SiOO
SiOO is an acronym for State Input Operator Output
SiOO is an embeddable cognitive architecture based on Soar 8.5.2. It is being enhanced with a *nix style command shell, an embeddable footprint, development tools and modifications to the actual cognitive architecture.
Sponsored By:
SiOO has been made possible by the generous support of Module Master LLC. http://modulemaster.com/rebuilds/about-us/ Module Master supports the principles of Liberated and Open Source and makes all changes to the underlying system available under the GPLv3 License: https://gnu.org/licenses/quick-guide-gplv3.html. SoarSuite 8.5.2 was licensed under a BSD License, that license is in the documentation and the original source.
Applications:
Example applications include specialized applications like energy management in the home or office or supervising scientific data collection activities across an entire project. The ability of the system to learn "changes the game" from adaptable but preprogrammed controllers - to full fledged software agents that can seek solutions that were not considered at design time.
Requirements:
SiOO uses a command line shell interface and is intended for use on GNU/Linux systems but it should work on Homebrew for MacOS and Cygwin on Microsoft products with simple modifications if any. The build system uses the GNU Autotools system to build the kernel. The build system uses a bin/sh script and sed. There are no external library requirements. Prebuilt GNU/Linux, (GNU libc) ELF executables are distributed in the sioo/out directory.
HOW-TO run the SiOO Command Line Demos
Note that you need to change to the build directory of the CLI system, and then run sioo from that directory or the system won't find the demos. (this hassle is a listed bug, yes, its simple, but its complicated too ;-).
Missionaries and Cannibals Puzzle
There is a demo for the Missionaries and Cannibals puzzle that comes as a binary rete network dump. Again, starting in the sioo/cli directory, run ./SiOO (start fresh from scratch) and try rete-net -load agents/mac.rete once it is loaded you can then run it, try 100 Decision Cycle groups: run 100 run this command 3 to 7 times (it varies) to see the system solve the problem. Notice that it takes hundreds of steps. To watch it run straight through, try init-soar to reset the system and then do run by itself. The system will blaze through the puzzle leaving a trail of steps.
Using Intelligence and Learning
Lets turn on the actual learning system, "chunking" and see what happens. try learn -on. Next type learn you should see a statement that says learn -on. Lets reinitialize the system to start all over from scratch with init-soar, now we are ready for the next step.
Learning changes the entire game in these toy domains. Now lets run the MaC simulation with learning on, try run 100 you may or may not have to type it again. Note how many Decision Cycles(DC) it took to solve the problem. Now type init-soar again and do run. There is the magic number; once this domain is fully chunked, the system has found and remembered a set of rules that solves this configuration in optimal execution time.
Other Demos Restart SiOO from scratch again, just to be sure we clear everything (This IS Alpha software). At the command line you can do either of these:
counter-demo and toh-demo
Next you will try run 1 and then repeat that to see these demonstrations perform their tasks. Remember you can use any number after run and no number at all to run until completed.
Visit the Wiki for More Info