Speaker
Description
Abstract— Autonomous fault and anomaly detection is critical for ensuring the safety and success of space missions, addressing the limitations of ground-based analysis due to bandwidth constraints and operational delays. The conventional approach in Space Operations involves using Out-of-Limits (OOL) alarms for anomaly detection, which may prove insufficient in identifying and responding to complex anomalies or unforeseen novelties within the range of nominal values [1], [2]. In our previous work [3], we proposed a Machine Learning (ML) approach for on-board telemetry anomaly detection that addresses these limitations. We demonstrated a proof-of-concept integration of a TensorFlow model onto a radiation-tolerant LEON 3 processor and benchmarked various unsupervised and semi-supervised techniques with respect to their performance, memory footprint, and runtime. Our recent advancements focus on bridging the gap between a proof-of-concept solution and a nearly production-ready system. Mainly, we focused on preparing a semi-automatic pipeline for model training and deployment, experimented with other types of machine learning models, and created a patch for TensorFlow Lite for Microcontrollers (TFLM) which allows integration to the LEON 3 processor while still following the guidelines regarding software safety. Additionally, we incorporated uncertainty quantification (UQ) techniques to provide a more reliable assessment of the black-box model’s outputs.
Index Terms— anomaly detection, novelty detection, machine learning, LEON processor, TensorFlow, TFLM.
