Forensic analysis of STM32F7 firmware failure modes in drone swarms
The article discusses the use of drone swarms in Kharkiv, highlighting their operation outside of US regulations. It explains that these drones cannot rely on traditional remote piloting due to extreme RF and GPS jamming. Instead, they utilize localized edge-compute systems for autonomous targeting and swarm coherence.
- ▪In the US, Part 107 regulations limit drone operations to one drone per pilot without a specific FAA waiver.
- ▪The drone swarms in Kharkiv operate under extreme RF and GPS jamming conditions, making remote control impractical.
- ▪Localized edge-compute systems, primarily using STM32F7s and NVIDIA Jetsons, are employed for autonomous operations.
Opening excerpt (first ~120 words) tap to expand
If you're flying in US airspace, yeah, absolutely. Part 107 strictly limits you to one drone per pilot. To run a swarm here, you have to get a specific waiver from the FAA proving you have solid collision avoidance algorithms and link fail-safes in place.But the 800-drone deployments happening in Kharkiv right now are obviously a completely different animal. They operate entirely outside civilian frameworks. Because the local RF and GPS jamming is so extreme (often >85% attenuation), they can't rely on remote pilots or cloud links anyway. Instead, they use localized, offline edge-compute loops—mostly using STM32F7s and NVIDIA Jetsons running Model Predictive Control (MPC)—to handle the targeting and swarm coherence autonomously.
Excerpt limited to ~120 words for fair-use compliance. The full article is at Ycombinator.