Issue: Definability

Robots and rapid prototype fabricators are currently expensive machines to buy or time consuming things to build (even if you have the expertise). If you have to ship a new one to a far location, they are also fragile, heavy and bulky. Ideally they should be cheap and quick to assemble, with little expertise required and mainly using readily available raw materials to minimize the amount that has to be purchased and transported to the target site. Improving this state of affairs by a factor is 10 is a nice simple criteria. Or is it?

What does it mean to say a machine should be able to 90% replicate itself? 90% of the parts? 90% by mass? Or is it by volume, by purchase cost, by effort required, by expertise required, or some other criteria? Is it a specific one of these, an 'either or' combination of these, or perhaps it has to meet all of these tests? Can all of these, such as expertise, even be numerically defined?

About what about fraud? Suppose we say the criteria is that the machine must be able to replicate 90% of itself by mass? What's to stop some pedant taking a perfectly working commercial robot arm, and then defining a concrete base that is 9 times as heavy to be a required part of his system? Similar evasions by adding spurious parts exist for the other criteria. You could ask a technical panel to judge whether an addition is truly spurious, but that's a judgment call and so rather ill defined.

I asked Hod Lipson of Cornell University university about this. His thoughts were:

Sounds very cool. The idea of breaking the requirement down to 90%/10% is
great. I think you would need to be more precise when you say 90% of the
machineā€¦ is it 90% of the weight? Of volume? Of # of parts? Its easy to
achieve any of these by adding dead weight/parts. What you are really
shooting for is 90% of the complexity. The closest to this is part count,
and 90% of these parts need to be automatically assembled. Even then it gets
complicated: Consider our 4-cube robot. I can argue that the 4-part machine
assembles a 4-part machine, so its above 90%. But then the complexity of the
parts themselves comes into question. So you need to factor in the
complexity of the "raw materials". We've actually written a paper on these
questions (ECAL 2003 on our website).

An alternative approach would be to pick a current machine that is commercially available that does a similar function, and set an absolute benchmark of compared to that. Eg, the complete system must be able to replicate and then assemble the resulting parts into a duplicate of itself using less than 100 dollars of non-self-manufactured parts and less that 1 hour of human intervention (which could be spread over a longer period. eg 20 minutes putting pieces in place, 5 mins calibration, 5 mins doing up a particular snap fastner half way through and 29 minutes of finishing off assembling the remainder and connecting computer leads, ready for it to start working on a second duplicate to verify it is a fully functional copy).

Problems I see with this second approach are defining what counts as a 'raw material', and deciding on the real price of parts supplied by a third party.

Your thoughts? What is it most important that the prize submission achieve? How would you define this unambiguously, in a way that captures the imagination (eg 100 dollar laptop, or man powered flight across the English channel, rather than just flying 25 miles over land). What other ways must the defined criteria be defended against fraud or pedantry?

Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License