BroncoSat-1: Enabling Technologies for On-orbit Artificial Intelligence Systems

BroncoSat-1, the flagship project of Cal Poly Pomona’s Bronco Space satellite program and the university’s first space mission, aims to test and demonstrate the use of a novel commercial-off-the-shelf (COTS) computing platform, the NVIDIA Jetson, to improve on-orbit operations of a CubeSat and open the possibility for implementing artificial intelligence (AI) in space. Access to space and the supply for space launches are at their highest levels, yet their demand has not been able to sustain the market as small satellites cannot be developed quickly enough. CubeSats, a standard of small satellites created by Cal Poly San Luis Obispo and Stanford University based on 10 cm³ units or “U’s,” are popular avenues for commercial and academic space programs like Cal Poly Pomona’s Bronco Space to conduct scientific experimentation and technological demonstrations using custom and COTS parts. However, the complexity of some missions in the CubeSat form factor often prolongs development schedules, while the constraints of common 1U and 3U satellites limit mission operations due to reduced onboard communication, computation, and electrical generation capabilities. The ARM CPU of the Jetson provides greater speed and flexibility than conventional microcontroller-based systems and operates with low power consumption. The desktop-class graphics processor on the Jetson can accelerate large data set calculations and analysis for image processing and tensor operations for machine learning algorithms. BroncoSat-1 shall execute multiple premade and homebrewed benchmark tests to verify its AI performance while in orbit and transmit the results via a Globalstar communication system. Critical operations of the satellite shall be carried out via a modified PyCubed, an open-source flight computer architecture, tailored to the needs of BroncoSat-1 and named “PolyCubed.” The mission can be extended to characterize the low-Earth orbit environment to create models for AI-based attitude determination and control algorithms for future CubeSat testing at Cal Poly Pomona’s Space System Laboratory, currently under development. From concept to delivery in less than a year, through the adversity of the COVID-19 pandemic, and with a launch planned for 2022, BroncoSat-1 would be among the first satellites in space using the NVIDIA Jetson platform and implementing AI for on-orbit operations. Upon verification that the Jetson is a qualified technology, further research in Jetson-based systems will be done at Cal Poly Pomona, contributing an option for other university space programs to build CubeSat and small satellite missions around. Cal Poly Pomona and Bronco Space are already working on their first use of the Jetson as the primary flight computer for the Bronco Ember project, one of three winning submissions in the 2021 NASA TechLeap competition.