Next: CA algorithms Up: CA research Previous: CA research


Our closest collaborator remains the U.S. Air Force. They have two of our machines, and would like another five as soon as we're able to arrange for it. Jeff Yepez and his group there have used CAM-8 to make a number of breakthroughs in lattice-gas modeling, including momentum conserving models of liquid/gas phase transitions, and the first lattice gas models of elastic solids (Fig. 1). They are trying to arrange to have their new large CAM-8 machine built by Lincoln Labs, with us overseeing this project. (The Air Force Initiative for building a large CAM-8 machine started about a year ago).

Our collaborators at the University of Wisconsin are making their CAM-8 the basis of what they call the ``Ecomachine'' project-a suite of software they are developing to make the CAM-8 machine useful to a large modeling community. This project was based at least partly on the fact that their reaction-diffusion models ran about 17,000 times faster on CAM-8 than on their workstations.

Several collaborators, including Joel Yellin at U.C. Santa Cruz and Bryant York at Northeastern, have been using CAM-8 for combinatorics calculations (adjacency-matrix calculations, jamming densities, etc.). We're just beginning to collaborate with mathematicians at MIT on these kinds of simulations (we recently helped Peter McCorquodale implement a CAM-8 version of a combinatorial dynamics due to James Propp in the MIT Math Department).

We've also begun setting up a collaboration with MGH (through Tuan Cao-Huu) to study the use of CAM-8 in analyzing PET scan data. The volume rendering techniques that we've developed to visualize physical systems in three dimensions (see title page) should work well in this context; and our machine should be very effective for CA-based image-segmentation algorithms.

Dan Rothman, Eirik Flekkoy, John Olson and Chris Adler (all at MIT-EAPS), have adapted a CM-2 algorithm due to Bruce Boghosian (Thinking Machines Corporation) to simulate 3-D lattice-gas fluids much more efficiently on CAM-8 than the original algorithm did on the CM-2.

In order to promote collaborations, CA modeling in general, and our approach in particular, CAM-8 has traveled extensively this year. The farthest its been was last September, when it was featured at the Second International Conference on Theoretical Physics in Beijing, China. It was presented at the lattice gas conference in Toronto last year, and will be featured in several talks at this year's meeting in Princeton, at the end of June.

Figure 1

Next: CA algorithms Up: CA research Previous: CA research
Wed Jul 27 11:18:54 EDT 1994