Estimating Initial Map Features for High-Efficiency mmWave Cellular SLAM

Abstract

<p>The intrinsic geometric connections between the environment and the millimeter-wave (mmWave) signals can be exploited for cellular simultaneous localization and mapping (SLAM) in 5G and future 6G networks. One central task in any mmWave SLAM filter is how new map features are initialized when they are observed for the first time. In this paper, we study this initialization problem and present low-complexity methods for characterizing density of the features. The presented methods cover the typical scenarios encountered in mmWave cellular SLAM, and two variants of the probability hypothesis density SLAM filter are implemented to demonstrate the proposed methods. We evaluate the new initialization methods using simulations and the results imply that the proposed methods have a low computational overhead and provide accurate initial estimates, enhancing the efficiency and/or accuracy of SLAM in future cellular systems.</p>