Reference Scan Matching for Global Self-Localization
2001 Technical Report CAROL-Project
University of Kaiserslautern, Computer Science Department
Joachim Weber, Lutz Franken, Klaus-Werner Jörg, Ewald von Puttkamer
Especially in dynamic environments a key feature concerning the robustness of mobile robot navigation is the capability of global self-localization. This term denotes a robot's ability to generate and evaluate position hypotheses independently from initial estimates, by these means providing the capacity to correct position errors of arbitrary scale. In the self-localization frame described in this report, the feature based APR scan matching algorithm provides for each new laser scan several possible alignments within a set of reference scans. The generation of alignments depends on similarities between the current scan and the reference scan, only, and is unconstrained by earlier estimates. The resulting multiple position hypotheses are tracked and evaluated in a hybrid topological/metric world model by a Bayesian approach. This probabilistic technique is especially designed to integrate position information from different sources, e.g. laserscanners, computer vision, etc.. Experimental results are given for robust self-localization in a partly monotonous environment based on a combination of odometry, APR and neural image classification.
Back to Publications