Other CABot components
- There are some other minor bits.
- I originally built the components and sewed them together
symbolically.
- I replaced the symbolic controller with a neural one.
- CABot1 went wait for command, parse it, set goal, achieve them,
and repeat.
- CABot2 can process goals while parsing.
- CABot2's learning dichotomy stuff is also different.
- Roman's work is based on his optimal learning theory.
- CABot2 has a neural version of this using spontaneous activation
and environmental feedback.
- CABot3 will have spatial cognitive mapping.
- That is, the agent learns the map of the game.
