Human-Free Construction (2018)
forconstructionpros.com
forconstructionpros.com
Mass-grading is the only part of construction anywhere close to this that I’m aware of. Manufacturing processes for building materials are really impressive, but it’s a different story on the job site where you constantly have to make judgments about the work environment. I think most of the low-hanging fruit for automation is in the design process.
I own a business framing houses, it would radically change the dynamic in that field. However i am confident i could execute in that new environment, given the technology.
Makes me want to start an open source 3d building printer foundation.
It might be useful for making concrete tanks. I think the problem is in including the steel reinforcements.
A house builder machine that could wheel in and do all but the electrical, plumbing, and appliances would be pretty impressive. It'd need a variety of attachments for digging and leveling, pouring concrete, printing walls, filling insulation, and manipulators for placing beams, windows, and doors. That last part probably needs human workers too, so the machine needs to be safe to work around. All that sounds very expensive.
Useful on the Moon and beyond, maybe so useful on Earth. Also worth looking at factory-built houses.
would require some new materials to be viable, most certainly.
There are so many unappreciated subtleties to physical work. Manufacturing changes the work to fit the machine. So much of the work in construction just is not amenable to that.
There's still a ton to be done, of course. But something as simple as installing an outlet during construction requires a level of physical skill that is so far beyond what any robot can do that we have no idea how it would be accomplished. Maybe we could design around these things, create building standards that prioritize mechanical feasibility, designs where every step is laid out so that a robot can perform the next necessary step given everything that's been done before (and probably dozens of dedicated robot types for different functions). But that requires perfect vertical integration from design to each point of construction and verification. It would be hard to apply incrementally.
It seems like the realistic improvements are: design; construction planning, scheduling, timelines; applying instructions to the worksite; managing materials; applying instructions to the materials; verification and measurement. All the information parts, not so much the physical parts. Maybe materials could be cut? But even that seems questionable when so many things are cut to fit.
But I did enjoy hearing the story of Dusty Robotics, which is working on worksite lay out (i.e., the chalking/etc to guide construction): https://www.youtube.com/watch?v=nokkftQVa8M
This could be the equivalent of the Turing test for robotics, perhaps. Although I'd extend it to "install a new outlet in any randomly selected existing house, as requested by the occupants and in accordance with local regulations and national electrical standards, without damaging anything. And clean up after yourself."
"Where are the robotic bricklayers?"[1] is illuminating, but there is much more in the other blog entries.
Many of Brian's other essays go into the detail of various aspects. The whole of them taken together make the following points: every construction project has site differences, materials differences, and requires different mix of workers (so group learning/workflow automation is pointless), with the workers doing a vast array of different tasks with bulky, heavy, and low-value materials, each task being a tiny percentage of total building effort and cost, and there are strict safety, performance, and aesthetic requirements on the product (the building). And there are always surprises.
These factors incentivize a tightly regulated, single-project contractual system, great conservatism, and little possibility for whole-system construction innovation to take off.
Successful construction innovations are for (parts of) a single task, e.g. nail guns replacing hammers for framing, rather than being a whole-system "house printer"which also prints the required permits from the local authorities.
Any attempt to replace humans is going to have to deal with all of that. That article was from 2018. Progress in robotics and AI since then has not been on pace to meet the OP's "widespread in 2028" goal.
--- The author uses manufacturing as an example of successful replacement of humans by machines and argues by analogy.
As an aside, it's not the shiniest one. Agriculture went from employing around 60% of adults before the industrial revolution to well under 2% today, a 30-to-1 ratio. Manufacturing peaked at around a third of the workforce and is now near 10%, so maybe a 4-to-1 ratio.
The more important point is that manufacturing automated by rigorously standardisng inputs, factory operation, and products. In each construction project, the site is different, different materials are used because of different climates and different regulations and different customer demands, there are different logistics, and the customer wants a different-looking product. Manufacturing's automation success is not an appropriate analogy.
TL;DR: writers need to get up from their desks and go and look at whatever it is that they're writing about.
1. https://constructionphysics.substack.com/p/where-are-the-rob...