Description: This collaborative effort proposes to develop an edge computing device that is specifically designed for CubeSats to maximize computing capability while also minimizing power and related thermal problems. The proposed computing system uses multiple low-power FPGAs, known as a cluster, housed in a 6U or larger CubeSat that can be independently reprogrammed on-orbit for new tasks without needing to stop the entire system. This allows for ongoing operation and adaptability to new tasks or changing conditions, a necessary capability for deep space missions. The system can interface with a variety of sensors such as cameras, software defined radios, and space weather instrumentation. The overall goal of this project is to develop and demonstrate a high-fidelity prototype of the reconfigurable computing system (hardware and algorithms) for two tasks: space weather awareness and maritime location of ships. A promising computing solution is beneficial to several SmallSat missions like Earth observation, space weather monitoring, communications, and autonomous operations. With the rise in complex CubeSat and SmallSat missions for beyond Earth exploration, there are several key factors that have boosted attention towards maturing edge computing for nanosatellites such as the increase in data volume for deep space nanosatellites, real-time decision making on the spacecraft, reconfigurability, spacecraft autonomy and reliability, and cost reduction.

Benefits: The proposed novel edge computing/ machine learning platform uses JPL’s F-prime spacecraft operating software capabilities and expands it to include machine learning capabilities, all the while running on low, commercially available Graphics Processing Units (GPUs). If successful, this technology could enable a whole new set of Machine Learning based functions in space for SmallSats. The cost-effective Edge computing/Machine Learning platform could enable low SWaP-C Machine Learning using COTS hardware and software aboard SmallSats, and significantly increase future adoption of Machine Learning aboard small spacecraft. A low power yet highly capable edge computing device is beneficial to numerous types of missions, particularly useful for dynamic environments where tasks and conditions can change rapidly.