A Pedagogical Framework for Modeling and Simulating Intelligent Agents and Control Systems

Classroom assignments are an effective way for students to investigate many important (and fun) aspects of artificial intelligence. However, for any project of reasonable breadth and depth, especially involving multiple agents and graphics, most of the programming goes into tedious, error-prone administrative tasks. Moreover, time constraints usually preclude an extensible and reusable design, so this overhead repeats itself throughout the semester. This pedagogy-oriented modeling-and-simulation framework provides all the necessary support capabilities to get students up to speed quickly on playing with AI content. It contains extensive, highly configurable, yet user-friendly, engineering, physics, and communication models for arbitrary components within a definable task environment. These components are managed automatically in a stochastic simulation that allows students to define, test, and evaluate their performance over a wide range of controlled experiments. Introduction and Background Student assignments and experiments in artificial intelligence can be fun, exciting, and educational. Most programming assignments, however, involve a disproportionate amount of mundane, administrative code that distracts students from the AI focus. This pedagogy­ oriented system provides a straightforward framework of highly extensible components and functionality to investigate many concepts and strategies in AI and intelligent control systems (Russell and Norvig 2003, Bourg and Seemann 2004). It also helps foster an understanding of proper methodology in the design, implementation, testing, and evaluation of formal experiments. This system is a lightweight version derived from a full­ scale, accredited simulator developed for the Department of Defense to evaluate components for the U.S. Army's Future Combat Systems program (withheld, Engle 2005). Copyright © 2008, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. Another derivative has also been used successfully in AI for interpreting spatial relations in natural language processing (withheld). For portability, the system is written entirely in Java, with Java 3D as the display engine.