Multivariate Subspace Analysis for Sensor Fusion in Subsurface Planetary Drilling Fault and Anomaly Detection

In extraterrestrial planetary environments, environmental, computing, energy, and operational constraints require robotic agents to complete tasks without human supervision. Specialized extraterrestrial robotic drilling agents like TRIDENT currently lack autonomous capabilities, primarily because research into subsurface environment representation for autonomous drilling remains in its early stages. We build upon previous field-validated work using subspace analysis methods to estimate drilling faults from drill avionics telemetry, and extend it to a multivariate subsurface anomaly detection method that leverages sensor fusion of avionics telemetry from extraterrestrial drilling robots. This method was evaluated at Mars planetary analog sites in Antarctica.