TU Berlin built a robotics masters course where students make real drones fly, using Rust, no black-box libraries.

Most robotics courses hand students a simulator and a flight stack. This one doesn’t. Students build every layer themselves, in Rust, then deploy it to a real quadrotor’s STM32 microcontroller.

The pipeline, four parts, one language: → Simulation : students write their own physics simulator, no libraries → Controls : a geometric nonlinear controller, the kind used in real PX4 flight stacks → State estimation : a multiplicative EKF running on SO(3), with real sensor data → Planning : quadratic-program-based trajectory planning through narrow passages

The same Rust code runs in simulation on a laptop and, largely unchanged, on 168 MHz / 192KB RAM hardware. That’s the part Python and C alone can’t do together.

Most students had zero Rust experience going in, majority came from C and Python backgrounds. The course still scored 1.1/5 (1 = best) in official student surveys across two years.

One student’s comment: “Was genuinely one of the best yet.”

Full teaching material is open: http://imrclab.github.io/teaching/flying-robots…

http://arxiv.org/pdf/2604.00032

#RustLang #Robotics #Embedded #Education #OpenSource #UAV

Flying Robots

Source: https://imrclab.github.io/teaching/flying-robots The course is largely practical, where students work towards implementing all the components that are needed on a modern flight controller in Rust. The different aspects are:

• Introduction to flying robots (multirotors and variants, fixed-wing, hybrid)
• Underlying physics models of the dynamics
• Control approaches (differential flatness, geometric controllers for attitude and position)
• State estimation techniques (Mahony filter, Extended Kalman filters)
• Motion planning for flying robots (optimization using splines, search-based methods, sampling-based approaches)

This course is being presented at the Robotics in Education (RiE) conference in 2026. arXiv