When construction sites are not captured and digitized while construction is in progress, there can be unforeseen and time-consuming impacts, especially when contractors are trying check their work. Automation and robotics startup Scaled Robotics is bringing automation to reality capture specifically geared towards construction that is in progress. Their automatic robot-mounted progress monitoring system can take what once used to be an imprecise task and turn it into a value-added dataset that helps workers to prioritize tasks, work smarter, and reduce re-work.
Addressing the digital disconnect on the construction site
The need for this type of technology is clear and addresses a crucial pain point in the construction process. While there have been huge advances in the modeling and design ends of construction, there is still a gap between the office and the field, says Scaled Robotics CEO Stuart Maggs.
“Essentially the problem we are trying to solve is that whilst we are mostly designing digitally, we’re still constructing manually.”
Maggs detailed a typical scenario he’d recently seen where existing tools – especially across large jobsites – simply don’t provide enough information. At a jobsite, the general contractor wanted to check if all the ducts in the ceiling were installed above 2 meters as specified by the plans. Without scans, they had to try to get the answer by walking around with a 2-meter-tall stick, walking around and stopping when it hit something lower than the 2 meter mark.
“People have developed tools to be able to solve this problem with laser measures or tape measures or rulers, but you just can’t do that for 100,000 elements in a building – it is not scalable or possible.”
The concept of Scaled Robotics was born out of the growing frustration of not having the tools available to build what was designed in the office. Scaled Robotics aims to bridge this gap between the digital and the physical world by automating progress monitoring. Rather than having a general contractor checking a representative sample of elements, the automated monitoring can look at all of the elements at once.
Getting to actionable information, quickly
The robot, which is a low to the ground 4-wheeled vehicle about the size of a small ATV, can be deployed into areas where data need to be captured. It can be either manually driven with a controller, or taught to follow a path to capture the needed information. It automatically avoids obstacles, people, holes in the ground and other construction hazards.
“Everyone gets obsessed with the robot, which is pretty cool, but ultimately it is just a device to capture high frequency information… the real value is being derived by the software platform.”
Once the data is captured, it is processed by Scaled Robotics’ software platform and is uploaded to the cloud. The data is then automatically processed and compared with the BIM, comparing what was intended to be built to what the robot was able to capture. The software can also take in data from more traditional laser scanners as well, providing even more inputs.
The software then uses machine learning and computer vision algorithms to analyze it, and then it is delivered in a web-based front end viewer that provides the user with actionable information. The user is getting more than just raw capture data – it can be searched, queried or manipulated to provide easy-to-understand information. The different layers of analysis can be used to produce high frequency information that improves decision making.
Maggs described the type of analysis the system can provide when working on a complex jobsite with high tolerances. On the first day that they captured information at the jobsite, they uploaded and analyzed every element within the floor that they scanned, and the system produced a report of every element and whether it was in or out of tolerance. This can be displayed as a list, but also visualized with color-coding to bring out of tolerance elements to the attention of anyone looking at the application. The information is also searchable and sortable, which meant that they could also rank them by the magnitude of the deviation. The highest risk elements at the top, with more trivial issues at the list. Within a few minutes, the system essentially produced an intelligent “punch list” that could then dictate what needed to be immediately corrected.