Due date: Tuesday, April 15, 2008.

Original project as reference

Ex. 1 This project implements a version of PacMan in Python. We'll implement a fuzzy logic controller for the PacMan.

Download the zip file from the link above and unzip it. Read the instructions on how to run the script with different layouts and different agents. In the same folder, download the following file:
fuzzyAgents.py

For example, to run the random agent from the file we're adding to the project on the tinyMaze layout we would use the command:
python pacman.py -l tinyMaze -p RandomAgent

To integrate our controller with the project, we need to define our own special agent called FuzzyAgent found in fuzzyAgents.py. Every agent needs to define the function getAction that returns the next step to be taken by the Pacman.

What you need to know (and you may use other ways to get the same information if you find them easier):

• The state parameter in the function getAction is of the class GameState defined in pacman.py.
• The state allows you to find all the directions in which the pacman can move:
  state.getLegalPacmanActions()
• To find the food information you need to call
  state.getFood()
  which returns a 2D list of Booleans. A similar function getWalls gives you information about the walls.
• The function calls state.getCapsules() and state.getGhostStates() give you the remaining capsules and the positions of the ghosts as lists of coordinates.
• The function call state.getPacmanPosition() returns the current coordinates of the pacman in that state.
• You can generate a state resulting from executing a move from the current state with:
  state.generateSuccessor(0, action)
  where action would be one of the possible directions that you get from the function above.

Fuzzy Agent To implement the fuzzy controller, you'll have to implement the following functions:

• fuzzify(action): This function should return a fuzzy value of attraction towards the direction given by the action parameter. In this function you must consider both the possible food or capsules in that direction and the danger from ghosts.
• defuzzify(self, actions, values): This function should choose one of the possible actions based on their fuzzy values as an index in the array of actions. You can use the past action and its value in the decision if you want. The easiest way to choose is to select the action with the maximum corresponding value.

Turn in: the file fuzzyAgents.py and any other file that you modify or add.

Extra credit: The agent with the best score will get 5 extra credit points, the second place 3 points, and the third place 1 point. There is a "shoot for the moon" option of aiming for the lowest possible score, and the winner in this category gets 3 points. You can submit two versions of the agent if you want, one for a high score and one for a low score.