Large scale graph-based SLAM using aerial images as prior information
Robotics: Science and Systems V
Robot navigation within large-scale scenarios in general requires a globally consistent map of the environment, e.g., for reaching target locations safely. This problem has been addressed by many researchers and is usually referred to as Simultaneous Localization And Mapping (SLAM). However, existing solutions to SLAM do typically rely on loop-closures in order to maintain global consistency, and more importantly, do not exploit prior information when available. In this paper we present a novel
... we present a novel SLAM approach gaining long-term global consistency by utilizing publicly accessible aerial photographs as prior information. This is achieved by inserting correspondences found from local 3D laser scanner observations with the aerial image as constraint points into a graph representation of the SLAM problem. Data association between local laser observations and the aerial image is achieved by matching features from the raw 3D point clouds with features extracted from the aerial image. The proposed method has been validated on a large dataset acquired in mixed indoor and outdoor environments by comparing global accuracy with state-of-the-art SLAM approaches and GPS.