Decentralized Information Filtering Under Skew-Laplace Noise

Abstract

Localization in large sensor networks requires decentralized computationally efficient filtering solutions. To model challenging indoor propagation conditions, including non-line-of-sight conditions and other channel variations, it may be necessary to consider non-Gaussian distributed errors. In this case, Gaussian filters cannot be considered as is and particle filters do not meet the system requirements on computational cost and/or available memory. In this article we explore decentralized Gaussian information filtering strategies under skew-Laplace errors, exploiting the hierarchically Gaussian formulation of such distribution. An illustrative example is considered to show the performance and support the discussion.