Development of a High-Performance, Heterogenous, Scalable Test-Bed for Distributed Spacecraft

To address modern mission challenges, high-performance computation devices (such as SoCs, GP-GPUs, FPGAs/ASICs, and neuromorphic processors) have become in-creasingly common in spacecraft while simultaneously, satellite constellation size continues to grow. The matched increase in spacecraft capability and cardinality requires concepts of operation that embrace a level of autonomy traditionally avoided in the space domain while tackling the additional complexity and uncertainty of controlling distributed systems. Development of these distributed autonomy capabilities necessitates multi-agent testing at scale for Validation & Verification (V&V) since the group's behavior is inextricably linked to the interactions among the individuals. Under this paradigm, characterizing emergent behavior in large, computationally-intensive satellite networks is paramount to mission success. The Distributed Spacecraft Autonomy (DSA) project at NASA Ames Research Center is maturing technology required for larger, autonomous constellations through orbital deployments and simulation studies. Previous work detailed the DSA simulation and test framework, capable of developing, testing, and verifying a four-spacecraft mission in a single developer environment using open-source software. DSA technology is also applied to a Lunar Position Navigation and Timing (LPNT) scenario with 100 spacecraft to further advance the state-of-the-art. Whereas the flight test hardware provided higher fidelity testing of the four-node DSA application before flight, no equivalent set of 100 engineering units exists for these large theoretical swarms, and the existing test framework falls short of the required number of participating agents by an order of magnitude. A higher fidelity test platform capable of scaling to 100 independent agents is required to support the development and characterization of distributed autonomy in large swarms of spacecraft. DSA is building a heterogenous, many-node, Processor-in-the-Loop (PiL) testbed to aid the development and verification of scalable distributed autonomy capabilities for multi-spacecraft missions, including LPNT, heliophysics, space situational awareness, and collision avoidance. Our three rigorously selected platforms represent a range of computationally-intensive processors. Of the 100 PiLs, 85% are high-performance nodes, 40% are real flight hardware engineering units, and all have been deployed to orbit. Here, we detail the design, implementation, and capability of DSA's 100 processor-in-the-loop scalability testbed, the Distributed Intelligent Spacecraft Simulation Test RACK (DISSTRACK).