A Robotics Environment for Software Engineering Courses

Stephan Goebel, Ruben Jubeh, Simon-Lennert RaeschAAAI 2011 · DOI

Introduction

Teaching Software Engineering (SE) through robotics is highly engaging because it turns abstract code into physical action. However, traditional robotics courses often face significant hurdles that distract from core SE principles. The authors identify three primary friction points: hardware scarcity, where student enrollment outpaces available kits; sensor noise, which forces beginners to spend excessive time on low-level calibration; and concurrency bottlenecks, as physical hardware testing is a serial process.

To address these issues, this paper presents a low-cost, scalable robotics environment. By offloading robot localization to a "global" webcam-based tracking system and providing a high-fidelity simulator, the framework allows students to focus on high-level object-oriented logic rather than low-level hardware debugging.

Methodology

The core of the proposed environment is a hybrid framework that bridges the gap between physical robots and virtual simulations.

Fiducial Indoor Positioning System (FIPS)

Instead of relying on onboard sensors, the system utilizes an overhead webcam and the reacTIVision framework. Robots and objects are topped with unique optical markers called "fiducials."

Lego Mindstorms robot equipped with a fiducial marker for tracking.

The FIPS system tracks these markers to provide each robot with its exact ID, absolute (x, y) coordinates, and rotation. This "Global View" abstraction broadcasts positioning data over a network, providing a "Global GPS." This allows students to program robots using a simplified sensory model, focusing on navigation and logic rather than signal processing.

The Fiducial Indoor Positioning System (FIPS) using an overhead webcam to track robot coordinates and orientation.

Unified Simulation and Abstraction

A 3D simulator built with Java and Fujaba serves as a first-class development environment. Crucially, the simulator uses the exact same interface as the physical FIPS system. Code developed in the simulator can be deployed to physical robots with minimal changes. To ensure high fidelity, the simulator includes "fuzzing" transformations to mimic real-world coordinate jitter. Furthermore, the framework includes a "Navigator" component that abstracts two-wheel differential steering into simple commands, hiding the complexity of motor synchronization.

Key Insights

The most significant contribution is the outsourcing of perception. By moving perception to a stable, external webcam, the authors lower the barrier to entry, allowing students to tackle complex multi-agent problems earlier in the semester. Additionally, the physical manifestation of object-oriented (OO) patterns—such as seeing a "Cat" robot chase a "Mouse" robot—is more intuitive for understanding OO interactions than abstract diagrams or UML models.

Results

While quantitative metrics were not the focus, the system demonstrated several qualitative successes in a classroom setting:

  • Scalability: The environment successfully supported courses where student enrollment exceeded available hardware kits.
  • Multi-Agent Coordination: The "Cat-and-Mouse" scenario served as a flagship application. A "Cat" robot hunts a "Mouse" robot, which moves randomly until a "Cheese" object (a box with a fiducial) is introduced. This demonstrated the system's ability to handle complex coordination and human-in-the-loop interaction.
  • Accessibility: The reliance on a simple webcam and printed markers makes the setup easy for students to replicate at home.

Conclusion and Future Work

The current system is limited to 2D environments and requires a clear line of sight between the webcam and markers. Future work includes extending the "Cat-and-Mouse" game to include more complex roles, such as a "Dog," to increase the complexity of required software patterns. The authors also propose broader pedagogical adoption through hands-on workshops using Java and Eclipse, suggesting the framework is ready for wider use in software engineering education.

Summary updated

BibTeX
@inproceedings{goebel2011robotics,
  author = {Stephan Goebel and Ruben Jubeh and Simon-Lennert Raesch},
  title = {{A Robotics Environment for Software Engineering Courses}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2011},
  doi = {10.1609/aaai.v25i1.7997},
}
Figure