딥러닝 기술을 적용한 운전자 맞춤형 긴급자동제동시스템 파라메터 추출 (Deep Learning based Customized Autonomous Emergency Braking System Parameter Extraction)

서지정보

· 발행기관 : 제어·로봇·시스템학회
· 수록지 정보 : 제어.로봇.시스템학회 논문지 / 25권 / 8호 / 671 ~ 676페이지
· 저자명 : 김창원, 서준호, 이동규, 추준욱, 권오원

초록

The autonomous emergency braking system is one of advanced driving assist systems. It is an advanced safety functiondesigned to prevent collision with a forward vehicle or a pedestrian in the event of driver’s carelessness or a sudden accident ahead.
Because the AEB stops the vehicle with a maximum deceleration command at the last possible time to brake so that collision avoidance ispossible, it can prevent a collision with a preceding vehicle. However, there are risks of a secondary collision with a trailing vehicle andinjury to a passenger due to sharp deceleration. In addition, sudden braking control can present a driving feeling gap to the driver. Theseproblems must be solved because they can amplify a user's rejection of automatic emergency braking (AEB) and negatively affect theacceptability of this technology. In this study, the driver's braking propensity is derived by means of deep learning and it is applied to acustomized AEB system to prevent secondary accidents caused by sudden braking and to improve user acceptance. In order to derive adriver's braking behavior, normal driving data and stationary target based braking test data are collected. Utilizing a deep neural network,the driver’s braking profile during ordinary driving and a sudden stop are extracted. The correlation between the two extracted driver’sbraking behaviors is analyzed. This provides the means to derive sudden braking behavior as a function of driving speed, leading tocustomized AEB parameters.

영어초록

The autonomous emergency braking system is one of advanced driving assist systems. It is an advanced safety functiondesigned to prevent collision with a forward vehicle or a pedestrian in the event of driver’s carelessness or a sudden accident ahead.
Because the AEB stops the vehicle with a maximum deceleration command at the last possible time to brake so that collision avoidance ispossible, it can prevent a collision with a preceding vehicle. However, there are risks of a secondary collision with a trailing vehicle andinjury to a passenger due to sharp deceleration. In addition, sudden braking control can present a driving feeling gap to the driver. Theseproblems must be solved because they can amplify a user's rejection of automatic emergency braking (AEB) and negatively affect theacceptability of this technology. In this study, the driver's braking propensity is derived by means of deep learning and it is applied to acustomized AEB system to prevent secondary accidents caused by sudden braking and to improve user acceptance. In order to derive adriver's braking behavior, normal driving data and stationary target based braking test data are collected. Utilizing a deep neural network,the driver’s braking profile during ordinary driving and a sudden stop are extracted. The correlation between the two extracted driver’sbraking behaviors is analyzed. This provides the means to derive sudden braking behavior as a function of driving speed, leading tocustomized AEB parameters.

참고자료

· 없음