Calculation of likelihood based on quasi-euclidean norm for robust particle-filer-based self-localization

Abstract

In this paper, we propose to calculate a likelihood based on a quasi-euclidean norm for a fast and robust particle-filter-based self-localization of mobile robots. The quasi-euclidean norm can be calculated faster than the conventional L₂ norm. Furthermore, such norm is much robust to noises than the L₁ or L₂ norm. In this work, we perform self-localization experiments based on the L₁ norm, L₂ norm and quasi euclidean norm likelihood to compare the robustness and processing time. Through these experiments, we confirm the effectiveness of our method.