the shape of each building. This process usually costs much time in the early design stage when the condition of a city has not been finally determined. Thus the designers spend a lot of time working forward and backward drawing sketches for different characteristics of cities. Meanwhile, machine learning, as a decision-making tool, has been widely used in many fields. Generative Adversarial Network (GAN) is a model frame in machine learning, specially designed to learn and generate image data. Therefore, this research aims to apply GAN in creating urban design plans, helping designers automatically generate the predicted details of buildings configuration with a given condition of cities. Through the machine learning of image pairs, the result shows the relationship between the site conditions (roads, green lands, and rivers) and the configuration of buildings. This automatic design tool can help release the heavy load of urban designers in the early design stage, quickly providing a preview of design solutions for urban design tasks. The analysis of different machine learning models trained by the data from different cities inspires urban designers with design strategies and features in distinct conditions.
Proceedings of the 1st International Conference on Computational Design and Robotic Fabrication (CDRF), 2019
3D Graphic Statics (3DGS) is a geometry-based structural design and analysis method, helping desi... more 3D Graphic Statics (3DGS) is a geometry-based structural design and analysis method, helping designers to generate 3D polyhedral forms by manipulating force diagrams with given boundary conditions. By subdividing 3D force diagrams with different rules, a variety of forms can be generated, resulting in more members with shorter lengths and richer overall complexity in forms. However, it is hard to evaluate the preference toward different forms from the aspect of aesthetics, especially for a specific architect with his own scene of beauty and taste of forms. Therefore, this article proposes a method to quantify the design preference of forms using machine learning and find the form with the highest score based on the result of the preference test from the architect. A dataset of forms was firstly generated, then the architect was asked to keep picking a favorite form from a set of forms several times in order to record the preference. After being trained with the test result, the neural network can evaluate a new inputted form with a score from 0 to 1, indicating the predicted preference of the architect, showing the possibility of using machine learning to quantitatively evaluate personal design taste.
Proceedings of the 18th International Conference on Computer-Aided Architectural Design Futures (CAAD Futures), 2019
With the rapid development of parametric design, Grasshopper, as a visual programming tool for ar... more With the rapid development of parametric design, Grasshopper, as a visual programming tool for architects, has been widely used. However, although Grasshopper is powerful for data processing, there is a weakness that the data only flows linearly from the first component to the last component, which means it's impossible to update the data iteratively by loop structure in native Grasshopper. So here, we introduce a Python based scripting plug-in Decodes, adding the function of loop construct into Grasshopper while integrating the basic graphical operations with faster mathematical matrix calculation. What's more, in order to bring Decodes into play as far as possible, four iterative patterns are researched and designed through Decodes scripting, demonstrating the strength and necessity of loop construct. The patterns include iterative subdivision patterns (center tiling and pinwheel tiling) and iterative growing patterns (semi-regular tiling and swarm behavior). Also, the core parts of their codes are revealed and deciphered in this article. 1 Grasshopper and its Data Flow Grasshopper is a visual programming plug-in, developed by David Rutten in McNeel and running in Rhino, whose aim at first was to record and visualize the operation history in Rhino. Although Rhino does have a command called 'history' to help users to check the operation history, there is no parameters recorded after the operation. Users cannot see the history tree, and the parameters inputted when using this command cannot be modified once it is determined. The invention of Grasshopper solved this problem. It records each operation history with a component (battery), which is visual, reusable, and modifiable. When we make connections between components, we are actually building a history tree, recording each operation, in order to view the contents of previous components.
Proceedings of the 24th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2019
Additive manufacturing has widely been spread in the digital fabrication and design fields, allow... more Additive manufacturing has widely been spread in the digital fabrication and design fields, allowing designers to rapidly manufacture complex geometry. In the additive process of Fused Deposition Modelling (FDM), machine movements are provided in the form of Gcode - A language of spatial coordinates controlling the position of the 3D printing extruder. Slicing software use closed mesh models to create Gcode from planar contours of the imported mesh, which raises limitations in the geometry types accepted by slicing software as well as machine control freedom. This paper presents a framework that makes full use of three degrees of freedom of Computer Numerically Controlled (CNC) machines through the generation of Gcode in the Rhino and Grasshopper environment. Eliminating the need for slicing software, Gcode files are generated through user-defined toolpaths that allow for higher levels of control over the CNC machine and a wider range of possibilities for non-conventional 3D printing applications. Here, we present Caterpillar, a Grasshopper plug-in providing architects and designers with high degrees of customizability for additive manufacturing. Core codes are revealed, application examples of printing with user-defined toolpaths are shown.
Proceedings of the 38th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), 2018
With the development of information technology, the ideas of programming and mass calculation wer... more With the development of information technology, the ideas of programming and mass calculation were introduced into the design field, resulting in the growth of computer aided design. With the idea of designing by data, we began to manipulate data directly, and interpret data through design works. Machine Learning as a decision making tool has been widely used in many fields. It can be used to analyze large amounts of data and predict future changes. Generative Adversarial Network (GAN) is a model framework in machine learning. It's specially designed to learn and generate output data with similar or identical characteristics. Pix2pixHD is a modified version of GAN that learns image data in pairs and generates new images based on the input. The author applied pix2pixHD in recognizing and generating architectural drawings, marking rooms with different colors and then generating apartment plans through two convolutional neural networks. Next, in order to understand how these networks work, the author analyzed their framework, and provided an explanation of the three working principles of the networks, convolution layer, residual network layer and deconvolution layer. Lastly, in order to visualize the networks in architectural drawings, the author derived data from different layer and different training epochs, and visualized the findings as gray scale images. It was found that the features of the architectural plan drawings have been gradually learned and stored as parameters in the networks. As the networks get deeper and the training epoch increases, the features in the graph become more concise and clearer. This phenomenon may be inspiring in understanding the designing behavior of humans.
Learning, Prototyping and Adapting, Short Paper Proceedings of the 23rd International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2018
3D printing is becoming increasingly popular for the manufacturing of small artworks and prototyp... more 3D printing is becoming increasingly popular for the manufacturing of small artworks and prototypes. For the further development of this technology, the authors of this paper drew inspiration from nature to envision biologically inspired 3D printing techniques. Many insects, for example, have developed fascinating fabrication methods to build cocoons and webs by secreting and spinning silk fibers. In some ways, 3D printers and silk-producing insects share comparable extrusion mechanisms to build intricate structures. Both have a moving frame or supporting exoskeletons, extruder heads or glands, and a control or nervous system. Based on this analogy, it is possible to use 3D printers to mimic the construction processes of insects and develop a bio-inspired fabrication strategy. To provide 3D printers with a greater ability for free and controlled movement, a new method for conformal GCode generation was developed. Built as a plug-in for Grasshopper, the program automatically translates 3D curves into custom GCode. To better demonstrate this technique, the authors conducted four experiments. The first example was inspired by the spawning behavior of dragonflies and was used to enable a 3D printer operate below its printing platform. The second example drew inspiration from the spawning behavior of mosquitos and focused on printing filament onto the surface of water. The third example was inspired by spiders and led to a method that allowed the 3D printer to spray continuous filament in mid-air. The last example was inspired by weaver ants that are capable of placing fibers on top of fibrous sub surfaces.
Proceedings of the 41st Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), 2021
This research aims to explore the quantitative relationship between urban planning decisions and ... more This research aims to explore the quantitative relationship between urban planning decisions and the health status of residents. By modeling the Point of Interest (POI) data and the geographic distribution of health-related outcomes, the research explores the critical factors in urban planning that could influence the health status of residents. It also informs decision-making regarding a healthier built environment and opens up possibilities for other data-driven methods. The data source constitutes two data sets, the POI data from OpenStreetMap, and the CDC dataset PLACES: Local Data for Better Health. After the data is collected and joined spatially, a machine learning method is used to select the most critical urban features in predicting the health outcomes of residents. Several machine learning models are trained and compared. With the chosen model, the prediction is evaluated on the test dataset and mapped geographically. The relations between factors are explored and interpreted. Finally, to understand the implications for urban design, the impact of modified POI data on the prediction of residents' health status is calculated and compared. This research proves the possibility of predicting residents' health from urban conditions with machine learning methods. The result verifies existing healthy urban design theories from a different perspective. This approach shows vast potential that data could in future assist decision-making to achieve a healthier built environment.
Proceedings of the 27th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2022
The regeneration of the industrial waterfront is a global issue, and its significance lies in tra... more The regeneration of the industrial waterfront is a global issue, and its significance lies in transforming the waterfront brownfield into an eco-friendly, hospitable, and vibrant urban space. However, the industrial waterfront naturally has comparatively unmanageable morphological features, including linear shape, irregular waterfront boundary, and separation with urban networks. Therefore, how to subdivide the vacant land and determine the land-use type for each subdivision becomes a challenging problem. Accordingly, this study proposes an application of machine learning models. It allows the generation of morphological elements of the vacant industrial waterfront by comparing the before-and-after scenarios of successful regeneration projects. The data collected from New York City is used as a showcase of this method.
Proceedings of the 3rd International Conference on Computational Design and Robotic Fabrication (CDRF), 2021
Health environment is a key factor in public health. Since people's health depends largely on the... more Health environment is a key factor in public health. Since people's health depends largely on their lifestyle, the built environment which supports a healthy living style is becoming more important. With the right urban planning decisions, it's possible to encourage healthier living and save healthcare expenditures for the society. However, there is not yet a quantitative relationship established between urban planning decisions and the health status of the residents. With the abundance of data and computing resources, this research aims to explore this relationship with a machine learning method. The data source is from both the OpenStreetMap and American Center for Decease Control and Prevention (CDC). By modeling the Point of Interest data and the geographic distribution of health-related outcome, the research explores the key factors in urban planning that could influence the health status of the residents quantitatively. It informs how to create a built environment that supports health and opens up possibilities for other data-driven methods in this field.
Proceedings of International Association for Shell and Spatial Structures Annual Symposia (IASS), 2021
This research investigates the use of graphic statics in analyzing the structural geometry of a n... more This research investigates the use of graphic statics in analyzing the structural geometry of a natural phenomenon to understand its performance and its relevant design parameters. Nature has always been the source of inspiration for designers, engineers, and scientists. Structural systems in nature are constantly evolving to optimize themselves with their boundary conditions. This optimization follows certain design rules that are quite challenging for a human to formulate or even comprehend. A dragonfly wing is an instance of a high-performance, lightweight structure that has intrigued many researchers to investigate its geometry and its performance as one of the most light-weight structures designed by nature. There are extensive geometrical and analytical studies on the pattern of the wing, but the underlying design logic is not clear. The geometry of the internal members of the dragonfly wings mainly consists of convex cell which may represent a compression-only network on a 2D plane. However, this property has not been geometrically analyzed from this perspective to confirm the hypothesis. In this research, we use the methods of 2D graphic statics to construct the force diagram from the given structural geometry of the wing. We use algebraic and iterative methods to report the topological and geometric properties of the form and force diagrams such as the degrees of indeterminacies of the network. For sample wings, we separate the internal and the boundary edges, construct the force diagram, and finally reconstruct the structural forms. Comparing the magnitude of the forces of the reconstructed network with the actual structure of the wing using the edge lengths of the force diagram will shed light on the performance of the structure. Multiple analytical studies will be provided to compare the results in both synthetic and natural networks and drive solid conclusions. The success in predicting the force flow in the natural structural pattern using graphic statics will expand the use of these powerful methods in reproducing the similar geometry of the natural structural system for the use in many engineering and scientific problems. It will also ultimately help us understand the design parameters and boundary conditions for which nature produces its master-pieces.
Proceedings of the 40th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), 2020
In this paper, we propose a geometry-based generative design method to generate and optimize a fl... more In this paper, we propose a geometry-based generative design method to generate and optimize a floor structure with funicular building members. This method challenges the antiquated column system, which has been used for more than a century. By inputting the floor plan with the positions of columns, designers can generate a variety of funicular supporting structures, expanding the choice of floor structure designs beyond simply columns and beams and encouraging the creation of architectural spaces with more diverse design elements. We further apply machine learning techniques (artificial neural networks) to evaluate and optimize the structural performance and constructability of the funicular structure, thus finding the optimal solutions within the almost infinite solution space. To achieve this, a machine learning model is trained and used as a fast evaluator to help the evolutionary algorithm find the optimal designs. This interdisciplinary method combines computer science and structural design, providing flexible design choices for generating floor structures.
Proceedings of the 40th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), 2020
The growing popularity of machine learning has provided new opportunities to predict certain beha... more The growing popularity of machine learning has provided new opportunities to predict certain behaviors precisely by utilizing big data. In this research, we use an image-based neural network to explore the relationship between the built environment and the activity of bicyclists in that environment. The generative model can produce heat maps that can be used to predict quantitatively the cycling and running activity in a given area, and then use urban design to enhance urban vitality in that area. In the machine learning model, the input image is a plan view of the built environment, and the output image is a heat map showing certain activities in the corresponding area. After it is trained, the model yields output (the predicted heat map) at an acceptable level of accuracy. The heat map shows the levels and conditions of the subject activity in different sections of the built environment. Thus, the predicted results can help identify where regional vitality can be improved. Using this method, designers can not only predict the behavioral heat distribution but also examine the different interactions between behaviors and aspects of the environment. The extent to which factors might influence behaviors also is studied by generating a heat map of the modified plan. Besides the potential applications of this approach, its limitations and areas for improvement also are proposed.
Proceedings of the 26th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2021
The free-form ice shell is the most challenging type in the design and construction of free-form ... more The free-form ice shell is the most challenging type in the design and construction of free-form buildings due to the regional and temporary nature of its materials. This paper presents a case study of the integration of design and fabrication of free-form ice shell. Taking the computational design and robotic fabrication of the ice shell as the main object, we discuss that combines the form-finding of the shell structure of graphic static with the tessellation technology of stereotomy, and propose a new method and workflow for the integration of discrete free-form ice shell design and construction. In the end, we built a free-form ice shell consisting of 116 discrete ice blocks. Practice has proved the feasibility of the integrated method of discrete free-form ice shell design and construction in the article.
Proceedings of the 26th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2021
Street vending is a recent policy advocated by city governments to support small and intermediate... more Street vending is a recent policy advocated by city governments to support small and intermediate businesses in the post-pandemic period in China. Street vendors select their locations primarily based on their intuitions about the surrounding environment; they temporarily occupy popular locations that benefit their business. Taking the city of Chengdu as an example, this study aims to formulate the rules governing vendors' location selection using machine learning and big data analysis techniques, thus identifying streets likely to become vital street markets. We propose a semantic segmentation method to construct heat maps that visualize and quantify the distribution of street vendors and pedestrians on public urban streets. The image-based generative adversarial network (GAN) is then trained to predict the vendors' heat maps from the pedestrians' heat map, finding the relationship between the locations of the vendors and the pedestrians. Our successful prediction of the vendors' locations highlights machine learning techniques' ability to quantify experience-based decision strategies. Moreover, suggesting potential marketing locations to vendors could help increase cities' vitality.
Style transfer is a design technique that is based on Artificial Intelligence and Machine Learnin... more Style transfer is a design technique that is based on Artificial Intelligence and Machine Learning, which is an innovative way to generate new images with the intervention of style images. The output image will carry the characteristic of style image and maintain the content of the input image. However, the design technique is employed in generating 2D images, which has a limited range in practical use. Thus, the goal of the project is to utilize style transfer as a toolset for architectural design and find out the possibility for a 3D modeling design. To implement style transfer into the research, floor plans of different heights are selected from a given design boundary and set as the content images, while a framework of a truss structure is set as the style image. Transferred images are obtained after processing the style transfer neural network, then the geometric images are translated into floor plans for new structure design. After the selection of the tilt angle and the degree of density, vertical components that connecting two adjacent layers are generated to be the pillars of the structure. At this stage, 2D style transferred images are successfully transformed into 3D geometries, which can be applied to the architectural design processes. Generally speaking, style transfer is an intelligent design tool that provides architects with a variety of choices of idea-generating. It has the potential to inspire architects at an early stage of design with not only 2D but also 3D format.
Proceedings of the 38th International Conference on Education and research in Computer Aided Architectural Design in Europe (eCAADe), 2020
Can a virtual city game be built by both the public and computer-based on real-site data? In the ... more Can a virtual city game be built by both the public and computer-based on real-site data? In the current process of deepening global connectivity, requirements for an effective urban design are no longer limited to functions or aesthetics, but a smart, dynamic complex with multi-interactions of data, group behaviours, and physical space. This paper introduces the logic of swarm intelligence and particle system for proposing a new urban design methodology. The platforms range from simulations that quantify the impact of the disruptive interventions of city activities to communicable collaboration between different users in a UI system, which creates virtual connections between optimized urbanscape and users. In the design system, based on the context data, the computer firstly simulates and optimizes the existing 2D activity joints between the people and analyzed the current spatial connection nodes into certain design rules. Through optimal programming for spatial connection and data iterations, the activity connection structures in the second simulation are abstracted into a set of interactive 3D topographic. The final data-visualization results are presented as a co-building megacity in a virtual construction game. Users can choose the virtual building unit types and intuitively influence the future urbanscape decision through virtual construction.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
Machine Learning, a recently prevalent research domain in data prediction and analysis, has been ... more Machine Learning, a recently prevalent research domain in data prediction and analysis, has been widely used in a variety of fields. In the design field, especially for architectural design, a machine learning method to learn and generate design data as pixelized images has been developed in previous researches. However, proceeding pixelized image data will cause the problems of precision loss and calculation waste, since the geometric architectural design data is efficiently stored and presented as vectorized CAD files. Thus, in this article, the author developed a specific machine learning neural network to learn and predict design drawings as vectorized data, speeding up the learning and predicting process, while improving the accuracy. First, two necessary geometric tests have been successfully done, which shows the central concept of neural network construct. Then, a design rule prediction model was built to demonstrate the methods to optimize the neural network and data structure. Lastly, a generation model based on human-made design data was constructed, which can be used to predict and generate the bedroom furniture positions by inputting the boundary data of the room, door, and window.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
When drawing architectural plans, designers should always define every detail, so the images can ... more When drawing architectural plans, designers should always define every detail, so the images can contain enough information to support design. This process usually costs much time in the early design stage when the design boundary has not been finally determined. Thus the designers spend a lot of time working forward and backward drawing sketches for different site conditions. Meanwhile, Machine Learning, as a decision-making tool, has been widely used in many fields. Generative Adversarial Network (GAN) is a model frame in machine learning, specially designed to learn and generate image data. Therefore, this research aims to apply GAN in creating architectural plan drawings, helping designers automatically generate the predicted details of apartment floor plans with given boundaries. Through the machine learning of image pairs that show the boundary and the details of plan drawings, the learning program will build a model to learn the connections between two given images, and then the evaluation program will generate architectural drawings according to the inputted boundary images. This automatic design tool can help release the heavy load of architects in the early design stage, quickly providing a preview of design solutions for architectural plans.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
Simulated Annealing is an artificial intelligence algorithm for finding the optimal solution of a... more Simulated Annealing is an artificial intelligence algorithm for finding the optimal solution of a proposition in an ample search space, which is based on the similarity between the physical annealing process of solid materials and the combinatorial optimization problem. In architectural layout design, although architects usually rely on their subjective design concepts to arrange buildings in a site, the judging criteria hidden in their design concepts are understandable. They can be summarized and parameterized as a combination of penalty and reward functions. By defining the functions to evaluate a design plan, then using the simulated annealing algorithm to search the optimal solution, the plan can be optimized and generated automatically. Six penalty and reward functions are proposed with different parameter weights in this article, which become a guideline for architectural layout design, especially for residential area planning. Then the results of several tests are shown, in which the parameter weights are adjusted, and the importance of each function is integrated. Lastly, a recommended weight and "temperature" setting are proposed, and a system of generating architectural layout is invented, which releases architects from building arranging work in an early stage.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
In the architecture field, humans have mastered various skills for creating unique spatial experi... more In the architecture field, humans have mastered various skills for creating unique spatial experiences with unknown interplays between known contents and styles. Meanwhile, machine learning, as a popular tool for mapping different input factors and generating unpredictable outputs, links the similarity of the machine intelligence with the typical form-finding process. Style Transfer, therefore, is widely used in 2D visuals for mixing styles while inspiring the architecture field with new form-finding possibilities. Researchers have applied the algorithm in generating 2D renderings of buildings, limiting the results in 2D pixels rather than real full volume forms. Therefore, this paper aims to develop a voxel-based form generation methodology to extend the 3D architectural application of Style Transfer. Briefly, through cutting the original 3D model into multiple plans and apply them to the 2D style image, the stylized 2D results generated by Style Transfer are then abstracted and filtered as groups of pixel points in space. By adjusting the feature parameters with user customization and replacing pixel points with basic voxelization units, designers can easily recreate the original 3D geometries into different design styles, which proposes an intelligent way of finding new and inspiring 3D forms.
Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2020
When drawing urban scale plans, designers should always define the position and the shape of each... more When drawing urban scale plans, designers should always define the position and the shape of each building. This process usually costs much time in the early design stage when the condition of a city has not been finally determined. Thus the designers spend a lot of time working forward and backward drawing sketches for different characteristics of cities. Meanwhile, machine learning, as a decision-making tool, has been widely used in many fields. Generative Adversarial Network (GAN) is a model frame in machine learning, specially designed to learn and generate image data. Therefore, this research aims to apply GAN in creating urban design plans, helping designers automatically generate the predicted details of buildings configuration with a given condition of cities. Through the machine learning of image pairs, the result shows the relationship between the site conditions (roads, green lands, and rivers) and the configuration of buildings. This automatic design tool can help release the heavy load of urban designers in the early design stage, quickly providing a preview of design solutions for urban design tasks. The analysis of different machine learning models trained by the data from different cities inspires urban designers with design strategies and features in distinct conditions.
Proceedings of the 1st International Conference on Computational Design and Robotic Fabrication (CDRF), 2019
3D Graphic Statics (3DGS) is a geometry-based structural design and analysis method, helping desi... more 3D Graphic Statics (3DGS) is a geometry-based structural design and analysis method, helping designers to generate 3D polyhedral forms by manipulating force diagrams with given boundary conditions. By subdividing 3D force diagrams with different rules, a variety of forms can be generated, resulting in more members with shorter lengths and richer overall complexity in forms. However, it is hard to evaluate the preference toward different forms from the aspect of aesthetics, especially for a specific architect with his own scene of beauty and taste of forms. Therefore, this article proposes a method to quantify the design preference of forms using machine learning and find the form with the highest score based on the result of the preference test from the architect. A dataset of forms was firstly generated, then the architect was asked to keep picking a favorite form from a set of forms several times in order to record the preference. After being trained with the test result, the neural network can evaluate a new inputted form with a score from 0 to 1, indicating the predicted preference of the architect, showing the possibility of using machine learning to quantitatively evaluate personal design taste.
Proceedings of the 18th International Conference on Computer-Aided Architectural Design Futures (CAAD Futures), 2019
With the rapid development of parametric design, Grasshopper, as a visual programming tool for ar... more With the rapid development of parametric design, Grasshopper, as a visual programming tool for architects, has been widely used. However, although Grasshopper is powerful for data processing, there is a weakness that the data only flows linearly from the first component to the last component, which means it's impossible to update the data iteratively by loop structure in native Grasshopper. So here, we introduce a Python based scripting plug-in Decodes, adding the function of loop construct into Grasshopper while integrating the basic graphical operations with faster mathematical matrix calculation. What's more, in order to bring Decodes into play as far as possible, four iterative patterns are researched and designed through Decodes scripting, demonstrating the strength and necessity of loop construct. The patterns include iterative subdivision patterns (center tiling and pinwheel tiling) and iterative growing patterns (semi-regular tiling and swarm behavior). Also, the core parts of their codes are revealed and deciphered in this article. 1 Grasshopper and its Data Flow Grasshopper is a visual programming plug-in, developed by David Rutten in McNeel and running in Rhino, whose aim at first was to record and visualize the operation history in Rhino. Although Rhino does have a command called 'history' to help users to check the operation history, there is no parameters recorded after the operation. Users cannot see the history tree, and the parameters inputted when using this command cannot be modified once it is determined. The invention of Grasshopper solved this problem. It records each operation history with a component (battery), which is visual, reusable, and modifiable. When we make connections between components, we are actually building a history tree, recording each operation, in order to view the contents of previous components.
Proceedings of the 24th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2019
Additive manufacturing has widely been spread in the digital fabrication and design fields, allow... more Additive manufacturing has widely been spread in the digital fabrication and design fields, allowing designers to rapidly manufacture complex geometry. In the additive process of Fused Deposition Modelling (FDM), machine movements are provided in the form of Gcode - A language of spatial coordinates controlling the position of the 3D printing extruder. Slicing software use closed mesh models to create Gcode from planar contours of the imported mesh, which raises limitations in the geometry types accepted by slicing software as well as machine control freedom. This paper presents a framework that makes full use of three degrees of freedom of Computer Numerically Controlled (CNC) machines through the generation of Gcode in the Rhino and Grasshopper environment. Eliminating the need for slicing software, Gcode files are generated through user-defined toolpaths that allow for higher levels of control over the CNC machine and a wider range of possibilities for non-conventional 3D printing applications. Here, we present Caterpillar, a Grasshopper plug-in providing architects and designers with high degrees of customizability for additive manufacturing. Core codes are revealed, application examples of printing with user-defined toolpaths are shown.
Proceedings of the 38th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), 2018
With the development of information technology, the ideas of programming and mass calculation wer... more With the development of information technology, the ideas of programming and mass calculation were introduced into the design field, resulting in the growth of computer aided design. With the idea of designing by data, we began to manipulate data directly, and interpret data through design works. Machine Learning as a decision making tool has been widely used in many fields. It can be used to analyze large amounts of data and predict future changes. Generative Adversarial Network (GAN) is a model framework in machine learning. It's specially designed to learn and generate output data with similar or identical characteristics. Pix2pixHD is a modified version of GAN that learns image data in pairs and generates new images based on the input. The author applied pix2pixHD in recognizing and generating architectural drawings, marking rooms with different colors and then generating apartment plans through two convolutional neural networks. Next, in order to understand how these networks work, the author analyzed their framework, and provided an explanation of the three working principles of the networks, convolution layer, residual network layer and deconvolution layer. Lastly, in order to visualize the networks in architectural drawings, the author derived data from different layer and different training epochs, and visualized the findings as gray scale images. It was found that the features of the architectural plan drawings have been gradually learned and stored as parameters in the networks. As the networks get deeper and the training epoch increases, the features in the graph become more concise and clearer. This phenomenon may be inspiring in understanding the designing behavior of humans.
Learning, Prototyping and Adapting, Short Paper Proceedings of the 23rd International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), 2018
3D printing is becoming increasingly popular for the manufacturing of small artworks and prototyp... more 3D printing is becoming increasingly popular for the manufacturing of small artworks and prototypes. For the further development of this technology, the authors of this paper drew inspiration from nature to envision biologically inspired 3D printing techniques. Many insects, for example, have developed fascinating fabrication methods to build cocoons and webs by secreting and spinning silk fibers. In some ways, 3D printers and silk-producing insects share comparable extrusion mechanisms to build intricate structures. Both have a moving frame or supporting exoskeletons, extruder heads or glands, and a control or nervous system. Based on this analogy, it is possible to use 3D printers to mimic the construction processes of insects and develop a bio-inspired fabrication strategy. To provide 3D printers with a greater ability for free and controlled movement, a new method for conformal GCode generation was developed. Built as a plug-in for Grasshopper, the program automatically translates 3D curves into custom GCode. To better demonstrate this technique, the authors conducted four experiments. The first example was inspired by the spawning behavior of dragonflies and was used to enable a 3D printer operate below its printing platform. The second example drew inspiration from the spawning behavior of mosquitos and focused on printing filament onto the surface of water. The third example was inspired by spiders and led to a method that allowed the 3D printer to spray continuous filament in mid-air. The last example was inspired by weaver ants that are capable of placing fibers on top of fibrous sub surfaces.
Uploads
Papers by Hao Zheng