More from the Harvard Self-Organizing Systems Research Lab, an interesting video presented at 13th International Symposium on Distributed Autonomous Robotic Systems (DARS 2016) in December.
As Evan Ackerman put it, “Assembly through self-disassembly”.
The Kilobot project is a swarm of 1024 small autonomous robots on a table. This is a testbed that is a step up from pure simulation, because gravity and other gnarly real stuff happens, so the algorithms have to be prepared to deal with anomalies.
The DARS presentation (I have seen the video, nothing more) reverses some earlier experiments in self-assembly. In this case, the swarm is programmed to assemble a shape through a subtractive, “disassembly” process, rather than the earlier “additive” assembly.
“If you are not part of the starfish, please leave now.”
This looks cool, but I had to think, “what is it good for?”
In general, this kind of self-assembly could be useful for creating temporary circuits or nanomachines (e.g., a funnel),that need to be deployed and used for a short time. This also could be used to create templates to mold fabrication. The swarm forms shapes for the part, and which is captured by some other material, and then the swarm is released to form the next part.
Obviously, it should be possible to make anything via either “additive” or “subtractive” algorithms (though some shapes will be more difficult for one or the other approach).
One advantage that subtractive methods usually have is that it may be easier to achieve uniform material throughout the shape. I’m thinking here that assembling a form (additively) over time will be subject to drift and errors, so the shape may not be uniform throughout. For one thing, I think that “removing” a piece should not disturb its neighbors unless they are somehow supporting each other. In contrast, adding a piece inevitably risks collision and displacement of neighbors.
In the subtractive method, you could initially pack the swarm very neatly, in a default, optimal pattern (perhaps even using macro processes to “manually” pack them). The subtraction only affects elements that are not part of the shape, and those on the edge. The interior is undisturbed by the process. I speculate that this could potentially create higher quality pieces.
So there you have are some speculations about why I might choose a subtractive algorithm over additive, if I had a choice.