Helios
Spanning risk modeling to downlink, Helios is a wildfire-watching CubeSat prototype built by a capstone team, with Ada flight software running on a NEORV32 RISC-V soft-core to capture and transmit images tasked by a ground-based risk classifier.
Penn State Behrend
Wildfire monitoring from low orbit is a real mission profile, and a Penn State Behrend capstone team built a working prototype of the whole loop. Helios, by Seth Ford, Justin Wyffels, Calvin Nordstrom, and Leonard Brennan, with AdaCore mentorship through the GNAT Academic Program, spans risk model to downlink.
On the ground, a logistic-regression classifier, built from scratch in Node.js, trained on fire detections, weather, and terrain context, with feature imputation and probability calibration, scores coordinates for wildfire risk. High-risk coordinates task the satellite. On the flight side, a NEORV32 RISC-V soft-core (synthesized from the team's own VHDL top, with dual UARTs and an open-drain I2C bus) runs Ada flight software: a command listener, a driver for a TTL serial JPEG camera, and an imaging pipeline that captures on demand.
The downlink is the professional touch: images stream over the MIN embedded framing protocol, wrapped for Ada over its C reference implementation, as chunked frames that a Dockerized Python ground station reassembles back into JPEGs, with dev containers, I2C debugging notes, and a logic-analyzer guide documenting the road there.
Tasking, capture, downlink, reassembly: a CubeSat mission in miniature, with Ada in the pilot's seat.