Building Hybrid B-Spline And Neural Network Operators

Control systems are critical in ensuring the safety of cyber-physical systems (CPS) across domains like airplanes and missiles. Safeguarding CPS necessitates runtime methodologies that continuously monitor safety-critical conditions and respond in a verifiably safe manner. Many real-time safety approaches require predicting the future behavior of systems. However, achieving this requires accurate models that can operate in real time. Inspired by DeepONets, we propose a novel approach that combines B-splines’ inductive bias with data-driven neural networks (NNs). Our hybrid B-spline neural operator serves as a universal approximator, validated on a 6DOF quadrotor.