Probabilistic multi-component extended strong tracking filter for mobile robot global localization

This paper proposes a multi-component extended strong tracking filter (MESTer) for global localization. It is the first time strong tracking filter (STF) is introduced into robotics domain and is fundamentally extended to be suitable for fusing observations with arbitrary time-varying dimensionality, based on equivalent space transformation and extended orthogonality principle. The resulted extended strong tracking filter (ESTF) is then combined with a probabilistic multi-component evolving mechanism and finally forms the MESTer localization method. Real robot experiments and comparisons with existing methods show that MESTer has high convergence speed, computational efficiency and definite robustness to sensor noises, kidnapped robot problem, system nonlinearities, and symmetric environments.