Real life “Constructicon” quadcopter robots being developed | Hacker News Reader

Real life “Constructicon” quadcopter robots being developed

extremetech.com

34 pointsevo_915 comments

I think I recall something like this posted some time ago. It is an interesting idea, with a few caveats :

* At least in the video at the bottom of the article, it is not actually a swarm - there is never more than one robot doing each task (pickup, place), and one can see them waiting to get a 'workzone' cleared before moving in. In a swarm I'd imagine each flier working independently, with interaction limited to collision avoidance and occasional cooperation when needed(lift heavier block, etc).

* I hardly see such bots battling even moderate windspeeds in a real-life construction project, while carying any kind of real-life construction material block - assuming they can even lift one.

* Intuitively, I'd say this way of building is very energy-inefficient (in terms of power used by the drones) - but I'd love to see some actual figures.

If you are referring to the University of Pennsylvania quadrotor construction video, they were doing pipelining, basically for collision avoidance. They also have some cooperative lifting videos.

Edit: this article has that video at the end.

If making a building involved stacking blocks then it would be pretty easy for humans to make buildings too. I don't see how these robots would be capable of installing an electrical socket, or applying caulk, or any number of important tasks. Not to say a robot couldn't accomplish these things, and maybe a quadcopter could deliver supplies in some limited circumstances, though a robotic crane could deliver things as well. One robotic crane could actually assemble the entire described structure, only with concrete blocks instead of polystyrene. It's the eclectic nature of buildings that makes them hard to build, not just piling materials into place.

I can imagine something like this being very useful in situations that are complicated by the environment. For instance, in a setting like Chernobyl or Fukushima. Or in a war-zone.

I picture a Google Chauffeur (self-driving) truck, bringing supplies to the site (if possible), and quadcopters doing the final assembly. Push a button here, and a structure is assembled there, with no human intervention.

And don't limit yourself to picturing these dog-sized toys. Picture giant, military-style helicopters, but quadcopters. I don't think there's anything in the physics that doesn't scale. A Chinook CH-47D can carry 26,000 pounds of payload externally.

I think you are absolutely right that this is only going to be used when other means fail. Why would anyone lift standard materials for a sky scraper with an aircraft when they can use a crane?

But once you start doing construction the advantages of quadcopters (mainly control) starts to get outweighed by their inefficiency. Regular helicopters are far more efficient - large rotors are exactly what you want for heavy lifting. The strongest helicopter I know of is the CH-53E Super Stallion, and that is a single rotor aircraft (even though it has multiple engines).

I think 10,000 quadcopters might be able to do something really amazing. I'm back down to "dog-sized," and I'm back to small payloads. It's a design challenge to come up with useful atomic units that such a robot could lift, but the number of them that can operate independently is a serious factor.

I'll recast this in terms I'm more familiar with: we used to scale software by getting faster CPUs with more powerful instructions. Now we're all being told to scale by numbers of CPUs. Can building construction be similarly recast?

Your CH-53E Super Stallion is effectively scaling up. Can we instead use quadcopters to scale out?

Like in Batteries Not Included?

The team has developed an algorithm that in their words is “limited only by the battery life of the quadcopters and the number of parts available.”

Wow, what an understatement. This is in fact the main problem.

Motors consume a lot of energy - in fact electric motors consume more than 2/3 of produced electricity - this is manufacturing, transportation, even most of the electricity consumed by computers goes to power cooling ventilators.

Electric Quadrocopters usually have flight times measured in minutes - and they carry a heavy accumulator, maybe a third or half their weight. And the picture of that quadrocopter lifting that block of concrete is actually funny if you have ever thought seriously about flying robots - where can I find those motors and propellers?

Do you have a citation for "in fact electric motors consume more than 2/3 of produced electricity". I would be interested in reading more about total energy consumption.

I very much doubt that these bots will ever replace humans -- and certainly not near term.

What they will properly do is become what computers are to us -- a tool which helps humans.

While most technical obstacles will be overcome the limitation of these robots are their use of energy -- lifting something up in the air is so much less efficient that almost any other method of transportation.

Anyone on HN a quadcopter aficionado? I would like to get into this hobby, and am eager to learn more.

I'm not really an aficionado, but I've done some basic feasibility research as part of school work in university (I'm studying Automation and Control Engineering, and we are tasked at presenting a possible quadrotor/quatrocopter solution for scrutiny).

So you will need to know a lot of stuff, but it depends on what you want to do. What you need to know: - Electric drives. (Or internal combustion engines, but I think they are way more complex, while undeniably more powerful, also more control problems) - Control Theory, mostly state-space. - basic electronics - Microcontroller programming. - you will probably need to know at least some digital signal processing. - AI - Radio communications (you have to communicate with the robot, even if its autonomous)

Shit, that sounds hard. There's no gentler way in then? I was hoping for ready-made kits running Linux :)

This? http://aeroquad.com/

I'm not sure how much effort it will replace thought. I doubt everything can be abstracted away in a software.

Don't they already use quad-copter (human pilots) for logging in remote areas?

Reply on news.ycombinator.com