Autonomous Aerial Robots to Monitor Construction
A video is available online that shows one of the research team’s Unmanned Aerial Vehicles (UAV) flying over a residence hall construction project on the campus of the University of Illinois at Urbana-Champaign.
Imagine a construction site in which small, computer-controlled aerial robots autonomously navigate the job site indoors and outdoors, conduct visual inspection with onboard cameras and more. These robots can measure construction progress and provide detailed and continuous performance data on workers and equipment. This futuristic scene is probably a lot closer than you think, thanks to research by CEE assistant professor Mani Golparvar-Fard and collaborators Timothy Bretl, associate professor of Aerospace Engineering, and Derek Hoiem, assistant professor of Computer Science.
Quadcopters, small aerial robots with four propellers ranging in size from two to four feet across and also known as Unmanned Aerial Vehicles, are being deployed already at some construction sites, but their use is currently limited to taking photographs and videos. In addition, all quadcopters now being used at construction sites can operate autonomously only when access to GPS data is available; otherwise they must be operated by people. Golparvar-Fard and his research team are developing something very different: quadcopters that will be computer-controlled to fully automate the data collection and, more importantly, the analysis and reporting of progress information on the construction site. They have also developed a way for these robots to install cameras on elements of the site automatically. They are envisioning the use of these cameras for other tasks such as tracking the location of workers and equipment and maximizing accuracy in activity interpretations for performance monitoring purposes.
Construction is a $900 billion industry with 25-50 percent waste in coordinating labor and equipment and in managing, moving and installing material, said Golparvar-Fard. In addition, he said, productivity arguably has been declining for many years at the industry level. This has led the National Research Council of the National Academies to identify improving the efficiency of construction as a key national need. Construction monitoring improves efficiency by characterizing the extent to which construction plans are being followed and the extent to which workers and equipment are fully utilized. Current methods can be expensive, time-consuming and subjective, resulting in less frequent monitoring than is optimal, Golparvar-Fard said.
“The need for prompt feedback about actual or potential performance deviations on job sites is growing because the margins of profit are getting smaller, and companies need to be more competitive. Also current data collection methods are tedious, arduous and prone to being done only intermittently. … Many companies end up having incomplete performance data, and you cannot mount any analytics or root-cause assessments on top of it,” Golparvar-Fard said.
The system Golparvar-Fard and his team are developing works this way: the quadcopters take photos and videos of the construction site, guided by a cloud-based computer program that can direct them to the rights spots, resulting in automatic data collection. The activities of the aerial robots are fully autonomous, including take off, navigation, landing and charging. The captured images and videos are then used to create an actual 3D model of the site under construction. The system compares this automatically generated 3D model to the as-designed 4D (3D plus time) Building Information Model, resulting in more frequent and complete progress monitoring information. The system also autonomously mounts battery-operated and WiFi-enabled surveillance cameras on different elements of the site to automatically capture videos of ongoing construction operations. Once the data is captured and transferred to the cloud, the system automatically detects and tracks workers and equipment in real time from the video feeds and categorizes activities of the resources automatically. The progress and activity monitoring results are visualized in a web-based, 4D augmented reality (D4AR) environment—a representation of the construction site in 4D with additional performance information superimposed on it. These D4AR models can also be made available to construction professionals through smartphones and tablets, enabling them to make effective control decisions on- and off-site more quickly and easily.
“The autonomous nature of the system in terms of data collection and the automated performance analytics will significantly improve monitoring and control practices in construction,” Golparvar-Fard said.
An expert in developing innovative construction monitoring methods, Golparvar-Fard (PhD 10) is an alumnus of both CEE at Illinois and the Computer Science department. His research interests include creating and developing computer vision, image processing and machine learning methods to automatically monitor building and construction performance. For this project, Hoiem’s contribution will be in visual analysis, which is necessary for enabling the computer system to recognize construction materials from the surfaces of images, identify the actions of workers and track construction equipment. Bretl’s focus will be automating the process of data collection with the aerial robots. His team will develop methods of navigation and control that get the robots safely from one place to another in the site, enable the placement and retrieval by robots of cameras on structural elements, and guarantee that enough video is taken to support the visual analysis.
In order to test and develop their system, the researchers have been granted access to several active construction sites around the U.S. being operated by Zachry Construction and Turner Construction, including the new Sacramento Kings arena in Sacramento, Calif. In addition, the University of Illinois Facilities and Services division has granted access to a current residence hall construction site. The result will be a new way to increase efficiency on the job site by giving the routine, tedious jobs to the quadcopters and the associated cloud-based computer systems, so construction professionals can focus on the more important tasks of decision-making and root-cause analysis of performance deviations.
“We want to minimize the time to access accurate performance information and we don’t want construction professionals to spend any time doing analytics on how things are running on the job site,” Golparvar-Fard said. “All we want them to do is focus on how things could be improved.”
The preliminary work for this project was funded through an interdisciplinary faculty fellowship grant from National Center for Supercomputing Applications. The National Science Foundation awarded the team a $1 million grant to kick off the full project from January of 2015.
Students bring the UAVs to a construction site on the University of Illinois campus.