What’s the future of robotics technology on construction sites?
BY: Corie Cheeseman
In Part I of our Trailblazers interview with Brian Ringley, construction technology manager at Boston Dynamics, Ringley revisits his career background and his thoughts on innovation in the construction industry.
In our follow-up article, Ringley shares his perspective on how technologies such as industrialized construction and field robotics could greatly impact the industry, and he explores the future for AI and machine learning in construction.
Burcin Kaplanoglu, executive director, innovation officer for Oracle Global Business Units, led the discussion.
BK: Which technologies do you see having the biggest impact on the industry?
BR: I see a lot of promise with industrialized construction. The bulk of the work I was doing at WeWork was thinking about how to culturally approach the development and delivery of the building as a product―something that can be learned from and continually refined.
At WeWork, we benefited from also being owners―having access to our space and measuring the success of each subsequent project. The ability to move as much of the work off the jobsite and into controlled conditions as possible, as well as the ability to re-imagine supply chains relative to unitized modes of construction, are the most important things.
Secondly comes field robotics. But if you really extrapolate into the future and think about what types of things robots will be doing on our construction sites, first you must imagine, what are we no longer doing on construction sites in the first place?
You start to narrow down the set of problems that robotics will be solving in 10-20 years, and then you can focus on those problems. That’s why I always think about industrialized construction first and then field robotics second. You need both; it’s not an either/or question.
Beyond that, there’s a lot of interest from our customers regarding connecting their BIM systems to robotics. That’s a great goal, but this approach presumes that robotic technology will fix some of the inherent business and communication problems with BIM.
Again, the ability to leverage that technology is only as good as the contract. If you can’t have multiple shareholders sharing risk and being in the same model performing actions, there’s not a lot you can do.
Reality capture data and augmented reality
More to the point is the ability for these platforms to ingest the reality capture data from the site daily and to have spatial awareness of the real world versus the digital asset. The ways we do that right now with laser scanning and traditional surveying methods are quite cumbersome.
There’s a subset of emerging browser-based platforms that can convert reality capture data and associate it with certain solutions through model parameter updates and intelligent understandings of which points of the point cloud relate to which building elements.
But then there’s also stuff that some of our customers are doing with augmented reality (AR) technologies and ways of providing modes of localization that start to tie in recognizable physical features in the environment with known locations in digital assets. This provides the kind of bi-directionality you want in those models.
You could drive a robot from that model, but then you could also consume the data that the robot collects to further enhance and augment that model.
BK: Where do you see the biggest benefit of AI and machine learning when it comes to the industry?
We’re seeing a lot of it right now in providing some semantic understanding of the environment, including the ability to run computer vision models over collected imagery data to look for certain things. We see this with companies like Smartvid.io that are focused on issues around job site safety, auditing, and better management practices with regard to COVID-19 and reducing the spread of disease.
We use it a little bit, like in Spot, for instance. The arm uses models like that to understand different types of door handles, including how to open and close doors as part of extending its ability, to be autonomous in human-purposed environments.
There’s also a lot of urgency around a better semantic understanding of point clouds– it’s one of the most common forms of reality capture data. It’s interesting to see efforts like scan-to-BIM go head-to-head with efforts like 360 imagery and video to photogrammetry, meaning using image and depth data to reconstruct 3D environments.
There are trade-offs in terms of accuracy, processing time, and ease of use, but it’s compelling. There isn’t a clear winner yet, so it’s something to keep an eye on.
See Boston Dynamics’ Spot robot in action at the Oracle Industries Innovation Lab, where Oracle works with customers and partners to co-innovate and transform industries.