Automation in Construction in brief – 14/06/2025
Automation and Digitization
Researchers have developed Cloud2BIM, an open-source tool for automated conversion of large-scale point clouds into IFC-compatible BIM models. The tool segments and classifies structural elements like slabs, walls, and openings, handling non-orthogonal geometries efficiently. It achieves up to seven times faster BIM conversion with minimal user input. Validated with benchmark datasets, Cloud2BIM shows high precision and scalability, offering a robust solution for digital documentation in the construction industry.
Source: https://doi.org/10.1016/j.autcon.2025.106303
Researchers have developed an automated framework to streamline assembly-oriented template models for digital twin implementation in Design for Manufacturing and Assembly (DfMA) construction. The framework reduces model size by 87.3% and loading time by 95.1%, while significantly decreasing geometric complexity and quality calculation. This advancement enables real-time cyber-physical synchronization and proactive quality control during dynamic assembly. Future work aims to integrate closed-loop control systems and automated assembly equipment to enhance construction automation.
Source: https://doi.org/10.1016/j.autcon.2025.106314
Researchers have developed an automated data acquisition method for Terrestrial Laser Scanners (TLS) in civil engineering. The technique uses a vision-language model to convert text prompts into image semantic masks, guiding the TLS to scan only targeted objects. This approach, validated in various civil scenarios, reduces redundant data and enhances efficiency. Field tests included capturing the position of a tower crane and extracting the pose of bridge piers. The method aims to streamline infrastructure monitoring and management.
Source: https://doi.org/10.1016/j.autcon.2025.106337
Researchers introduced a new AI model for object detection in complex construction scenes. The Complex Construction Scenes Transformer (CCS-TR) uses a lightweight transformer-based method, improving detection accuracy by 5.1%-8.8% compared to existing methods, while maintaining lower computational costs. The model addresses challenges such as boundary information sharing and scene sensitivity, making it suitable for real-time on-site detection. Future work will focus on multi-modal collaboration strategies for extreme lighting conditions.
Source: https://doi.org/10.1016/j.autcon.2025.106330
Meta has introduced its V-JEPA 2 AI model, designed to enhance AI agents’ understanding and prediction of physical world events. Trained on over 1 million hours of video, this model is 30 times faster than Nvidia’s Cosmos model. Meta anticipates that this technology will significantly advance robotics, enabling AI agents to assist with real-world tasks without requiring extensive training data. This approach is beneficial for autonomous systems operating in dynamic environments, such as construction sites.
Source: https://techcrunch.com/2025/06/11/metas-v-jepa-2-model-teaches-ai-to-understand-its-surroundings/
OpenAI has delayed the release of its first open model in years until later this summer. CEO Sam Altman announced that the company’s research team had achieved an unexpected breakthrough, making the wait worthwhile. The competitive landscape has evolved with Mistral and Qwen releasing advanced models. OpenAI aims to outperform these rivals and is considering additional features for its open model. The release is crucial for OpenAI to improve its standing with researchers and developers.
Source: https://techcrunch.com/2025/06/10/openais-open-model-is-delayed/
Construction Robotics
Warp, a tech-enabled supply chain streamliner, is introducing robots to automate its warehouse network. After creating a digital twin of its Los Angeles facility, Warp experimented with various automation methods, ultimately finding success with retrofitted off-the-shelf robots. The company, which serves enterprises like Walmart and HelloFresh, aims to reduce labor costs and mitigate staffing issues for its warehouse partners. Warp has raised $10 million in Series A funding, co-led by Up.Partners and Blue Bear Capital, to deploy these robots in core networks this year. The technology, currently not for sale outside Warp’s hub, is expected to provide a competitive advantage.
Elon Musk has announced that Tesla plans to launch its long-awaited robotaxi service in Austin, Texas, on June 22. The service will use driverless Tesla Model Y SUVs, equipped with an “unsupervised” version of the company’s Full Self-Driving (FSD) software. Initial operations will be limited to small fleets in geofenced areas. Musk’s claims about Tesla’s self-driving capabilities have been scaled back over the years, and the company is currently under investigation for crashes involving its FSD software.
Amazon’s Zoox begins robotaxi tests in Los Angeles. The company, known for its purpose-built autonomous vehicles, will use retrofitted Toyota SUVs with human drivers for initial testing. Zoox, already testing in five other cities, is taking a slower rollout approach than competitors Waymo and Cruise.
Source: https://www.planetizen.com/news/2025/05/134982-amazon-owned-robotaxis-begin-testing-la
Bedrock Ocean, a startup mapping the seafloor with robots, has raised $25m in Series A-2 funding led by Primary and Northzone. The company’s autonomous underwater vehicles (AUVs) use sonar and magnetic sensors to map the ocean floor more efficiently and cheaper than traditional ships. Bedrock’s AUVs, costing less than $1m each, can cover the same area as one mapping ship, and are less disruptive to marine life. The company has conducted paid surveys for offshore wind, oil and gas, and environmental assessments, and has drawn interest from the U.S. Navy.
Source: https://techcrunch.com/2025/06/10/bedrock-ocean-dredges-up-25m-to-map-the-seafloor-with-robots/
Scientists at ETH Zürich have developed ANYmal, a quadruped robot capable of playing badminton. Equipped with a racket-holding arm and a stereoscopic camera, the robot uses reinforcement learning to predict shuttlecock trajectories. However, ANYmal’s 0.35-second reaction time pales compared to human players’ 0.12-0.25 seconds. The team aims to improve ANYmal’s perception and reaction time, but commercial badminton-playing robots seem unlikely.
XRobotics, a San Francisco-based robotics company, has gained traction with its countertop pizza-making robot, xPizza Cube. The machine, leased at $1,300 monthly, can produce up to 100 pizzas hourly, saving staff time by automating repetitive tasks. Unlike competitors, XRobotics focuses on assistive technology, making its device compact and affordable. After a successful pilot, the company raised $2.5 million in seed funding and plans to expand to Mexico and Canada. Despite market size, XRobotics remains committed to the pizza industry.
Source: https://techcrunch.com/2025/06/09/xrobotics-countertop-robots-are-cooking-up-25000-pizzas-a-month/
Researchers have developed a new motion planning system for mobile robots in dynamic construction environments. The approach uses low-feature mapping and pose-based positioning, reducing CPU usage by 19%, memory usage by 8%, and energy consumption by 9.5% compared to traditional methods. The system aims to enhance productivity, safety, and flexibility in construction tasks. The study highlights the potential of multi-sensor data fusion and multi-constraint path planning in improving autonomous construction operations.
Source: https://doi.org/10.1016/j.autcon.2025.106334
A comprehensive review on Bluetooth Low Energy (BLE)-based indoor positioning systems (IPS) highlights the technology’s evolution, challenges, and advancements. Key techniques include geometric mapping and fingerprinting, with machine learning and deep learning algorithms improving accuracy in complex environments. The integration of BLE with other technologies like Wi-Fi, UWB, and IMU is noted for enhancing precision and robustness. The study also emphasizes the role of datasets in refining IPS algorithms and the benefits of advanced BLE technologies.
Source: https://doi.org/10.1016/j.autcon.2025.106316
Space Construction
SpaceX plans to demolish structures at Cape Canaveral’s SLC-37 for Starship launches. The US Air Force, overseeing the environmental review, found no significant impacts on local interests or competitors. SpaceX may conduct up to 76 annual launches from SLC-37, pending final approval. The company is also expanding facilities in Texas and California, potentially having nine active launch pads across three states. Meanwhile, competitors ULA and Blue Origin operate only one or two pads. SpaceX’s expansion could lead to over 100 annual launches at Cape Canaveral, with the Air Force anticipating no significant disruptions to other operations.