부산대학교 도서관

Automatic vehicle following and cruising under all road geometry characteristics (curvature, slope, and superelevation) and traffic conditions without putting the vehicle occupants or other vehicles in unsafe situations are fundamental for the success of autonomous vehicles. Despite the advances in sensor technologies, we do not have the accuracy needed to guarantee safety under all driving conditions. In fact, for certain road geometries such as going around sharp curves, forward-looking beams and vision cannot see what is around the curve no matter how accurate they are. To ensure safety and driving comfort, the following challenges need to be addressed under all road geometry characteristics and traffic conditions: Effective detection of in-path objects, collision avoidance with objects not visible with exteroceptive sensors, adopting a cautious approach in high noise environment and avoiding high longitudinal and lateral accelerations for driving comfort. This paper presents a methodology and control design that meets these challenges. It incorporates geo-referenced maps to identify road geometries and detect in-path objects. It analyses the impact of noise on sensor data and uses a probabilistic approach to determine the desired velocity. It then proposes a vehicle following controller that achieves robust vehicle following and cruising under all road geometries and in particular sharp curves. In the case of sharp curves, where a possible obstacle is not visible by the on-board sensors, slowing down to a certain speed preemptively to have time to stop is part of the proposed methodology. Simulations show that proposed system fulfills all safety and comfort constraints.