Keynote lecture: Holistic perception for active safety in intelligent vehicles

Summary form only given. Designing fully autonomous robotic vehicles which can drive on regular roads does not require models of drivers and how they interact with vehicles. In contrast, design of advanced intelligent driver assistance systems for active safety for preventing accidents requires human behavior understanding, human-vehicle interaction modeling and human intent prediction. We will present an overview of a “human-centered” framework for a distributed intelligent system with the driver, vehicle and environment as three key components. We will emphasize the need and implications of utilizing a holistic approach where driving in naturalistic context is observed over long periods to learn driving behavior and to predict driver intentions and interactivity patterns. The approach necessitates design and development of novel vision systems for simultaneously “Looking-In and Looking-Out” (LiLo) of a vehicle. The presentation will include presentation of selected ongoing studies and will conclude with some pointers to important outstanding research challenges.