I have been working for some time on the problem of "unrolling" complex task models into Markov Decision Processes, which can be solved analytically. This paper describes some work on this problem in the context of the DARPA Coordinators program for multi-agent schedule repair. computer science research planning artificial intelligence http://rpgoldman.real-time.com/papers/ss07-gtdt-coordinators.pdf Fri, 26 Jan 2007 17:37:34 -0600 Mozilla Firefox and Thunderbird have lots of extensions, and there doesn't seem to be a good way to determine whether a particular one is good or not. This page gives the beginnings of a list of extensions that I have vetted. http://rpgoldman.real-time.com/mozilla-extensions.html open source Mozilla Tue, 16 Jan 2007 15:16:59 -0600 I am just about to update my desktop to a new version of Mandriva Linux. Others seem to find this to be a smooth process, but I find it painful in the extreme, and often find it's weeks before the new system is fully productive and configured the way I like it again. So this time, before I start doing anything irrevocable, I'm trying to come up with a checklist of things to do before and after the actual upgrade. http://rpgoldman.real-time.com/linux/mandriva-update-process.html linux Mandriva open source Wed, 08 Mar 2006 10:47:04 -0600 Added a short paper, to appear in the proceedings of ICAPS-2006, describing how to plan with durative actions in the hierarchical task network (HTN) planner, SHOP2. http://rpgoldman.real-time.com/papers/icaps06-short.pdf planning computer science research Tue, 28 Feb 2006 16:48:05 -0600 My paper on Incremental Verification (with David Musliner and Mike Pelican) has been published in the Electronic Notes in Theoretical Computer Science. In controller synthesis, as we have approached it in the CIRCA system, one conducts repeated verifications of partly-constructed controllers (automata). For this reason, one can achieve significant speedups by caching partial verification results and restarting the verification where the last one left off. The actual process is somewhat more complex than that sounds, because there may be backtracking in the synthesis process, but that's the general idea. http://tinyurl.com/akas3 computer science research timed automata planning verification Sun, 05 Feb 2006 20:48:49 -0600 I have just co-authored a paper, with teammates from the DARPA Coordinators program, that describes using multi-agent MDP techniques for multi-agent coordination over hierarchical tasks. http://rpgoldman.real-time.com/index.html#sss06IU lisp computer science research artificial intelligence I managed to translate some of the old Garnet documents, originally in Scribe, into LaTeX and thence to HTML. http://rpgoldman.real-time.com/lisp/index.html#garnetDocs Sun, 22 Jan 2006 00:00:00 -0600 lisp I finally got around to making ASDF-installable, and adding to www.cliki.net my implementation of the Floyd-Warshall all-pairs shortest-paths algorithm. http://rpgoldman.real-time.com/lisp/index.html#floydWarshall Mon, 23 Jan 2006 00:00:00 -0600 lisp The Minneapolis Public Library has a pretty good web interface for managing the books you take out, etc. Unfortunately, it has a terrible login protocol. This Greasemonkey script helps you avoid the large amount of typing you usually need. http://rpgoldman.real-time.com/mpllogin-scrubbed.user.js web utility software open source Tue, 24 Jan 2006 20:58:57 -0600 http://rpgoldman.real-time.com Robert P. Goldman's homepage: updates Wed, 08 Mar 2006 10:47:04 -0600 http://blogs.law.harvard.edu/tech/rss