Title: Heterogeneous On-board Processing for Distributed Measurement and Multi-Satellite Missions
Presenting Author: Matthew French
Organization: USC Information Sciences Institute (ISI)

Abstract:
While prior research demonstrated that heterogeneous on-board computing solutions could deliver an increase of 100 to 1,000x performance for Earth science applications that stressed higher resolution instruments and persistent measurements, emerging missions which emphasize distributed measurement and multi-satellite operations are expected to increase onboard computing requirements by an additional 10 to 100x. This research extends the on-board computing analysis framework developed in our AIST-14 effort to address the needs of emerging distributed measurement and multi-satellite NASA Earth Science missions. Next generation missions invoke requirements for distributed sensing, data continuity, satellite constellation station keeping, and intelligent sensor control. Distributed sensing missions require application portability across platform types; however, platform constraints lead UAVs and satellites to utilize different co-processor accelerators (GPUs vs. FPGAs). A common framework that supports both FPGAs and GPUs will facilitate migration between these platform types. Multi-satellite missions enable diurnal and multi-angle measurements, and invoke complex communication and control logic that must be processed on-board. Intelligent sensor control creates complex, ad hoc processing, which necessitates experimentation on prototype hardware. This presentation provides an overview of our research to address these challenges by extending the evolvable testbed to include GPU support, model distributed sensors and high bandwidth communication links, develop prototype hardware, and demonstrate the technology on distributed sensing applications such as FluidCam or MiDAR.