CS340 Project 1 - The Vacuum World

Design a more sophisticated agent for the vacuum world. You may assume that the room the agents are to vacuum is a rectangle with no furniture or other obstacles. You agent may have memory. That is, the agent may have a “state” which can change and remember information from previous steps.

Here is a template that you can use:

// My Vacuum Agent

package ai.worlds.vacuum;

import java.util.*;

class MyVacuumAgent extends VacuumAgent

{

// declare any state variables here

void determineAction()

{

/* This method uses the percept (stored in the Vector percept)

and state values to determine the agents next action.

The member variable action is assigned a string (either “suck”,

“forward”, “turn left”, “turn right”, or “shut-off”) to indicate

the agent’s next action */

}

}

Give your vacuum agent class your name followed by VacuumAgent, for example, JillVacuumAgent. The class needs to be contained in file with the same name with the .java extension, for example JillVacuumAgent.java. In order to test out your agent, you will need to add this file to an Eclipse project which contains all the java files that I have provided and then make a couple of small modifications to one of my files.

Download the project files from http://phoenix.goucher.edu/~jillz/cs340/ai/ai.zip. You will need to use Winzip to extract the files.

You then should use Eclipse to import the project. Then compile the files and run the class AI.

Your new VacuumAgent file should be placed in the ai.worlds.vacuum package. Also in the ai.worlds package select the file WorldCreatePanel.java and make the following changes to include your agent.

· Add a string with name of your agent to

String[] vacuumStrings = {"Random Vacuum Agent",

"Reactive Vacuum Agent"};

· In the method createAgent, add an else if clause before the else like

else if (agentName == "Jill Vacuum Agent")

return new JillVacuumAgent();

You can compare your agent to the stupid ones provided by using the agent trials. Simply select the agents to be in the trials and run the trials. The average score of trials on identical environments will be displayed.

When you are satisfied with the intelligence of your vacuum agent, email your source code file to me at jill.zimmerman@goucher.edu . We will run agent trials in class among all your agents to see who has the best vacuum. The top agent will earn bonus points.