Claytronics on Slashdot
The Claytronics Project has hit Slashdot (again) today. See the post by clicking here. Jason Campbell (Intel Research Pittsburgh) is quoted in the article, and some of the comments that have been posted have been interesting.
I’ve been collaborating with several people on the problem of how to write programs for Claytronics systems. Two people who deserve special note are CMU grad students Michael DeRosa, who has been working on a language called LDP, and Michael Ashley-Rollman, who has a language called Meld. (With Seth Goldstein, I co-advise both students.) Both LDP and Meld are interesting as languages for very large-scale distributed systems programming. To give you an idea of what I mean, check out this video of one million Claytronics atoms forming into a complex shape. The program, written in Meld in only a few pages of code, reads a CAD/CAM object file and then rapidly computes the motion plan for forming the 3D shape.
Other project videos can also be found by searching YouTube for “claytronics”.
Peter Lee @ March 9, 2009
That is a great program! Is there currently a production application for this? Since this reads CAD will this be used in 3D design R and D for new products/Inventions and other sorts of product design?
Unfortunately there isn’t yet a production application as this is a research project that we are still actively working on. The algorithm shown here is described in the paper here: http://www.cs.cmu.edu/~claytronics/papers/dewey-iros08.pdf
It works remarkably well for something like this trumpet, but there is still room for significant improvement, particularly in how Claytronics atoms are moved around. The current algorithm is far from optimality in this respect and could be made to run much faster. We may also need to augment Meld in order to be able to make these changes in the right way. Meld is still very much under development, although you can learn some about it here:
http://www.cs.cmu.edu/~claytronics/software/meld.html, and we are actively looking to understand what we need to do to make Meld better suited to writing programs like this. It currently works very well for this particular program, but shortcomings may arise as we augment it in which case we will have to find good ways of addressing them.
Very exciting stuff, especially after reading some of your articles! I will definetely follow closely your group when I start my grad studies in CMU.
It made me wonder that do you have collaboration with biologists? Seems nature is very good in solving this kind of distributed algorithm: an embryo starts as a bunch of anonymous generic cells which divides and divides until cells start to specialize (activate certain genes and passivate others) as muscle cells, skin cells, etc. and take their position in a relatively well defined geometric shape (for some better defined than with others :)). Of course, this process does not preserve mass as more energy and material is brought into the system in food. Actually my understanding is that science does not very well understand how this happens. Am I running in Meld?
[…] to give a distinguished lecture for this year’s Saul Gorn Memorial Lecture. My lecture was on programming languages for coordinating very large numbers of distributed computing nodes (which I also gave at KAIST earlier this year), based largely on the work by my students, Michael […]
[…] lots of papers in research conferences and journals. The project has also gotten more than its fair share of attention in the popular press as well as various blogs and websites. (See the news links on the project’s web page for some […]
Great movie
i like this movie thanks