Hyperdimensional AI for Autonomous Anomaly Response in Smart Habitats

This proposal introduces a novel brain-inspired anomaly detection and response framework based on Hyperdimensional Computing (HDC) for deployment in autonomous deep-space smart habitats. Unlike conventional AI models that require high-power GPUs and retraining for new anomalies, the proposed HDC-based system enables ultra-low-power, real-time, and explainable fault detection, isolation, and recovery (FDIR) capabilities. HDC’s distributed, symbolic vector representations provide exceptional robustness to noise, adaptability to previously unseen failures, and human-interpretable reasoning—critical for supporting astronaut-autonomy collaboration in long-duration space missions. Phase I funding will be used to design, implement, and validate an HDC prototype that performs onboard anomaly detection across multi-modal spacecraft telemetry, supports one-shot learning, and operates on reconfigurable hardware like FPGAs and embedded processors. The system will undergo rigorous fault injection testing and hardware-in-the-loop validation using NASA-relevant datasets to ensure operational resilience and real-time responsiveness in spaceflight conditions. Target markets include NASA, DoD, and commercial space companies requiring intelligent onboard health monitoring systems, as well as broader terrestrial markets in industrial IoT, aerospace, and critical infrastructure. This innovation aims to enhance mission autonomy, reduce operational costs, and extend AI-driven fault management to edge devices in both space and Earth-based applications.