Inside Agents - Basic Strategies Explained

The Heroes of the Wumpus Environment

This page is the first in a line of pages that explains how agents act in the wumpus environment. If you are not yet familiar with the rules for this environment, please start at the introduction page that describes the rules and elements of the Wumpus Environment.

A lot of agents for the Wumpus-Environment have already been implemented. The goal in creating these agents was to show the development from very inefficient agents to agents that succeed in nearly all environments, despite those where the way to success is nearly - or at all - impossible to find.

Therefore a dumb agent and a reflexive agent as described in Artificial Intelligence - A Modern Approach by Stuart J. Russel and Peter Norvig were implemented first followed by more intelligent agents that have an - repeatedly updated - internal representation of their environment. Those latter agents represent a class hierarchy were every agent excels over his successor by either extending the methods responsible for updating the internal map or by superior reasoning or by both.

Finally, the last generation of agents is able to search for a goal field.

The above described strategy brings with it a lot of different classes - especially for representing the environment. This could be found unnecessary but it shows how an agent can be made 'more intelligent' than its successor by simple extracting a little more information out of the percepts it receives and tuning its reactions according to this more of information.

By following one of the supplied links you will get a description of how the agent works.


Last modified: Sa. June 06 14:34:54 1997 by Christian Schwaiger

email to