http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
도로 참여자의 경로 예측을 위한 도메인 적응 및 일반화 기술의 필요성에 대한 연구
박노경(Nokyung Park),김진규(Jinkyu Kim) 대한전자공학회 2023 대한전자공학회 학술대회 Vol.2023 No.6
Trajectory Prediction has become an essential component of autonomous driving system in recent times. However, despite the availiability of state-ofthe-art models, their performance drops significantly when evaluated on unssen city data. In the real world, it is impossible to gather data about all cities, and people often find themselves visiting a city for the first time. In this paper we explicitly demonstrate the performance drop that occurs when there is a distribution difference between the training and evaluation datasets. Additionally, we highlight the need for safe autonomous vehicles and proposes a solution to mitigate the performance drop by applyig domain adaptation techniques, such as DANN and VREx, to the existing trajectory prediction models.