Development of AI-powered Predictive Model via Convolutional Recurrent Network for Mobility Application

Technology development and research are rapidly expanding into eco-friendly hydrogen/electric, autonomous vehicles, and urban air mobility as the advancement of the future mobility industry accelerates. Active noise control (ANC) is one of the advanced technologies to provide a quieter interior environment of mobility. Recently, the ANC has been applied to mass production vehicles and is expected to become an essential technology for the future mobility industry. The FxLMS (Filtered-x Least Mean Square) al gorithm is a key method for active noise control. Although the stability and performance of the FxLMS algorithm using a linear adaptive filter have been verified through a number of studies, some nonlinear systems have limitations in their application. Herein, a deep learning-based network model can be applied, which has advantages in processing input/output data in nonlinear relationships. This can improve the accuracy of active noise control algorithms. In this paper, a nonlinear active noise control model based on a convolutional recurrent network (CRN) will be presented. The performance of CRN will be evaluated by predicting interior noise from acceleration signals measured through actual driving tests.