Hybrid localization using the hierarchical atlas

A unified Bayesian framework for global localization and SLAM in hybrid metric/topological maps

“…The third class (Lisien et al 2005;Tully et al 2007;Werner et al 2009a,b;, to which we believe our work is most closely related, associates vertices in the topological graph with true anchored points in the environment with physical meaning that the robot can precisely navigate to in order to deterministically transition to a new subspace in the global map. Thus, the representative topology is unique for the environment, is repeatable upon subsequent experiments, and is, most importantly, stable upon revisiting vertices via sensor-based control.…”

“…Thus, topological state space representations generally scale well with environment size. A common method is to use topological features or fingerprints to heuristically localize a robot to a vertex in the graph (Tully et al 2007;Choset and Nagatani 2001). In topologically rich environments, though, similarity in topological vertex characteristics is common, and additional information is necessary to resolve these ambiguities.…”

“…Existing solutions typically rely on one of three map representations: obstacle-based grid maps or particle maps (Burgard et al 1996(Burgard et al , 1998Fox et al 1999b;Grisetti et al 2005;Eliazar and Parr 2003), featurebased metric maps (Smith et al 1990;Dissanayake et al 2001;Montiel et al 2006;Civera et al 2007;Tully et al 2008Tully et al , 2010Montemerlo et al 2002Montemerlo et al , 2003, and topological maps (Choset and Burdick 1995;Nourbakhsh et al 1995;Simmons and Koenig 1995;Cassandra et al 1996;Choset and Nagatani 2001;Dudek et al 1996;Tomatis et al 2002;Ranganathan et al 2006;De et al 2008;Kuipers and Byun 1991;Tully et al 2007;Lisien et al 2005;Blanco et al 2008Blanco et al , 2007Angeli et al 2008;Kouzoubov and Austin 2004). Arguably, the most efficient of the three is topological mapping, as it concisely represents an environment as a graph, whose vertices are interesting "places" and whose edges represent the paths between them (Kuipers and Byun 1991).…”

“…In (Tully et al 2007), we focused on hybrid filtering for metric/topological localization. This original work did not solve the full SLAM problem, only intermittently updated the robot's pose estimate, and could not robustly recover from unmodeled disturbances.…”

UWB-assisted real-time localization in wireless sensor networks

“…In our work we are trying to adopt the UWB in addition to the existing RSSI measurement to form a hybrid localization scheme. Hybrid localization is a term that has been adopted in some existing work, such as [13][14] [15], where most of them focus on higher level integration of hybrid data post-processing methods or localization algorithms rather than combining different ranging approaches. Solution in [16] integrates various ranging techniques, but only uses them separately in different areas.…”