Deputy-Centric Fault Detection and Isolation for Fault-Tolerant SmallSat Swarms

Autonomous SmallSat swarms are a natural fit for imaging fast, transient targets such as near-Earth asteroids and cislunar traffic, but short science windows and limited ground contact make onboard Fault Detection and Isolation (FDI) a primary mission constraint. This paper develops and evaluates a deputy-centric FDI architecture for mothership-plus-swarm missions built on Hill-frame relative navigation. Each deputy carries a common mid-rate Estimation bus that runs a Hill-frame EKF using Line of sight-only relative measurements and cross-links, achieving tens-of-metres RMS position accuracy across an Apophis-class flyby and outperforming a GNSS-like range+LOS configuration under the assumed sensor models. On top of this bus, a lightweight, filter-agnostic FDI core ingests Extended kalman Filter (EKF) states and covariances, reconstructed burn histories, attitude-tracking errors, and comm heartbeats to form six residual channels (NAV, ATT, PROP, COMM, FORM, PEER). These are processed by leaky CUSUM detectors and a persistence-based state machine that classifies each deputy as HEALTHY, SUSPECT, FAULTED, or DEAD with an associated fault label and confidence, yielding a compact health vector suitable for burn gating and swarm reconfiguration. The approach is demonstrated on an eight-deputy ARISE-class swarm executing a high-speed Apophis encounter. An offline FDI sweep shows that, in the nominal case, all deputies remain HEALTHY while flyby-aware FORM and PEER residuals confirm that the fleet meets terminal stand-off constraints without outliers. A Monte Carlo campaign with $2\times10^4$ trials per scenario injects single-deputy NAV, ATT and PROP faults at random phases along the encounter. For representative NAV and ATT faults the FDI core achieves near-unity detection probability with no observed misses and detection delays of order one to two minutes, while maintaining low false-alarm rates on healthy deputies. These results indicate that the proposed deputy-centric FDI layer can provide timely, reliable health cues to downstream fault-tolerant trajectory planners for future cislunar SDA, debris-monitoring, and NEO-reconnaissance missions.