Neural Network Aided Kalman Filter to Maximize Accuracy

Abstract

With the growing need for data, and ever-growing demand for prediction and error-correction, Kalman Filters are undoubtedly at the forefronts of real-time estimation. While these filters are designed to achieve convergence shortly after getting exposed to the data, the filters might not be able to maximize all the data it can extract from the system. In order to extract most of the information that is otherwise unusable by the Kalman Factor, an initial assumption of a pre-trained Machine Learning model that correlates a feedable parameter with the unusable data is made. The feedable parameter is then given to the Kalman Filter along with the other standard parameters which boosts the accuracy by adding another dimension to the filter.