Personal Driving Style-based ADAS Customization in Diverse Traffic Environments using SVM for Public Driving Safety

Abstract

The advancement of autonomous driving and Advanced Driver Assistance System (ADAS) technology leads to improved public driving. However, due to the lack of public trust in self-driving vehicles, the actual use of Autonomous Vehicles (AVs) is still surprisingly limited. Based on such needs, this paper proposes a new attempt for a customized ADAS taking each driver's driving style into account in order to make individuals feel much more comfortable with autonomous driving. In this paper, a novel customized ADAS algorithm using Support Vector Machine (SVM) with high classification accuracy is proposed, which categorizes drivers into assertive and defensive driving styles. Since the importance of ADAS parameters that affect driving propensity varies depending on the driving situation of each driver, this paper compares and analyzes the driving styles of drivers in three driving scenarios. Each driver's driving data is collected using CARLA, an Unreal Engine-based realistic simulator that can imitate real-world scenarios. Based on this precise categorization, the ADAS sensors can enable more advanced driving safety support. The suggested scheme in this paper is particularly significant since the present state-of-the-art in autonomous driving is at level 3, which calls for sophisticated and advanced functions that can assist drivers using ADAS technology.