[not-so-mindless automatons]
while studying for one of my comp sci classes i was trying to justify in my own mind why anyone would want to know about things like fa''s (finite automatons) and pda''s (push-down automatons). then i was struck by an idea: why not use a simple dfa (deterministic fa) model for ai? try this on for size:
-the ai is composed of ''mental states'' (equivalent to the states of the dfa. there would be no final state per se, but there would be a start, or default, state.
-the system then switches states based upon ''input'' from the npc. for example, if their environment does not change significantly then no real input is given to the system and it stays in it''s mental state. if, however, it is by itself and ''hears'' a noise behind it, the system may get a ''frightened'' input and it will switch to the mental state corresponding to an agitated state. this may cause the npc to start to wander away if they are cowardly, or investigate if they are brave, etc. if they ''hear'' another noise and they are already ''agitated'' they could again switch states to a ''scared'' state which will send them to find more friends.
-each state must have (in good dfa fashion) possible moves for each possible input so it doesn''t get locked into some state unintentionally.
this may seem kinda ugly but i think it would work nicely (haven''t had a chance to code it yet). lemme know what you guys think
<(o)>
<(o)>
April 13, 2001 02:21 AM
Any agent-based AI I''ve ever seen has been essentially a DFA. They all have basic [receive input][process input][react according to input] designs that may vary in appearance, but perform the same duties.
Making formulated mental states will just make the code more readable; it will not add any functionality compared to what exists.
Making formulated mental states will just make the code more readable; it will not add any functionality compared to what exists.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement