of Fundamentals Artificial Intelligence - Spring 2004 Instructor: Craig A. Struble, Ph.D. Lab 1: Intelligent Agents
Assigned: January 20, 2004 Due: February 18, 2004
Introduction
Central concepts for artificial intelligence are intelligent agents and rational behavior. In this lab, you will explore these concepts by building intelligent agents to perform a task in a simulated environment.
Outcomes
By completing this lab, students should Understand intelligent agents as described in Chapter 2 of Russell and Norvig. Be familiar with different agent programs, specifically table-driven agents, simple reflex agents, and model based agents. Understand the meaning of rational behavior. Learn to simulate intelligent agents in a software environment. Understand the advantages and disadvantages of table-driven, simple reflex, and model based agents.
Prepatory Reading
Chapters 1 and 2 in Russell and Norvig.
Materials
The following materials will be used in this lab: A Java based simulation environment for use with this assignment. Students wishing to use other languages may do so, but will need to implement their own simulation environment.
Pre-lab Questions
These questions should be answered before you perform the lab assignment. Record your answers in the introduction section of your lab assignment in your lab notebook. 1. Exercise 1.1 on page 30 in Russell and Norvig.
1
2. Exercise 1.7 on page 31 in Russell and Norvig. 3. Graduate Students Only: Exercise 1.2 on page 30 in Russell and Norvig. Write up to two pages addressing each of the questions mentioned in the exercise. The journal Mind is available online through MARQCAT. It is your responsibility to find the proper bibliography entry in the text, to locate the journal through MARQCAT, and to download the article by Turing. Include citations for any other literature you use as a reference. 4. Exercise 2.2 on page 56 in Russell and Norvig. 5. Exercise 2.5 on page 57 in Russell and Norvig. A Word template for PEAS descriptions is available from the course website. 6. Exercise 2.6 on page 57 in Russell and Norvig. A Word template is available from the course web site.
Procedure
This section provides the steps to take for this lab assignment. As you carry out each step, record observations you make in your lab notebook. Your notes do not have to be completed writing, but you'll use them to generate your final lab report.
Environment Description
You are responsible for building an Mars rover type agent to collect rocks on the surface of Mars. The agent's initial position is in the "center" of a rectangular grid. In grid locations around the agent is the Martian surface. Some of the locations contain rocks, which the agent should pick up. A sample environment is depicted in Figure 1. The rover is only able to grab and hold one rock at a time. So, every time the rover picks up a rock, it must be returned to the rover's landing spot. A correctly working rover should collect every rock in the entire grid and then return to the landing location. Each grid location contains at most one rock. If the agent attempts to drop a rock in a location already containing one, it will be unable to do so. When a rock is deposited in the landing location, the rock is stored in a collection bin. Percepts The agent perceives the following information: what it currently holds, what objects are in the same its location, heading (north, south, east, west), and if it tries to go past the edge of the rectangular region. This information is provided to the agent by the simulation environment. Actions The agent can perform the following actions: move forward, turn left, turn right, grab rock, release rock, stop. Performance For each rock returned, the agent scores 4hw points, where h is height of the Martian grid, and w is the width of the Martian grid. For each rock not returned, a penalty of 4hw points is scored. A penalty of one (1) point is scored for each action performed by the agent. If the agent stops in a grid location other than the center, a penalty of 4hw is scored. 2
Figure 1: A 3 3 grid with the rover starting at the center location. Rocks are in some of the grid locations.
Table Driven Agent
Agent programs for table driven agents consist of a table of actions that is indexed by a percept sequence. 1. Implement a table driven agent for the 3 3 environment provided by the instructor. 2. Simulate your agent in the environments provided by the instructor. Record the performance of your agent in each environment.
Simple Reflex Agent
The agent programs for simple reflex agents consist entirely of condition-action rules based on the current percept only. 1. Implement a simple reflex agent to collect rocks in all possible Martian environments. 2. Simulate your agent in the environments provided by the instructor. Record the performance of your agent in each environment.
Model Based Agent
In this part of the lab you are to implement a model based agent for Martian rock collection environments. Model based agents can take advantage of knowledge about the evolution of the world; that is, models of how the world behaves. In this case, you are free to incorporate memory and an understanding of how changes occur into your agent. 1. Implement a model based agent that can collect Martian rocks in all possible Martian environments. 3
Agent Table-driven Simple reflex Model-based
Environment 1 2 3 24 - - 18 -...