Machine Learning methods are transforming many aspects of engineering by enabling the design
and operation of systems that feature more advanced objectives thanks to the methods' predictive,
autonomous, and data-driven capabilities. Cube/Nano-satellites are a type of aerospace system
especially positioned to benefit from the use of these methods, given their very limited budget
and onboard computational power. This article presents the application of the technique known
as Compressed Sensing to the estimation of the inertial parameters of UC3M's ST3LLARsat1
"Boira" CubeSat during the detumbling phase. Its physical characteristics, detumbling controller,
sensor/actuator models, and orbital parameters are implemented and simulated on an open-source
CubeSat simulator, and a Monte Carlo campaign is performed randomly varying the initial conditions.
The results are very promising, with a maximum estimation error of 1%.