도심 자율주행 제어시스템을 위한 컨볼루션 신경망 기반 도로 차선모델 추정

Abstract

The lane keeping system (LKS) simulates the shape of the road as a cubic polynomial through a camera sensor. The control input, steering wheel angle, is calculated to follow the reference based on the vehicle motion model using the road coefficients. LKS can maintain the lane safe only when the lane information is valid and cannot guarantee the stability of the control system when there is no lane information at all or the sensor fail for a long period of time. In this paper, we propose a parallel deep convolutional neural network (P-DCNN) that generates the road lane model for urban driving environments. The data for training and validation is collected by our in-vehicle logging systems and experiment was performed through the computational simulation.