주행 데이터 학습을 통한 주행 성향 판단 및 경로 예측

Abstract

While there are various drivers on the road and it is difficult to predict the future behavior of the vehicle because of different driving style even in the same situation. Therefore, it is necessary to judge the driving style and predict the trajectory using the driving style. In this paper, an algorithm for judging the driving style and predicting the trajectory of the vehicle by reflecting the style is proposed. The proposed algorithm is constructed by integrating the driving style judgement network and the trajectory prediction network. The driving style is classified into normal, aggressive, and careless driving, and CAN-bus data is used to determine the driving style. In addition, trajectory prediction is performed using a conditional generation model based on the determined style and driving data. The driving data is collected using CarMaker/HIL and the performance of the algorithm is verified in simulations with the proposed network structure.