Evaluation of a Distributed Kalman Filter for Autonomous Satellite Navigation Using DASEE

Distributed Application Simulation and Evaluation Environment (DASEE) is an edge computing testbed built on a cloud-native software platform. Distributed applications at the edge are resource constrained and failure-prone, have disadvantaged communications, often require local autonomy, and exhibit non-deterministic global behavior. DASEE was created to support mission concept of operations (CONOPS) development, architecture trades, and demonstration of distributed compute algorithms in a realistic test environment. DASEE was recently used to simulate the performance of a distributed Kalman filter (DKF) for autonomous satellite navigation in a 112-node Low Earth Orbit (LEO) spacecraft constellation. Preliminary results show that the DKF algorithm converges in a centralized simulation but clearly diverges when executed in a decentralized simulation. Divergence is caused by the covariance matrix some-times losing positive-definiteness. Root cause of the divergence is traced back to asymmetric partner updates between nodes, in which one node successfully completes the update while its partner fails. This result demonstrates the need for an edge computing simulation environment for distributed computing applications.