Thank you.
>Like a fool I have agreed to program an agent (like the Wumpus world example
>in Artificial intelligence a modern approach - Russell & Norvig), but this
>project is COMPLETELY different to the little programming that I have done,
>I would be grateful for any pointers
The AIMA home page has links to source, including Prolog source for the
Wumpus world: http://www.cs.berkeley.edu/~russell/aima.html
Unfortunately the state of agents is implemented using destructive
update (assert etc). A cleaner design is for the environment program to
call the agent(s) with their previous state (a Prolog term) and the
current percept and have the agent return the new state explicitly as
well as the action. I would like a version of the code which does this,
but have not got around to doing it myself (yet).
lee
The MSc course on multi-agent systems I teach (with Graem Ringwood)
regularly sets programming a Wumpus world agent as a coursework.
See the web page for the course:
http://www.dcs.qmw.ac.uk/~gar/MAS/
We program it in a concurrent logic language rather than Prolog, however.
Matthew Huntbach