Paper accepted to IEEE SMC 2023

A paper titled “Causal Deep Operator Networks for Data-Driven Modeling of Dynamical Systems” was accepted to the IEEE International Conference on Systems, Man, and Cybernetics (SMC) 2023. The paper proposes a novel physics-informed machine learning method for modeling dynamical systems. It enhances the standard deep operator network (DeepONet) by enforcing temporal causality, an innate physical property of dynamical systems, and adding the initial condition. Using a case study of learning a model of a room in a heating, ventilating, and air conditioning (HVAC) system, the proposed method was shown to outperform other popular machine learning methods, such as the standard DeepONet and the recurrent neural network (RNN), especially for multi-step predictions and predictive controls.