CABot 2

CABot 1 is the version of the project finished in July 2008.
It is slightly more sophisticated than CABot 1, works more or less all the time, and learns a goal action set.
The agent is in a very simple crystal space game.
The agent itself does almost all of its processing in fLIF neurons.
It can execute a few commands given by the user, sees the room ( multiple pyramids and stalactites), and uses a spreading activation net for plans and goals, which are set by the user.
To run the agent you'll need to either:
1. Build the source. Follow the CABot1 CS procedure but use the simple2 crystal space game that we developed; the new simple2.cpp and simple2.h should be used. You might want to keep a copy of the simple2 files from CABot1.
2. Or, you can download the much bigger version (about 25 megs) that has all the dlls included cs.tar.gz. I'm pretty sure that it'll only run from a cygwin bash prompt, but you don't need to install Crystal Space (which has been surprisingly difficult to do in cygwin). Just gunzip cs.tar.gz then tar -xvf cs.tar. Then go to CS and run ./simple2. The vfs.cfg file will almost certainly need to be modified to point to the cantAgent directory to where the cantAgent puts and receives data.
You will also need to install and run the CABot2 agent.
I'm afraid there is no planned Tech Report for CABot 2..
As far as I can tell it runs perfectly on the base commands Turn left/right. and Move forward/backward.; perfectly on the compound commands Go left/right.; when the vision is right (almost all the time) on Turn toward the pyramid/stalactite.; almost all the time on Go to the pyramid/stalactite. though it sometimes overruns (but then stops); and it learns to Center the pyramid/stalactite. quite well.
The parser is perfect on all the tested sentences.
Vision is more rudimentary, but largely succeeds in recognizing stalactites and pyramids.