Multi-agent frameworks and Python

The 3 second status update: I had a truly wunnerful December holiday in South Africa, featuring an abundance of sun, sea, crayfish, beer, meat, fish, various tasty molluscs, a very exciting jaunt in a 4-seater Cessna plane (thanks Dave!), good friends and fun-loving family.

After re-adjusting to Dutch weather and getting back into the work thing for a few days, we spent a weekend in Koeln (or Cologne; that’s in Germany for the geographically challenged), worked some more and spent this past weekend debauching terribly with a large group of friends in Bradford-on-Avon, a picturesque little town about an hour’s drive from Bristol (that’s in the UK, again for the geographically challenged readers).

I’ll be spending at least the next few weeks (or months?) investigating the possibilities of multi-agent visualisation and image processing systems. It seems like some of the better multi-agent frameworks have been implemented in Java. At the moment, I’m leaning towards Jade, mostly because Cougaar, otherwise a good system, doesn’t support any of the FIPA (The Foundation for Intelligent Physical Agents) standards.

Now I just have to figure what’s the thinnest possible layer required to inter-operate with Jade agents. Also, if I’m seriously going to play with this, I have to make it work with Python. No, Jython is not the answer. Really. The options seem to be getting Jade to build in GCJ and then wrapping with SWIG (a la PyLucene). Ouch. I think JPype might be more what I’m looking for, but I’ll keep you up to date.