Deep Neural Network를 이용한 자율 주행 차량의 차선 유지 시스템 성능 향상

Abstract

In this paper, we propose the method to compensate for unmatched disturbance caused by lateral control of autonomous vehicles using deep neural network. We utilize dynamic lateral model based lane keeping system considering look-ahead distance for autonomous driving vehicle. On the curved road, it makes it difficult to control the system because of unmatched disturbance caused by external signals. To mimic the strategy of human driver, The deep neural network computes an additional control input added to a conventional control system. The proposed algorithm is verified in MATLAB/Simulink environment and the error of state is reduced in the curved road.