Machine Learning Benchmarking for Spacecraft Processors During Spacecraft Rendezvous and Proximity Operations

This work benchmarks explicit computation for Machine Learning (ML) applied to spacecraft Rendezvous and Proximity (RPO) mission. In general, spacecraft processing is severely limited, which contrasts demanding computational resources via AI/ML techniques. Using an off-the-shelf NIVIDIA Jetson Orin baselined for several Cubesats, this works measures traditional computation metrics such as execution time as well as non-traditional metrics such as memory and CPU. These are measured as the algorithm is implemented recursively to achieve its RPO goal. The result quantifies explicty the resources needed to be implemented on a spacecraft.