Map-based precision vehicle localization in urban environments

Abstract

Many urban navigation applications (e.g., autonomousnavigation, driver assistance systems) can benefitgreatly from localization with centimeter accuracy. Yet suchaccuracy cannot be achieved reliably with GPS-based inertialguidance systems, specifically in urban settings.We propose a technique for high-accuracy localization ofmoving vehicles that utilizes maps of urban environments. Ourapproach integrates GPS, IMU, wheel odometry, and LIDAR dataacquired by an instrumented vehicle, to generate high-resolutionenvironment maps. Offline relaxation techniques similar to recentSLAM methods [2, 10, 13, 14, 21, 30] are employed to bringthe map into alignment at intersections and other regions ofself-overlap. By reducing the final map to the flat road surface,imprints of other vehicles are removed. The result is a 2-D surfaceimage of ground reflectivity in the infrared spectrum with 5cmpixel resolution.To localize a moving vehicle relative to these maps, we present aparticle filter method for correlating LIDAR measurements withthis map. As we show by experimentation, the resulting relativeaccuracies exceed that of conventional GPS-IMU-odometry-basedmethods by more than an order of magnitude. Specifically, weshow that our algorithm is effective in urban environments,achieving reliable real-time localization with accuracy in the 10-centimeter range. Experimental results are provided for localizationin GPS-denied environments, during bad weather, and indense traffic. The proposed approach has been used successfullyfor steering a car through narrow, dynamic urban roads.