Backing up GPS in urban areas using a scanning laser

Abstract

This paper studies the use of lidar for the egolocalization of car-like vehicles in conjunction with GPS. We consider a laser scanner installed at the front of a vehicle that detects both sides of the road. We present a method that fuses the lidar information to improve the localization process by providing additional exteroceptive information and by rejecting bad GPS fixes. The strategy is inspired by a SLAM paradigm and is efficient if the vehicle navigates often in the same area. We present the different main stages of such a strategy: lidar data processing, map data representation, and augmented Kalman filtering scheme. Finally, experimental results are reported to illustrate the performance of this approach.