Course Description

The unifying theme of this course is the concept of an intelligent agent. Based on this perspective, the problem of AI is seen as describing and building agents that receive perceptions from an environment and perform appropriate actions based on them. This course will begin by examining the traditional methods for implementing this mapping from perceptions to actions including blind search, informed search, logical inference, and production systems.

Then the course will consider some of the fundamental problems with these traditional approaches and explore the concept of embodied agents, or robots. We will build simple, Lego-based robots controlled by Handy boards and implement behavior-based control. Finally the course concludes with an examination of various learning techniques including decision trees and neural networks.

There will be frequent programming assignments illustrating the concepts presented. Many assignments will be in Scheme, and others will be in C or Java.

Texts

• Required: Artificial Intelligence: A Modern Approach, by Stuart Russell and Peter Norvig
  
• Packet: Purchase from Bridget Rothera, the departmental administrative assistant in Sproul Room 2.
  • Excerpt from An Introduction to Genetic Algorithms, by Melanie Mitchell
  • Excerpt from Artificial Intelligence, by George Luger
  • Excerpt from Understanding Intelligence, by Rolf Pfeifer and Christian Scheier
  • Excerpt from Exercises in Rethinking Innateness: A Handbook for Connectionist Simulations, by Kim Plunkett and Jeffrey Elman

Grading

Homework Policy

On some homework assignments you will work jointly with other students (with each assignment I will explicitly tell you if working as a team is allowed). When you work as a team, then you and your team members may share code for that particular assignment. The team should submit a single assignment with all names clearly indicated at the top of the program file(s). All team members will receive the same grade for that assignment.

Programs will be graded with respect to both style (40%) and correctness/efficiency (60%). Good programming style includes the following features:

• an opening, summary comment for each file
• use of white space to separate sections of code
• descriptive variable names
• logical organization
• consistent indentation
• avoidance of long, wrapped lines
• avoidance of inline comments and over commenting

READINGS KEY
R&N = Russell and Norvig, P&S = Pfeifer and Scheier, P&E = Plunkett and Elman

Schedule