부산대학교 도서관

While automated driving systems (ADSs), which correspond to level-two automated driving of Society of Automotive Engineers (SAE), are being developed actively, the ADS that learns the preference of drivers at an opportune time is required. In addition, there is a lack of assessments on ADS comfort. In this study, we focused on collision avoidance and developed an ADS that learns driver preference by driver intervention. By applying the gain-tuning and learning methods for gradual learning, the ADS can stably modify its behavior at an opportune time. The results of a driving simulator experiment indicated that driver comfort improved by using the proposed ADS. In addition, we confirmed that the ADS can guarantee usability during authority transfer in the learning process.