Our project is inspired by Wumpus World.
The AI will be required to navigate through a platform while they are under the effects of the blindness potion. They must locate the gold that is placed randomly on the platform while avoiding the golem. If the golem is in the way of the gold and there is no other way except to kill the monster, the AI must kill the golem, retrieve the gold, and return to their starting point (where they initially spawn).
Our input will be information about the map and the agent’s surroundings. The program will indicate whenever the agent smells a wumpus, feels breeze, encounters a golem, or discovers the gold. Our output will be the actions of the agent.
We will be using Reinforcement Learning and Dijkstra’s Shortest Path Algorithm.
The project will use a cost and reward system. If the AI dies they will lose a large number of points. However if they survive and attain the gold they will receive a large reward. If they die while trying to return to their starting point with the gold then they will lose a large number of points. If they survive and manage to return to their starting point with the gold then they win the game and receive a large reward. We believe that with this system the AI will learn how to navigate through maps more accurately and quicker.
To verify that the project works we will begin the AI on a controlled map that will not change (training data). If the AI succeeds they will move on to radomized maps for testing and learning. Our moonshot case is when the AI stops dying and is always able to retrieve the gold for every map.
Wednesday, April 24, 2019 DBH 4204