Publications

30 must-read publications on digital and circular building

The European construction industry is facing significant challenges in sustainability, innovation adoption, and labour shortages, with almost half of its jobs in short supply. However, by accelerating the sector's digital and green transformation, we can significantly improve its competitiveness and resource efficiency and appeal to a new generation of highly skilled workers, thereby creating a more innovative, sustainable, and attractive industry.

At HumanTech, we have joined forces with seven other European-funded projects addressing these challenges in the Tech4EUConstruction cluster. In this article, we share 30 of our scientific publications, providing new insights into the scientific foundation of our project's innovations for researchers and industry professionals.

The science behind the Tech4EUConstruction cluster’s innovations

At the Tech4EUConstruction cluster, we aim to create a lasting impact by exchanging our expertise and technical innovations. This article explores the science behind the groundbreaking innovations we are developing in areas such as building renovation, sustainability monitoring, digital innovation technologies, energy efficiency, renewable energy, and materials & design.

What advancements in AI and robotics will shape the future of the construction industry? Check out the articles below to find out!

Topic 1: Building renovation and sustainability monitoring

1. Monitoring the sustainability of building renovation projects — A tailored Key Performance Indicator repository

This publication, developed by the InCUBE project, aims to assist in identifying suitable key performance indicators (KPIs) that can be used to assess the sustainability performance of buildings as they transition into zero-carbon, resource-efficient, and resilient structures.

2. Towards the digitalization and automation of circular and sustainable construction and demolition waste management – project RECONMATIC

This publication presents Reconmatic, a Horizon Europe Research and Innovation Action project that aims to develop novel tools, technologies, and methodologies that can contribute to multiple construction phases and project types or material and product life cycle stages.

3. Assessing the construction and demolition waste volume for a typical Mediterranean residential building

Released by Reconmatic, this study estimates the construction and demolition waste (CDW) produced by a typical multi-storey residential building in Greece, built in the mid-20th century, made of reinforced concrete and filling masonry walls. It also considers renovation procedures and presents challenges related to the disposal, recycling, and reuse potential of CDW types.

Topic 2: Digital innovation technology

4. From 3D surveying data to BIM to BEM: The InCUBE dataset

This paper introduces the InCUBE dataset, resulting from the project's activities, focused on unlocking the EU building renovation through integrated strategies and processes for efficient built-environment management (including the use of innovative renewable energy technologies and digitalisation). The dataset contains raw and processed data from an Italian demo site in Trento's Santa Chiara district, enabling multiple potential uses, investigations, and developments.

5. Introducing Noise for AirSim’s 3D LiDAR Sensor to Reduce the Sim2real Gap

In robotics, modeling sensor noise is important as it can affect the accuracy and reliability of a robot’s perception of its environment. It also allows for more accurate simulations of robotic systems, which can help improve their performance in real-world scenarios. The Beeyonders’ project proposes introducing a noise model for the 3D LiDAR (Light Detection and Ranging) sensor supported in AirSim to help the community develop more accurate, reliable, and cost-effective solutions.

6. Online Ergonomic Evaluation in Realistic Manual Material Handling Task: Proof of Concept

Work-related musculoskeletal disorders are a major cause of work-related injuries. To address this issue, work task ergonomic risk indices have been developed, but they are subjective and challenging to perform in real time. This work, released by Beeyonders, provides a technique to digitalize this process by developing an online algorithm to calculate the NIOSH index using inertial sensors, which can be easily integrated into the industrial environment.

7. REINCARNATE: Shaping a sustainable future in construction through digital innovation

The heart of the Reincarnate project is the Circular Potential Information Model (CP-IM), a digital platform designed to assess and enhance the recyclability of construction materials, products, and buildings. The CP-IM utilizes advanced technologies to revolutionize the handling of construction waste, turning it into valuable resources and reducing the sector's environmental footprint. Its features include digital tracing, material durability predictions, and CO2 reduction materials design, showcased in eleven European demonstration projects, highlighting significant reductions in construction waste and CO2 emissions.

8. Presenting SLAMD – A Sequential Learning Based Software for the Inverse Design of Sustainable Cementitious Materials

The composition of concrete has become more complex, especially with formulations aimed at reducing carbon footprint. Inverse Design techniques offer a solution by allowing for a comprehensive search to create new and improved concrete formulations. This publication introduces the concept of Inverse Design and demonstrates how an open-source app called SLAMD, developed by Reincarnate, provides necessary workflow steps to adapt it in the laboratory, lowering the barriers to its application.

9. Single Frame Semantic Segmentation Using Multi-Modal Spherical Images

In recent years, the research community has shown much interest in panoramic images that offer a 360º directional perspective. Multiple data modalities can be fed, and complementary characteristics can be utilised for more robust and rich scene interpretation based on semantic segmentation. In this study, HumanTech proposes a transformer-based cross-modal fusion architecture to bridge the gap between multi-modal fusion and omnidirectional scene perception.

10. U-RED: Unsupervised 3D Shape Retrieval and Deformation for Partial Point Clouds

In this paper, HumanTech proposes U-RED, an Unsupervised shape REtrieval and Deformation pipeline that takes an arbitrary object observation as input, typically captured by RGB images or scans, and jointly retrieves and deforms the geometrically similar CAD models from a pre-established database.

11. Annotation rules and classes for semantic segmentation of point clouds for digitalization of existing bridge structures

Germany needs to digitize its extensive bridge infrastructure using BIM due to political requirements. This transformation involves using point cloud data and exploring available open-source datasets and various approaches to semantic segmentation. HumanTech aims to bridge the gap between theoretical research on point cloud data and manual inspection by proposing a set of object-oriented classes for semantic segmentation in this study.

12. OPA-3D: Occlusion-Aware Pixel-Wise Aggregation for Monocular 3D Object Detection

Monocular 3D object detection has advanced with the use of pre-trained depth estimators for pseudo-LiDAR (Light Detection and Ranging) recovery. HumanTech proposes a method that jointly estimates dense scene depth, depth-bounding box residuals, and object-bounding boxes, enabling a two-stream detection of 3D objects.

13. Ontology-Based Semantic Labelling for RGB-D and Point Cloud Datasets

Deep learning applications have recently surged in the construction field. Supervised semantic segmentation of 2D or 3D data acquired from buildings requires using annotated data for training, validation, and testing. However, existing datasets lack a common convention and definitions based on construction ontologies. In this work, HumanTech presents a guideline for ontology-based semantic annotation of RGB-D and point cloud datasets, bridging the gap between deep learning and computer science.

14. When Machine Learning Meets Raft: How to Elect a Leader over a Network

The Raft consensus algorithm is widely used to keep data consistent across multiple distributed nodes by having a leader node coordinate operations. However, the system pauses during leader elections, which can happen if the leader fails or gets disconnected from other nodes. In this paper, Reconmatic explores using Machine Learning to monitor and classify the causes of these leader elections, aiming to reduce unnecessary elections and increase system availability.

15. Machine-learning-assisted classification of construction and demolition waste fragments using computer vision: Convolution versus extraction of selected features

Reconmatic has developed a machine-learning-assisted procedure for identifying construction and demolition waste (CDW) fragments using an RGB camera. This approach improves waste sorting efficiency and accuracy, promoting sustainable resource use and reducing environmental impact.

16. Review of Concepts for Construction and Demolition Waste and the Circular Economy

This paper, developed by Reconmatic, examines the classification and management of construction and demolition waste (CDW) and the concept of circular economy (CE) in the construction sector. Its findings can guide practical measures to enhance waste management and inform planning and decision-making for waste reduction and recovery.

17. On Using Hyperledger Fabric Over Networks: Ordering Phase Improvements

Blockchain is increasingly being used in various research disciplines, such as the Internet of Things (IoT) and Software Defined Networking (SDN). Hyperledger Fabric is a popular enterprise-grade blockchain framework known for ensuring transparency in secure communication. One of its key features is the three-phase transaction flow architecture. This study released by Reconmatic focuses on improving the ordering phase by proposing a mechanism for faster communication.

18. RoBétArmé Project: Human-robot collaborative construction system for shotcrete digitization and automation through advanced perception, cognition, mobility and additive manufacturing skills

The shortage and rising costs of skilled workers, along with the need for new infrastructure and the maintenance of ageing infrastructure, are driving an increasing demand for construction automation. The RoBétArmé project aims to revolutionize construction with advanced technologies for shotcrete (sprayed concrete) automation. This paper showcases a novel robotic system for automating all phases of shotcrete application.

19. Adaptive BIM/CIM for Digital Twining of Automated Shotcreting Process

Developing digital twins for construction requires accurately replicating real-world spaces. This study, developed by RoBétArmé, emphasizes the importance of using Building/Civil-Construction Information Modeling (BIM/CIM) to create digital twins for construction, particularly for automated shotcreting of civil infrastructure projects. It highlights the need for simulations, visualizations, and adaptive modeling to monitor and control assets in real time.

20. Leveraging Multimodal Sensing and Topometric Mapping for Human-Like Autonomous Navigation in Complex Environments

Autonomous vehicles need to understand complex outdoor environments and follow traffic rules. RoBétArmé's approach is to imitate human driver behaviour using RGB and LiDAR (Light Detection and Ranging) data combined with a rough topometric map for route planning. Their method shows potential for safer and more human-like autonomous behaviours in urban and semi-structured environments.

21. Cognitive Fusion-based Path Planning for UAV Inspection of Power Towers

The use of Unmanned Aerial Vehicles (UAV) for inspecting critical power infrastructure has advanced significantly in recent years. This paper by RoBétArmé presents a novel path planning method that leverages robot vision derived from LiDAR (Light Detection and Ranging) and RGB data for inspecting power tower insulators.

22. Comparative Study of Surface 3D Reconstruction Methods Applied in Construction Sites

This research from RoBétArmé provides a comprehensive assessment of key methodologies for 3D reconstruction of construction sites. It evaluates monocular and binocular computer vision techniques for their ability to extract detailed 3D surfaces while considering their computational efficiency. The findings contribute to advancing 3D reconstruction techniques in the construction industry, which is essential for its digitalization.

23. Real-time 3D Reconstruction Adapted for Robotic Applications in Construction Sites

Integrating robot vision techniques, especially focused on 3D reconstruction, in the construction industry is crucial to meeting the digitalization needs of Industry 4.0. This study from RoBétArmé introduces a real-time 3D reconstruction pipeline that uses both RGB and depth information using common algorithms.

Topic 3: Energy efficiency and renewable energy

24. Dynamic Energy Analysis of Different Heat Pump Heating Systems Exploiting Renewable Energy Sources

Renewable energy source-fed heat pumps (HPs) may perform up to very high efficiency standards, offering a promising tool in the broader residential heat decarbonization effort. In this context, this paper, by InCUBE, investigates different heating configurations using various renewable thermal sources and an HP-based system to find the most efficient setup. Its findings can guide the ongoing design efforts for green residential heat solutions at the research and commercial implementation levels.

25. An integrated life cycle assessment and life cycle costing approach towards sustainable building renovation via a dynamic online tool

Building stock retrofitting is crucial for achieving the building sector sustainability goals due to its high energy consumption rates. This paper by InCUBE introduces VERIFY (Virtual intEgrated platfoRm on LIfe cycle AnalYsis), an online tool for dynamic life cycle analysis and global warming impact assessments. It evaluates energy retrofitting measures for a multi-family residential building in Athens, Greece, aiming to reduce environmental impact and achieve near-zero energy consumption.

26. BIM-Based Construction Quality Assessment Using Graph Neural Networks

In this paper, HumanTech presents a novel approach for automating construction quality control. This method improves element-wise quality assessments by utilizing the semantic information in Building Information Models (BIM). The approach involves representing the as-designed Building Information Models (ad-BIM) as a graph, encoding elements' topological and spatial relationships. By using this representation, the paper proposes an algorithm based on Graph Neural Networks (GNNs) to infer element-wise built quality status.

Topic 4: Material science and design

 

27. Data driven design of alkali-activated concrete using sequential learning

Released by Reincarnate, this paper presents a novel approach to developing sustainable building materials through Sequential Learning. The approach can be immediately implemented in practical applications and can be translated into significant advances in sustainable building material development.

28. LLMs can Design Sustainable Concrete - a Systematic Benchmark

In the context of a circular building material economy, resource flows' complexity and material composition variability pose significant challenges. This Reincarnate study demonstrates how Large Language Models (LLMs) can advance material design by adopting a Knowledge-Driven Design (KDD) approach that outperforms traditional Data-Driven Design (DDD) methods.

29. Beyond Theory: Pioneering AI-Driven Materials Design in the Sustainable Building Material Lab

This work, by Reincarnate, focuses on using Artificial Intelligence (AI)-driven materials design to improve the sustainability of building materials with complex formulations. It provides insights into the real-world application of data-driven design, showcasing the successful integration of AI to advance sustainable materials science and boost sustainable building in the construction industry.

30. 14 Examples of How LLMs Can Transform Materials Science and Chemistry: A Reflection on a Large Language Model Hackathon

Large-language models (LLMs) such as GPT-4 have garnered interest from scientists for their potential in chemistry and materials science. Reincarnate organized a hackathon that showcased various LLM applications, including predicting molecule and material properties, designing novel tool interfaces, and extracting knowledge from unstructured data. This demonstrates the broad impact of LLMs across scientific disciplines beyond materials science and chemistry.

---

Share these research developments with your network!

Use our hashtag #Tech4EUConstruction on social media.

---

by andrea@australo.org