jaysulk / wumpus

Artificial intelligence agent for the Wumpus game

Hunt the Wumpus

Artificial intelligence agent for the Hunt the Wumpus game.

AI Agent

To implement your own AI agent you just need to create a custom class that implements the wumpus.Agent interface.

public class CustomAgent implements Agent {
    /**
     * Do something before take the action, probably debug...
     */
    public void beforeAction(Player player) {
        System.out.println(player.render());
        System.out.println(player.debug());
    }

    /**
     * Do something with the player
     */
    public Action getAction(Player player) {
        // Feel the perceptions
        boolean bump = player.hasBump();
        boolean breeze = player.hasBreeze();
        boolean stench = player.hasStench();
        boolean scream = player.hasScream();
        boolean glitter = player.hasGlitter();

        if (bump || breeze || stench || scream || glitter) {
            // TODO: Do some action based on the players perception...
        }

        return Action.NO_OP;
    }

    /**
     * Do something after taking the action, probably debug...
     */
    public void afterAction(Player player) {
        // Players Last action
        System.out.println(player.getLastAction());
        // Show a very happy message
        if (player.isDead()) {
            System.out.println("GAME OVER!");
        }
    }
}

The return of the getAction method is the next player move. From this method you can access all Player perceptions and informations.

To run your custom Agent call it from the Main class at some World instance.

// Instantiate the AI agent
Agent agent = new CustomAgent();
// Create an world and execute the agent
World world = new World(4, 4);
world.execute(agent);
// Print the results
System.out.println(world.renderScore());

Output

Results for *CustomAgent*:
+----------------------------+
| Outcome | Score    | Steps |
| ------- | -------- | ----- |
| WIN     | 943      | 50    |
+----------------------------+

Debug

There are some methods to aid debug the agent strategy (don't forget System.out.println):

// Renders the player board perspective
player.render();
// Returns current player information
player.debug();
// Returns the last action taken null if none
player.getLastAction();
// Step-by-step debug
Environment.trace();

Output

+---+---+---+---+
|   |   |   |   |
+---+---+---+---+
|   |   |   |   |
+---+---+---+---+
|   |   |   |   |
+---+---+---+---+
|   | → |   |   |
+---+---+---+---+
// player.debug()
Position: (1,3,E)
Score: -1
Perceptions: [BREEZE]
GO_FORWARD
Press ENTER to continue...

Deterministic board

By default, all the dangers and the gold are setup in random fashion at each run. You can setup the board as you like by calling the setters methods:

// Creates a deterministic world
World world = new World(4, 4);
world.setWumpus(0, 1);
world.setPit(2, 2);
world.setPit(3, 0);
world.setGold(1, 1);
// Show the board
System.out.println(world.renderAll());

Output

+-----+-----+-----+-----+
|     |     |     |   P |
|   ~ |     |   ≈ |     |
+-----+-----+-----+-----+
|   W |   $ |     |     |
|     |   * |   ≈ |   ≈ |
+-----+-----+-----+-----+
|     |     |   P |     |
|   ~ |   ≈ |     |   ≈ |
+-----+-----+-----+-----+
|     |     |     |     |
| →   |     |   ≈ |     |
+-----+-----+-----+-----+

Disclaimer

This game structure and implementation is based on the Wumpus Lite v0.21a of James P. Biagioni of the University of Illinois at Chicago.

Available at https://www.cs.uic.edu/~jbiagion/wumpuslite.html.

License

Copyright (c) 2016 Edson Hilios. This is a free software is licensed under the MIT License.

• Edson Hilios. Mail me: edson (at) hilios (dot) com (dot) br