Artificial Intelligence
Whether you have experienced construction at home, with an office, a manufacturing plant, or other large project, you have most likely seen a problem with the process. It is, unfortunately, normal to run over budget and over schedule. With large projects, it becomes process and project management and is even more problematic. In a house, it is easy to know what’s missing and needs to be done. An office building, a road, a large manufacturing plant, and other highly complex projects make it very difficult to even know what is missing. Planning becomes an almost impossible job. Artificial intelligence (AI) companies are working to address that challenge.
One critical way contractors work to understand whether or not schedules can be met is through visual inspection. However, matching a space to the specifications is not easy. It’s here that computer vision and machine learning can improve accuracy and efficiency.
First let’s talk about what technology can’t do. While construction sites are becoming more complex, and connectivity to the on-site construction office is standard, broadband access throughout a site is still rarely possible. Having a live link between CAD models and an inspector is not something that can be relied upon.
Buildots is a company working within that constraint to improve process management through speeding and increasing the accuracy of the inspection process. The founders, Aviv Lebovici, Yakir Sudry and Roy Danon, were looking at manufacturing processes when somebody pointed out to them that construction was a different type of manufacturing. Their shift was then to understand how best to manage and track the construction process. One key area was matching information from inspections to the plans.
A camera, with storage, attached to a safety helmet is the front-end component to the system. Live video is not needed, so the camera is slowed to 2fps, collecting plenty of images for analysis while keeping the space down to what can be kept on flash drives. “Not every camera application needs live video,” said Aviv, Lebovici, CPO & Co-founder. “By minimizing the frames per second, we are able to provide a light, wearable, camera while still capturing the necessary details for our artificial intelligence engine.”
Video capture for AI in construction
Back at the office, those images are extracted from the camera and the AI analysis begins. There is, of course, the expected vision component that identifies objects such as walls, windows, and lighting fixtures. However, that doesn’t teach anything new and is becoming a normal component, as explained in a column last year about relocation moving estimates. It is an important piece, but quickly becoming a “must have” rather than a “nice to have.”
What is key is that Buildots then works with CAD systems, such as those from Autodesk. Inspecting a construction site is more than making sure an HVAC vent has been built. It also must be built in the correct place. By using AI to match images to the software models, the system can provide architects and construction teams the ability to understand the progress and accuracy of the project, minimizing cost and time overruns.
When I asked the Buildots team about advancement in both more dynamic and automated procedures, I received two answers. The first relates to the message mentioned above about construction networking. While fiber is now more frequently being run to large construction sites, that broadband goes to the main office. Even wifi is not something that is yet standard, so the current state of inspections needs to rely on a delayed feed. “While there are plans to link AI vision capture directly to CAD models and back to mobile devices, that is not practical at this point,” added Roy Danon, CEO and Co-founder. “The industry is changing, but things happen slowly. AI can address certain issues now and, as infrastructure changes, systems can be improved in the future.”
There’s also a question of drones, something that was pushed for so many applications in recent years. Large construction sites are complex, with lots of moving parts and constant change. Drones will need to become both more intelligent and smaller before they can be safely and cost effectively applied to the construction industry.
Many well established industries are hesitant to try new technology. While retail has incentive to rapidly adapt new technology to better address customers, construction has reasons, from safety to high cost and more, to be a slower adopter of technologies. What is shown is that artificial intelligence can help the industry and is beginning to make inroads.
