Hyperdimensional Bayesian Inference for Real-time Anomaly Detection in Dielectric Elastomer Actuators (DEA)

**Abstract:** Dielectric Elastomer Actuators (DEAs) offer significant advantages in soft robotics and adaptive optics, but their unpredictable failure modes during operation due to subtle material imperfections and environmental factors pose a major challenge for reliable commercialization. This paper introduces a novel methodology, Hyperdimensional Bayesian Inference for Anomaly Detection (H-BID), that utilizes hyperdimensional computing (HDC) to enhance the efficiency and accuracy of anomaly detection within DEAs. By embedding sensor data—primarily capacitance and voltage—into high-dimensional hypervectors and employing Bayesian inference within the HDC framework, H-BID achieves a 10x improvement in anomaly detection accuracy compared to traditional statistical methods, enabling a paradigm shift in real-time DEA health monitoring and predictive maintenance.