Navigation and localization for autonomous vehicle at road intersections with low-cost sensors

This paper addresses the problem of navigation and global localization for intersection driving with an autonomous vehicle equipped with low-cost sensors in urban environments. The intersection driving is an important part of outdoor autonomous navigation, when it travels along structured roads and needs to reach the intended destinations. Previous almost approaches are based on the high-accuracy localization by using expensive high-performance GPS/ INS sensors. In this paper, we propose a novel approach to enhance the position accuracy of vehicle using the relative position from vehicle to lane and stop-line makings instead of them. When the vehicle autonomously approaches at the intersection, it first is able to know where the vehicle approximately is on the digital map and what the global position of lane and stop-line in the nearest intersection using a GPS and Odometer equipped. The current global position of vehicle is calculated by adding a lateral/longitudinal offset value by means of lane and stop-line detection to global position of them above obtained. This paper presents a way of generation for the optimal feasible path using Bezier Curve to safely avoid obstacles and across the intersection. We demonstrated the validation of these methods by performing experiments with a custom built autonomous vehicle using experiments conditions based on an actual intersection driving in our test site.