A Real-Time Robust SLAM for Large-Scale Outdoor Environments
Résumé
The problem of simultaneous localization and mapping (SLAM) is still a challenging issue in large-scale unstructured dynamic environments. In this paper, we introduce a real-time reliable SLAM solution with the capability of closing the loop using exclusive laser data. In our algorithm, a universal motion model is presented for initial pose estimation. To further refine robot pose, we propose a novel progressive refining strategy using a pyramid grid-map based on Maximum Likelihood mapping framework. We demonstrate the success of our algorithm in experimental result by building a consistent map along a 1.2 km loop trajectory (an area about 100,000 m2) in an increasingly unstructured outdoor environment, with people and other clutter in real time.
Origine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...