Wumpus World AI

This project was completed in python with a partner as part of the Intro to AI class.
Environment Assumptions

• The agent always started in the bottom left corner of the field of the cave and started facing right
• The cave consisted of a grid of 4-7 rooms by 4-7 rooms
• The agent could only move to adjacent squares
• The wumpus’ adjacent squares had a stench percept
• Each room had a 20% chance of being a deadly pit
• All pits’ adjacent squares had a breeze percept
• +1000 for collecting the gold
• -1000 for death by either wumpus or pit
• -1 for every action take (moving, changing direction, picking up gold,…) Our AI’s strategies:
• Use a depth first traversal to move around the cave, backtracking when a percept is observed
• Of the adjacent rooms that have not been visited, choose the one that requires the least amount of turning
• Use a probability matrix to represent which rooms have been visited (probability of gold) as well as which rooms probabily have pits or the wumpus
• If a stench is smelled immediately, shoot the arrow to the right and continue with a depth first traversal
• Once the wumpus’ location is known, ignore all stenches and just ignore the wumpus square