Educating engineering students in physical robotics demands hands-on practice, but hardware integration and debugging complexities often impede progress and limit the project's scope. This study evaluates an intensive two-semester project-based learning course. Students designed, built, and programmed physical autonomous vehicles for maze navigation and lane-following tasks, systematically using Large Language Models (LLMs) as collaborative tools. Using Python, OpenCV, and PID control, students employed LLMs to accelerate code generation (especially for hardware drivers), enhance debugging (notably for intricate hardware-software interfacing), and clarify complex concepts. Crucially, pedagogical strategies emphasized rigorous critical evaluation and safe testing of AI-generated outputs intended for physical systems. A mixed-methods evaluation, blending robot performance metrics and student feedback, revealed that strategic LLM integration significantly enhanced project sophistication, enabling advanced implementations, reduced coding and debugging time (particularly for hardware interactions), and fostered greater problem-solving resilience. The findings demonstrate that LLM can enhance the effectiveness of PBL in physical robotics education, accelerate learning, and prepare students for AI-augmented engineering workflows, provided that critical thinking and safety are rigorously prioritized.

Authors:
Chen Giladi, Sami Shamoon College of Engineering, Israel
Shayke Bilu, Shamoon College of Engineering, Israel

About the Presenter(s)
Dr. Chen Giladi, lecturer in the Mechanical Engineering Department at Sami Shamoon College of Engineering (SCE), focuses on AI in education, agricultural & medical robotics and develops advanced machine learning algorithms in these fields.

See this presentation on the full schedule – On Demand Schedule

Bookmark