Syracuse humanoid learns construction skills

Syracuse University said on August 5 that a research team led by civil and environmental engineering professor Yizhi Liu is developing a humanoid robot that learns construction skills by watching human workers. The university said the system lets the robot observe, process and reproduce physical techniques demonstrated by skilled workers, rather than relying on remote control or fixed pre-programmed routines. Syracuse described the work as among the first of its kind in construction robotics. The findings were published in *Computer-Aided Civil and Infrastructure Engineering*. ### How does the robot actually learn from workers? Yizhi Liu and Ph.D. student Yanxi Liu built what they call a vision-based “perception-and-action” system, Syracuse said. The university said the framework enables a humanoid robot to watch workers, interpret body movements in three dimensions, translate those motions to fit the robot’s own body proportions, and then carry out the task. Yanxi Liu said humanoid robots are suited to construction because job sites are built around the human body. (news.syr.edu) Syracuse said that, unlike industrial robots designed for controlled factory floors, humanoid machines can move through scaffolding, ladders and uneven terrain that are common on work sites. ### What has the system done so far? Syracuse said the robot successfully learned and executed 30 distinct construction skills and tasks demonstrated directly by human workers on-site. (news.syr.edu) The university’s account did not list all 30 tasks in the summary it published, but it said the work covered basic construction techniques and physical task sequences. The study page on ScienceDirect says the project addresses a gap in research on giving humanoid robots the practical capabilities needed for construction work. (news.syr.edu) That description matches Syracuse’s account that the system combines perception and action so the robot can learn task execution from worker demonstrations. ### Why are construction researchers pursuing this now? Associated Builders and Contractors said in January that the U.S. construction industry would need to attract 349,000 net new workers in 2026 to meet demand for construction services. (news.syr.edu) Syracuse cited that labor shortfall in describing the project’s relevance to the industry. The International Labour Organization has said construction accounts for about 30% of all occupational fatal injuries worldwide. (sciencedirect.com) Syracuse also cited that figure as part of the case for robotics that can take on hazardous tasks or assist workers in risky settings. ### Who is behind the project at Syracuse? Syracuse said the research was developed by Yizhi Liu and Yanxi Liu. The university said Yizhi Liu joined the College of Engineering and Computer Science in 2024 after earning a doctorate in architectural engineering from Pennsylvania State University, and that his A-STAR lab focuses on robotics, artificial intelligence and human-robot interaction in construction settings. (abc.org) (ilo.org) A separate Syracuse profile published in 2025 described Liu as an expert in construction robotics and human-robot interaction. That article said he had also received a National Science Foundation grant for an AI-enabled roof-inspection robot, another project aimed at hazardous construction tasks. ### What comes next for the research? Computer-Aided Civil and Infrastructure Engineering is the publication where Syracuse said the team reported the new system. (news.syr.edu) The university said the work builds on earlier projects in Liu’s lab, including brain-computer interfaces for hands-free robot control and AI systems for detecting fall hazards on job sites. Syracuse said Liu’s A-STAR lab will continue work on robotics, artificial intelligence and human-robot interaction for construction environments. (news.syr.edu) The next visible milestone is likely to come through follow-on lab tests, additional peer-reviewed papers, or broader worksite validation tied to the university’s construction robotics research program. That expectation is an inference based on the lab’s stated research focus and the newly published paper. (news.syr.edu)