A preprint posted to arXiv (not yet peer-reviewed) describes a master's-level projects course that takes students through the full robotics pipeline — dynamics, state estimation, controls, and planning — while requiring them to run their code in real time on a cheap STM32 microcontroller. Titled 'Rusty Flying Robots,' the class makes students implement nonlinear algorithms similar to those in the PX4 flight stack, but without relying on pre-built black-box libraries. The goal is literal: get a small robot to fly using only the constrained hardware.

The methodology is project-based. Over two years, 10 students per year (total sample size of 20) completed the 9 ECTS course. Students write Rust code both for PC-based simulation and direct deployment on the STM32, receiving feedback through surveys. The authors note high student engagement but acknowledge the small cohort and single-institution setting. Limitations include the lack of long-term career outcome tracking and potential selection bias toward motivated learners.

Traditional robotics education often stays in simulators or uses expensive platforms running ROS on laptops, creating a gap between classroom theory and real-world embedded constraints. This course forces students to confront timing, memory, and computational limits immediately. What existing coverage misses is the broader accessibility impact: STM32 boards cost just a few dollars, dramatically lowering the barrier for universities in resource-limited regions compared to multi-thousand-dollar drone labs.

Synthesizing with the official PX4 documentation, which teaches comparable algorithms in C++, the Rust approach stands out for its memory safety guarantees without sacrificing real-time performance — critical for flying systems where crashes have consequences. The course also builds directly on concepts from 'The Embedded Rust Book,' applying them to a complete robotics stack rather than isolated examples.

This fits a larger pattern of Rust's growing role in safety-critical domains, from Linux kernel modules to automotive software. By teaching the language across both simulation and embedded targets, the program aligns education with industry shifts toward safer systems programming. The real novelty is showing that a full graduate-level real-time robotics curriculum does not require high-end hardware, potentially democratizing access to the skills needed for next-generation drones, automation, and exploration robots. While the preprint focuses on curriculum details and immediate feedback, it underplays the long-term potential for scaling such hands-on training globally.