Aerobrakes — Lesson 8: Noisy Data & Edge Cases

---

All three algorithms now work against clean simulated data — but real sensors are never that clean. This session finds out which algorithms actually hold up.

Objective: harden all three algorithms against real-world sensor imperfection.

Session Activities

Inject synthetic sensor noise into the simulation — Gaussian noise on readings, dropped samples, and added latency. Then run each of the three algorithms from Lessons 5–7 against this noisy data and see what breaks.

Define and explicitly test three categories for each algorithm: a happy path (everything working normally), an unhappy path (moderate noise/latency), and an edge case (a real sensor failure or a wind gust mid-coast). Don't assume an algorithm that passed the happy path will survive the other two — test all three deliberately.

Deliverable

A notebook writeup: which assumption from Lessons 5–7 broke first under realistic conditions, and what the team changed as a result. Being specific about what broke and why is more valuable here than just reporting that something “got worse.”

Maps to milestone P1-M5.