Abstract
Researchers at the Air Force Human Resources Laboratory developed an Intelligent Tutoring System (ITS) that utilizes a "microworld" simulation to teach students how to qualitatively deduce a satellite's orbital elements from its 2D ground track. By shifting the pedagogical focus from pure mathematics toward graphical pattern recognition, the system helps students build the same mental models used by expert satellite controllers. This paper outlines the system's architecture, its instructional philosophy, and early evaluative feedback.
Introduction
Monitoring the "exoatmospheric arena" is a critical mission for the Air Force, requiring personnel to interpret ground tracks—the two-dimensional path a satellite traces over the Earth's surface. Traditionally, students were taught to calculate orbital elements, such as eccentricity or inclination, using complex mathematical formulas. However, a significant gap exists between classroom theory and field expertise. Human experts rarely perform mental number-crunching; instead, they store ground tracks as graphical representations indexed by features like shape, size, and symmetry. The challenge addressed here was to build a system that bridges this gap, teaching students to "see" the physics behind the curves through discovery-based learning rather than rote calculation.
System Architecture and Methods
The system was developed on a Xerox 1108 workstation using the Lisp Object-Oriented Programming System (LOOPS). It employs a six-part architecture common to classic Intelligent Tutoring Systems:
- Expert Module: Contains the rules and qualitative models used to deduce shape descriptors (e.g., "lean-right," "hinge-symmetry") from orbital parameters.
- Curriculum Model: Organizes concepts by difficulty, ensuring students master simpler circular orbits before tackling complex elliptical ones.
- State Model: Tracks the student’s "experimental behavior"—whether they are exploring, predicting, or testing.
- Diagnostician: Analyzes student errors by comparing their inputs to the Expert Module and identifying specific "bugs" in their reasoning.
- Student Model: Maintains a record of the student's current knowledge and their effectiveness in using the provided tools.
- Coach: The "pedagogical engine" that decides when to intervene, offering advice if a student is stuck or using tools inefficiently.
The interface allows students to manipulate parameters (like the semi-major axis or inclination) using a mouse and immediately observe the resulting ground track in a simulated "microworld."
Key Insights and Instructional Philosophy
Qualitative vs. Quantitative Expertise
The most significant insight of this research is that expertise in orbital mechanics is primarily visual. The system prioritizes "shape descriptors" over raw numbers, helping students develop an intuition for how a 3D orbit projects onto a 2D map. Students are encouraged to use a "Why" menu to interrogate the Expert System. If they disagree with a descriptor, the system provides the expert's underlying reasoning, as shown in the interface below.
The "Microworld" and Monitoring Strategy
Rather than just presenting static problems, the system provides a sandbox. Students can "Generate a Ground Trace" to see the impact of their changes, fostering an inductive learning process. Furthermore, the "Coach" doesn't just look at whether a student's guess is right; it monitors how they experiment. If a student makes random, unsystematic changes to parameters, the Coach intervenes to teach better inquiry skills.
Key Results and Evaluation
While the system was in the evaluation phase at the time of publication, several key findings were reported:
- Expert Validation: NORAD crews and instructors confirmed the system's qualitative model accurately reflected professional heuristics.
- Adaptive Environment: The system successfully demonstrated the ability to switch instructional strategies based on whether a student was in "Discovery" mode or "Testing" mode.
- Tool Effectiveness: The inclusion of a "History Tool" (to overlay previous tracks) and an "Orbit Window" (to see the 3D orbit) was found to be critical for helping students recognize patterns.
Limitations and Future Work
The expert rules are currently highly specific to orbital mechanics; porting the system to a different field would require a complete rewrite of the Expert and Curriculum modules. A major question for future research is whether the "experimental behaviors" learned in this tutor—such as systematic parameter testing—would transfer to other domains. Additionally, the authors noted the need to investigate the optimal timing for intervention: specifically, whether it is better to diagnose a lack of domain knowledge or a lack of inquiry skills first.
Summary updated