MODEL-BASED MONITORING AND REFINING OF MODEL PREDICTIVE CONTROL SYSTEMS AND APPLICATIONS

Abstract

In various examples, the systems and methods of this disclosure can use a machine learning model (e.g., a Kolmogorov-Arnold network) to monitor and improve the performance of a model predictive control (MPC) system associated with a machine. For example, the machine learning model can be trained and used to predict errors between measured and predicted states of the machine. Based at least on the predicted errors, the systems can update one or more parameters of a predictive model associated with the MPC system to reduce the differences between the measured and predicted states in subsequent iterations.In some examples, the systems of the present disclosure can retrain the machine learning model using new data when the magnitudes of the predicted errors reach or exceed a threshold, thereby enabling the machine learning model to adapt to changing conditions in real time.