Mission
Visual orbit and attitude determination for small spacecraft
Argus is a 1U CubeSat technology demonstration of a low-cost autonomous vision-based orbit determination system. It uses onboard Earth imaging, machine-learning-based landmark extraction, IMU measurements, and batch nonlinear least squares estimation to recover spacecraft orbit and attitude.
The lost-in-space problem
When a small spacecraft deploys, it often has no prior knowledge of its state and must recover its orbit relying solely on onboard computing and sensing. Space-rated GNSS receivers are bulky, consume 1–2 W, and can cost around $10,000 — while consumer GPS modules are legally restricted at orbital speeds and altitudes. Ground-based radar and radio ranging can take weeks and still leave position errors of 10–20 kilometers. Visual sensing offers a complementary, autonomous path.
The Argus approach
Argus replaces expensive tracking hardware with a simple onboard RGB camera weighing under 15 grams, using under 5 watts, and costing under $100. An EfficientNet region classifier determines which of 16 highly salient MGRS regions the satellite is viewing; a YOLOv8s landmark detector then pinpoints ground features to produce bearing measurements. A batch nonlinear least squares estimator fuses these with IMU data to solve lost-in-space initialization, and an EKF (with a multiplicative variant for attitude) keeps the state updated recursively.
Validation
A GNSS receiver flies only as the truth reference. Argus compares its vision-based solution against onboard GNSS to evaluate estimation accuracy in orbit — the mission targets orbit determination errors under 50 km and attitude determination errors under 15 degrees.
Secondary mission 1
Doppler-based ground OD
Ground stations receive GMSK signals and study how SNR, bandwidth, and duty cycle affect orbit determination precision.
Secondary mission 2
Amateur radio packet repeater
Argus will support an amateur packet repeater as operational constraints allow.
An educational mission
Argus is developed in the context of CMU's Spacecraft Design-Build-Fly graduate-level course, in collaboration between the Carnegie Mellon Nanosatellite Lab and the Instituto Superior Técnico NanoSatLab. The spacecraft is designed to make spacecraft development accessible: Python flight software, COTS-based hardware, and open-source documentation.