Welcome to CEE595: AI in Construction Seminar Series!! Thanks to the rapid progress in the field of Artificial Intelligence (AI) around Machine Learning, Computer Vision and Natural Language Processing, many discussions have sparked in the past few years around the role and the impact of AI in the Architecture, Engineering, Construction (AEC) industry. While advancements are made, many fundamental questions about developing and leveraging AI technology in various design, construction and operation workflows in the built environment have remained unexplored. In this first-ever seminar series, we plan to explore the readiness of the current AI methods and solutions for the AEC industry. We will host a range of leading industry and academic speakers to seek clarity around three main areas:
Learning about the latest AI technologies that are proactively used in today's practice.
What startup and technology companies are active in offering solutions around various workflows involved in the design, engineering, construction and operation of the built environment?
Are recent academic works competing with these industry-grade solutions? What are the areas in which academic research is leading the charge on breaking the ground on offering the next generation of AI solutions?
What are the potential areas to explore in the near-term? What is the value of solving these problems for the AEC industry?
What fundamental AI problems are unsolved? Are these problems tied to clear business opportunities and will they have impact on the AEC industry?
Differentiable Architecture
The discussion will identify how Artificial Intelligence has helped, is helping, and may help solve future problems in Architectural Design. In general, AI will be called to bridge the gap between architects' and design experts' domain knowledge and statistical patterns in large-scale data. We will discuss current AI-driven approaches for shape and scene modeling, address their advantages and limitations, and suggest potential future opportunities such as using explainable AI models in design or using AI to spark creativity.
Alberto Tono is the president of the Computational Design Institute. He served as the Research and Computational Design Leader in Architectural and Engineering organizations, receiving the O1-visa for outstanding abilities with HOK and HDR. Tono obtained his Masters in Building Engineering - Architecture from the University of Padua and the Harbin Institute of Technology. He has been working in the computational design and deep learning space since 2014. Furthermore, he is focused on the improvement of BIM/VDC workflows with a statistical approach. He is an international multi-award-winning “hacker” and speaker, and his work within Architecture and Artificial Intelligence brought him to companies in China, the Netherlands, Italy, and California.
As-Is Information
Automatically generating structured information on the as-is status of facilities from visual data
Despite technological advancements, physical reality and its digital representation are currently two disconnected domains. We are quickly transitioning into a hybrid world, where we will live with virtual and augmented reality, robots, and automation along physical reality. This requires a seamless interaction between the physical and the digital. A big challenge toward their interaction is an accurate, detailed, consistent, and up-to-date knowledge of the as-is status of facilities in the digital domain, structured in a representation that is interpretable by both humans and machines. State-of-the-art practice attempts to acquire the digital representation with manual, time consuming, non-scalable, and error-prone methods. This has a direct, negative, and quantifiable impact on how architects, engineers, constructors, and facility managers (AEC/FM) design, construct, and use facilities. It is a global and ubiquitous challenge since it applies to the entire built environment, and therefore requires automatic, fast, scalable, accurate, and generalizable solutions. My vision is the seamless interaction of the two domains; but what does this mean? In this presentation, I will discuss about my past work on automatically producing and structuring semantic information on objects (e.g., wall, floor, chair) and rooms within entire buildings using visual data, to facilitate AEC/FM processes.
Iro Armeni completed her PhD at Stanford University, conducting interdisciplinary research between Civil Engineering and Machine Vision. Her area of focus is on automated semantic and operational understanding of buildings throughout their life cycle using visual data, toward establishing a seamless interaction between the physical world and its digital representation. Prior to enrolling in the PhD program, Iro received an MSc in Computer Science (Ionian University-2013), an MEng in Architecture and Digital Design (University of Tokyo-2011), and a Diploma in Architectural Engineering (National Technical University of Athens-2009). She is the recipient of the 2019 ETHZ PostDoctoral Fellowship, the 2017 Google PhD Fellowship on Machine Perception, and the 2009 Japanese Government (MEXT) scholarship.
Leveraging existing knowledge and artificial intelligence to enhance our understanding of structural performance
Existing knowledge regarding conventional materials and structures under most loading conditions has been amassed over years via experimental tests and computational analyses; however, extreme loads are changing drastically, and new materials and structural designs are being created swiftly. To maintain pace with this innovation while continuing to provide robust and resilient structures, there is a need for a rapid and reliable approach to understanding the behavior of new materials and structural designs under new, extreme loads. Furthermore, in recent years, significant advancements have been made in artificial intelligence (AI). With the convergence of AI into the civil engineering domain comes the capability to learn the highly nonlinear, complex relationships between material, structural, and load characteristics and a structures’ behavior. Moreover, this can potentially be done with little human intervention, few, if any, new experimental tests, and with excellent computational efficiency and generalization accuracy. To help realize this long-term vision, it is critical to predict the performance of existing and new structural components and designs due to extreme loads, where large amounts of data are not available (i.e., small datasets). This presentation will focus on the delicate integration of physics-based knowledge (by way of existing datasets, empirical relations, etc.) and AI-based approaches.
AI has revolutionized every field it has touched, from lending money to security infrastructure, to advanced manufacturing and open gameplay, such as chess. Still, the result of this effort could be learning that, regardless of the abundant adjacent innovation, there is minimal potential for successful application of AI within the civil engineering domain. Alternatively, it could very likely cause revolutionary innovation in the civil engineering fields, with benefits to all civil infrastructure as well as to the taxpayers who fund that infrastructure. The AI models developed previously by Dr. Paal to predict the seismic performance of reinforced concrete structures will be presented in this seminar, and the potential for AI-based approaches to act as conduits for enhanced understanding of physical behavior will be illustrated through these successful approaches. This work not only illustrates the potential impact for further research efforts in this direction, but more importantly, demonstrates that direct observation and modeling of real-world structural performance via novel, AI-based, data-driven methodologies greatly reduces the uncertainty in predictions and can lead to enhanced, near-real-time understanding of the physical behavior of structures, specifically under seismic events.
Dr. Paal is a tenure-track Assistant Professor in the Zachry Department of Civil & Environmental Engineering at Texas A&M University. She joined the faculty at Texas A&M in the Fall of 2016 after completing a post-doctoral fellowship in the School of Architecture, Civil and Environmental Engineering at the Ecole Polytechnique Federale de Lausanne (EPFL) in Switzerland. She received her Master’s and Doctoral degrees in Civil Engineering from the Georgia Institute of Technology in Atlanta, Georga in 2011 and 2013, respectively. Additionally, she received a B.S. in Architectural Engineering from the University of Texas at Austin in 2009. Dr. Paal has extensive background knowledge and expertise in machine learning and machine vision and applications of these technologies in infrastructure and structural condition assessments and other infrastructure-related practices. Her research focus is on mitigating the effects of natural and man-made disasters on our built infrastructure by integrating traditional civil engineering practices with emerging techniques and technologies such as artificial intelligence, augmented reality, unmanned aerial systems, and additive manufacturing. Her current research interests are towards hybrid artificial intelligence-physics-based approaches, understanding the impact of integrating artificial intelligence models and methodologies in civil engineering design, analysis, and evaluation operations, and developing advanced modeling approaches grounded in real-world data. She teaches classes on ‘Structural Analysis’ and ‘Structural Dynamics’ at Texas A&M University, and introduced a new graduate course on ‘Machine Intelligence and Applications in Civil Engineering’ to the Civil Engineering graduate program. This course focuses on both machine vision and machine learning within the realm of civil engineering. Her research has been supported by numerous state and federal agencies such as the National Cooperative Highway Research Program, Texas Department of Transportation, the National Association of Home Builders, and the National Science Foundation. She has 18 journal papers published or under review, four published book chapters, and has been invited to speak regarding her research over 30 times nationally and internationally. She has served on numerous workshop organizing committees and chaired several conference sessions. In 2020, she was granted an NSF Early CAREER award for her research towards enhancing our understanding of our infrastructures performance under natural hazards by leveraging available experimental data and artificial intelligence.
Structured Architectural Modeling: Past, Current, and Future.
I will talk about the history of Structured Modeling techniques for architectural scenes. In the past, primitive detection and fitting via combinatorial optimization was the standard, which unfortunately did not lead to the impact. The advent of deep learning enables a new computational approach, which finally becomes robust enough for production. Lastly, I will talk about my vision on the future of structured architectural modeling techniques: what interesting directions are and how things will be in the next 5 years.
Bio: Dr. Yasutaka Furukawa is an associate professor in the School of Computing Science at Simon Fraser University (SFU). Prior to joining SFU in 2017, he was an assistant professor at Washington University in St. Louis USA, and a software engineer at Google. He completed his Ph.D. at Computer Science Department of the University of Illinois at Urbana-Champaign in 2008. Dr. Furukawa received the best student paper award at ECCV 2012, the NSF CAREER Award in 2015, CS-CAN Outstanding Young CS Researcher Award 2018, Google Faculty Research Awards in 2016, 2017, and 2018, and PAMI Longuet-Higgins prize in 2020.
Oct 8
12:30-1:30pm
Bringing Digital Twin Technology and Artificial Intelligence to the Civil Infrastructure Asset Management Community
Unmanned aerial vehicles (UAV) are now a viable option for augmenting bridge inspections. Utilising an integrated combination of a UAV and computer vision can decrease costs, expedite inspections and facilitate bridge access. Any such inspection must consider the design of the UAV, the choice of cameras, data acquisition, geometrical resolution, safety regulations and pilot protocols. The Placer River Trail Bridge in Alaska recently served as a test bed for a UAV inspection methodology that integrates these considerations. The end goal was to produce a three-dimensional (3D) model of the bridge using UAV-captured images and a hierarchical Dense Structure-from-Motion algorithm. To maximise the quality of the model and its benefits to inspectors, this goal guided UAV design and mission planning. The resulting inspection methodology integrates UAV design, data capture and data analysis together to provide an optimised 3D model. This model provides inspection documentation while enabling the monitoring of defects. The developed methodology is presented herein, as well as analyses of the 3D models. The results are compared against models generated through laser scanning. The findings demonstrate that the UAV inspection methodology provided superior 3D models with the accuracy to resolve defects and support the needs of infrastructure managers.
Dr. Ali Khaloo is CEO and co-founder of Aren.ai. He is an engineer and researcher with a diverse background in structural engineering focused on developing frameworks to combine robotics, computer vision and artificial intelligence to conduct a quantitative condition assessment of infrastructure systems. Expert in utilizing Unmanned Aerial Vehicles (UAV) and 3D imaging techniques for automated as-built modeling followed by quantitative geometric analysis of generated models. Research interests include robotics and autonomous condition assessment of structures, 3D data processing, feature extraction, artificial intelligence and machine learning applications in civil engineering, pattern recognition, computer vision, data mining, computational geometry, image and video processing, Augmented Reality (AR) and Virtual Reality (VR) for structural integrity assessment, photogrammetry, and remote sensing.
AI Research at Autodesk
We (`Dr. Mehdi Nourbakhsh and two members of his team`) are going to talk about Autodesk Research and how we work with companies and industry partners. We are also going to talk about one of the research projects we recently completed. The research project's goal was to use AI and machine learning to expedite the bidding process for DFMA contractors. We are planning to demo the prototype that we built in this project.
Dr. Mehdi Nourbakhsh is a Research Lead at Autodesk Research. He is passionate about bringing AI-assisted technologies to the hands of Autodesk customers. With his several years of experience as a structural engineer and construction manager, he leads a research team focusing on the applications of AI and ML in the future of design and construction. He holds a PhD in building construction and a master’s in computer science from the Georgia Institute of Technology.
3D Reconstruction in the Age of Recognition
Recent advances in sensing and computer vision make image-based 3D scene modeling a core technology for augmented reality, infrastructure inspection, and robotics. I will present the capabilities and limitations of mature multi-image techniques, as well as nascent recognition-based single-image techniques for 3D modeling. I will also describe how, in the future, these disparate techniques can be unified to provide complete, precise, geometric, and semantic scene modeling.
Dr. Derek Hoiem is Chief Science Officer and co-founder of Reconstruct. He is also an associate professor Computer Science at the University of Illinois at Urbana-Champaign, since January 2009. Derek received his PhD in Robotics from Carnegie Mellon University in 2007 and completed a postdoctoral fellowship at the Beckman Institute in 2008. Derek's primary research goal is to model the physical and semantic structure of the world, so computers can better understand scenes from images. In particular, he researches algorithms to interpret physical space from images and to relate objects to their environment and to each other. Example applications include creating 3D models of scenes and objects from one image, photorealistic rendering of object models into images, robot navigation, and creating and matching as-built 3D models of construction scenes to planned models. Derek has published dozens of papers and several patents, and his work has been recognized with awards including an ACM Doctoral Dissertation Award honorable mention, CVPR best paper award, Intel Early Career Faculty award, Sloan Fellowship, and PAMI Significant Young Researcher award. Derek Hoiem is also co-founder and CTO of Reconstruct, which visually documents construction sites, matching images to plans and analyzing productivity and risk for delay.
