Automation in Construction in brief – 10/05/2025
Automation and Digitization
Jiang et al. developed a tool that uses a fine-tuned large language model to automate building energy modeling—simulating how buildings consume energy. Their system handles complex designs, different materials, and changing schedules, tasks that normally require heavy manual input. Tested on 402 cases, it achieved 100% accuracy while cutting workload by 98%. The platform could streamline sustainable design across diverse building types.
Source: https://doi.org/10.1016/j.autcon.2025.106223
Kim et al. developed an AI-driven method using reinforcement learning to optimize the placement of elements in BIM floor plans. Their approach reduced manual adjustments by over 90%, cut work time by 25%, and improved design clarity. By minimizing human error and boosting productivity, the system marks a step toward automating construction documentation in architecture and engineering.
Source: https://doi.org/10.1016/j.autcon.2025.106242
Researchers at Carnegie Mellon University have developed LegoGPT, an AI model that generates physically stable LEGO structures from text prompts. Trained on a dataset of over 47,000 designs, the system ensures buildability through physics-based validation and rollback mechanisms. The resulting models are not only aesthetically pleasing but also feasible for manual and robotic assembly, marking a significant advancement in AI-driven design.
A study by Duke University reveals that employees perceive colleagues who use AI tools as less competent, leading many to conceal their AI usage due to fear of professional stigma. This phenomenon, termed the “AI aversion effect,” suggests that despite AI’s growing presence in the workplace, social perceptions may hinder its open adoption and integration.
Source: https://arstechnica.com/ai/2025/05/ai-use-damages-professional-reputation-study-suggests/
Simphony-Dynamic-as-a-Service is a new cloud-based simulation tool designed to enhance construction project management. It integrates real-world data with advanced forecasting, helping decision-makers assess scheduling, resource allocation, and risk. The tool’s scalable architecture, standardized data formats, and user-friendly interfaces improve accessibility for construction staff. By enabling real-time tracking and scenario testing, it supports better-informed decisions, bridging the gap between simulation and practical construction management.
Source: https://doi.org/10.1016/j.autcon.2025.106198
An integrated framework combining Ground Penetrating Radar (GPR) and fuzzy logic has been developed to improve pavement thickness evaluation. The system automatically detects layer interfaces and calculates a Thickness Condition Index (TCI), aiding in more accurate condition assessments. Tested on 12 road sections, the models show reliable results, though challenges remain with thin layers and low contrast. The framework enhances pavement maintenance decisions and future research may improve robustness through advanced signal processing techniques.
Source: https://doi.org/10.1016/j.autcon.2025.106236
An AI-driven VR platform enhances customer-centric design in construction, making it more accessible by using smartphones instead of costly hardware. The system predicts user design preferences with neural networks and ranks alternatives based on cost, time, risk, and aesthetics using the TOPSIS method. A case study with 30 participants showed 79% satisfaction and 80% alignment with system recommendations, improving decision-making and customer involvement in construction projects.
Construction Robotics
Ersoz and Pekcan developed a system that uses drones and deep learning to monitor earthwork progress on construction sites. Their method removes construction machinery from drone images using AI, creating accurate terrain models. Tests showed less than 6% deviation from laser scans, while cutting labor time from days to hours. The approach improves safety and lowers costs by automating a once manual, risky task.
Source: https://doi.org/10.1016/j.autcon.2025.106211
Talamkhani and Liu developed an AI method to improve image analysis for underwater bridge inspections. Their system enhances poor-quality underwater images and accurately identifies structural components, overcoming challenges like murky water and low light. Tested against other methods, it achieved high accuracy, with over 90% segmentation performance. The technology could support cheaper, safer robotic inspections of underwater infrastructure.
Source: https://doi.org/10.1016/j.autcon.2025.106230
Space Construction
NASA’s TRIDENT drill, developed by Honeybee Robotics, successfully completed tests on the Moon during the Artemis mission. The 1-meter-long, lightweight drill employs both rotary and percussive mechanisms to extract lunar regolith. It halts at user-defined depths and deposits samples for analysis by the MSolo mass spectrometer. This achievement marks a significant step toward sustainable lunar exploration by enabling in situ resource utilization and advancing lunar science capabilities .
Source: https://www.nasa.gov/missions/artemis/nasas-lunar-drill-technology-passes-tests-on-the-moon/
NASA has selected 12 student teams as finalists for the 2025 Human Lander Challenge, aiming to develop solutions for storing and transferring super-cold liquid propellants essential for future lunar missions. Each team has received a $9,250 development stipend to advance their concepts. The finalists include teams from institutions such as California State Polytechnic University, Pomona; Colorado School of Mines; Embry-Riddle Aeronautical University; Jacksonville University; Massachusetts Institute of Technology; Old Dominion University; and Texas A&M University .
DLR researchers have developed TerraMind, an AI model that enhances the interpretation of Earth observation data.Capable of processing images, text, and time series, TerraMind outperforms existing models in accuracy while requiring less computational power. Its open-source availability aims to support climate research and environmental monitoring by enabling more efficient analysis of satellite data. The model represents a significant advancement in integrating AI with Earth system sciences.
Source: https://www.dlr.de/en/eoc/latest/news/2025/new-ai-model-for-a-better-understanding-of-our-planet
At the 2025 FIRST Robotics World Championship in Houston, NASA showcased its lunar exploration technologies to over 55,000 students and 75,000 mentors. Exhibits featured lunar rovers, robotic arms, and discussions on shaping the future of space discovery. The event highlighted NASA’s commitment to engaging the Artemis Generation, inspiring the next wave of engineers and scientists in STEM fields.
NASA is advancing artificial intelligence (AI) to enhance space exploration and address global challenges. Collaborations with companies like KX Systems have led to AI-driven tools capable of predicting space weather and analyzing satellite data. Additionally, AI technologies are being applied to monitor manufacturing processes and improve disaster response through real-time satellite data analysis. These innovations demonstrate NASA’s commitment to leveraging AI for both space missions and terrestrial applications.
ESA’s Advanced Concepts Team has launched the ELOPE (Event-based Lunar OPtical flow for Egomotion estimation) competition, inviting participants to reconstruct a lunar lander’s descent trajectory using only data from an event camera.Unlike traditional cameras, event cameras capture changes in brightness at high speeds, making them ideal for challenging lighting conditions. The competition, starting on 20 May 2025, offers a platform for global collaboration in lunar navigation technology development.
