Robust Global Urban Localization Based on Road Maps

This paper presents a method to perform global localization in urban environments using segment-based maps in combination with particle filters. In the proposed approach the likelihood function is generated as a grid, derived from segment-based maps. The scheme can efficiently assign weights to the particles in real time, with minimum memory requirements and without any additional pre-filtering procedure. Multi-hypothesis cases are handled transparently by the filter. A local history-based observation model is formulated as an extension to deal with ‘out-of-map navigation cases. This feature is highly desirable since the map can be incomplete, or the vehicle can be actually located outside the boundaries of the provided map. The system behaves like a global localizer for urban environments, without using an actual GPS. Experimental results show the performance of the proposed method in large scale urban environments using route network description (RNDF) segment-based maps. Accurate localization is a fundamental task in order to achieve high levels of autonomy in robot navigation and robustness in vehicle positioning. Localization systems often depend on GPS due to its affordability and convenience. However, it is well known that GPS is not fully reliable, since satellite positioning is not available anytime, anywhere. This is the case of extreme scenarios such as underwater or underground navigation, for instance. In the context of urban navigation, GPS signals are often affected by buildings (‘urban canyon’ effect) blocking the reception or generating undesirable jumps due to multi-path effects. Fortunately, the use of a priori maps can help in the localization process. There are maps already available for certain environments, such as the digital maps used for road positioning. Moreover, accurate maps can be built using GIS tools for many environments, not only urban, but also off-road settings, mining areas, and others. Usually the vehicle position is evaluated by combining absolute information such as GPS with onboard sensors such as encoders and IMUs. Since GPS information is not permanently available, significant work has been carried out in order to integrate external sensors such as laser and sonar in the localization process such as in (Leonard & DurrantWhyte, 1991), (Guivant & Nebot, 2001). A priori information, such as digital maps, has been used to obtain accurate global localization, usually fusing information into Bayesian filters (Fox et al., 2001). Maps of the 14