drone

Multi-Chip Avionics Stack

Three-layer heterogeneous compute: AI inference, FPGA sensor fusion, MCU safe-landing autonomy.

Multi-Chip Avionics Stack

Reference avionics package embodying graceful degradation: Jetson Orin NX 16 GB (~100 TOPS, 10–25 W), Xilinx Zynq UltraScale+ ZU3EG (EKF <1 ms), STM32H7 flight-layer watchdog with NVM Return-to-Home.

Layer 1 executes compressed Tiny AI (YOLOv8n / MobileNetV3-Small, INT8) with inference <100 ms and model size <20 MB.
Layer 2 guarantees deterministic EKF fusion in FPGA fabric independent of OS jitter.
Layer 3 isolates flight safety: if both upper layers fail, MCU lands the aircraft from non-volatile memory.

Source: SWARMind Drones Proposal Draft § Multi-Chip Architecture.

Mission profile

Embedded on each swarm node; laboratory gate requires physical layer-disconnect demonstrations of expected survival behaviour.