Benchmarking Onboard Science Data Retrieval Algorithms on the Snapdragon Platform

The coming decade of planetary missions will havestringent data requirements. Optimizing limited bandwidth is amajor motivation for the use of onboard science data analysisalgorithms for deep space missions. Under the traditionalmission operations paradigm, all data is sent to Earth beforefollow-on activities are planned by ground operators. Thisoperational model creates challenges for time and bandwidthconstrained mission events. The Near Earth Asteroid Scout(NEAScout) mission is a prime example of a mission with thesechallenges. NEAScout is an interplanetary CubeSat manifestedon Artemis-1. The mission will use a solar sail to navigate to aNear Earth asteroid and perform flyby reconnaissance imaging.Vehicle size, configuration, trajectory and distance from Earthall impose bandwidth constraints. To minimize theselimitations, the Project decided to move image processingcapabilities onboard the spacecraft. This onboard sciencesoftware enables image calibration, image coaddition, imagesubtraction, data compression, downsampling and cropping allonboard the spacecraft. The goal is to reduce the data downlink,without compromising science contents. NEAScout utilizes theSPHINX computing platform, whose processing performance isakin to the RAD750, but tailored for CubeSat resources.Although sufficient to enable these sorts of algorithms, theSPHINX takes multiple minutes per image, which diminishesthe prospect of using image content for rapid responsereasoning. This effort ported, characterized and benchmarkedthe performance of the NEAScout science software on theSnapdragon platform, a next-generation higher performingprocessor. Between 10x and 200x processing time speedup wasachieved, depending on the image processing algorithm, with a50x speedup for a notional image processing pipeline. This workhighlights new performance baselines for onboard science dataprocessing algorithms which missions could expect to achievewith modern computing hardware. Future avionics computerswith Snapdragon processors would enable onboard science dataanalysis to be used in more time constrained mission scenariosand with larger onboard data sets.