The University of Wisconsin-Madison is employing the work of a unique robot dog at the Kellner Family Athletic Center’s construction site near Camp Randall.
The robot is a part of a study led by Zhenhua Zhu, an assistant professor in the Engineering school and director of UW-Madison’s Digital and Robotic Construction team, to see how robots could support construction workers by performing repetitive, time-consuming or hazardous tasks, such as site scanning and progress monitoring.
Zhu’s research initiative works to increase jobsite safety and avoid human casualties. In 2023, the U.S. Bureau of Labor Statistics reported the construction industry was responsible for roughly one in five workplace deaths.
The Kellner Center, set to open in 2027, will be an indoor practice facility for all 23 UW-Madison varsity sports. The robot dog, developed by Unitree Robotics, is working with construction firm JP Cullen.
Unitree Robotics focuses on the development and production of high-performance quadruped robots, giving their B2 models the nickname of “robot dog.” UW-Madison became affiliated with the B2 through a grant from the Wisconsin Alumni Research Foundation.
“It requires human oversight, but our study’s purpose is precisely to identify and reduce potential risks so future robots can operate more safely and independently,” Liqun Xu, a graduate student who works closely with the robot, said.
The project captures “construction workers’ physical stress exposures in the field over extended periods,” Xu said.
“The primary goal is to understand the real-world navigation challenges that quadruped robots face on active construction sites,” Xu said.
Xu said the project focuses on building better human-robot relationships in real-world environments. “Eventually, this work can support safer human-robot collaboration in high-risk sectors like construction and infrastructure inspection.”
The Engineering school research team has already noticed improvements while working with Unitree B2.
“The robot’s mobility is excellent. It can handle stairs, uneven ground and tight indoor spaces, making it well-suited for real construction sites,” Xu said. Working with the robot allowed them to reveal navigation hazards that cannot be determined in simulation or lab settings.
But the B2 isn’t without its setbacks. “The robot’s navigation system relies on a geometry-based autonomy stack,” Xu said, “which often struggles with interpreting context, such as fragile objects, symbolic markers like red tape or human work zones.”
Other challenges include insufficient processing power that led to failures in mapping surroundings and required human intervention when the robot became stuck in difficult positions.
The primary objective of the research is to bridge robot collaboration with humans. Unless robots are made context-aware and socially adaptive, they risk becoming ineffective or even dangerous, Xu said.
“Our research helps…ensure robotic systems are not just functional but trustworthy and safe,” Xu said. “By identifying specific hazards like terrain instability, material contact risks and boundary violations, we aim to inform the design of more adaptive and context-aware robot autonomy systems that can safely coexist with human workers in complex environments.”
