GPS-Based Localization of Autonomous Vehicles

Abstract

Global Positioning System (GPS)-based positioning is commonly used for the localization and navigation of autonomous vehicles for intelligent transport systems and for mobile robotic applications. The reasons for the widespread use of GPS for positioning is due to its low cost, easy availability, and ease in use. Although the accuracy of GPS-based positioning is acceptable in most cases, in cities where the mobile robot or the autonomous vehicle is moving in an environment surrounded by an urban canyon of buildings, trees, and dynamic obstacles, the GPS positioning errors are of the order of tens of meters, which is not acceptable in many applications. In addition, in GPS-denied places like tunnels, subways, and canopies, GPS positioning must be augmented using vision-based methods. In this chapter, the need for localization for autonomous vehicles, the role of GPS, its issues and challenges, and the methods to correct for GPS positioning errors have been discussed. The GPS positioning errors are caused by various factors like multipaths, atmospheric delays, clock-offset errors, reflection of signals from surrounding building walls, etc. A statistical approach based on k-means clustering and quartiles to reduce errors due to multipaths is discussed. Also, a geometric approach using a laser range sensor that scans the three-dimensional (3D) environment around the GPS antenna mounted on the top of an autonomous vehicle and identifies the reflected GPS signals is discussed to correct for positioning errors due to reflections from the building facades. The use of wireless sensor networks and inertial navigation systems with GPS was found to enhance the localization accuracy in urban environments. The multisensory approach toward positioning for localization of vehicles is also discussed. The computer vision-based methods are quite helpful in localization where GPS signal is not reachable. This chapter also discusses how computer vision-based and artificial intelligence-based methods augment the localization task in the navigation of autonomous vehicles.