딥 러닝을 이용한 노면 거칠기 분류 로직 개발

Abstract

It is important to classify road condition to establish suspension control strategy. Today, using vertical acceleration sensor of sprung mass and unsprung mass is most commonly used method to classify road condition. However, when using only vertical acceleration sensor, it is hard to classify exact road condition. So, in this study, the logic which is used deep learning to classify road condition is proposed. Deep learning is a technology used to classify objects or data by learning artificial neural networks designed with multi-layer structure. It has advantage of increasing accuracy of road condition classification result by considering vehicle’s IMU sensors at once. IMU sensors are used as features which are deep learning input. To reduce network size and secure classification accuracy, feature selection method is proposed