Dirichlet Process Clustering-based Radio SLAM with Arbitrarily-Shaped Reflectors

Paper in proceeding, 2023

This paper proposes a Dirichlet process (DP)-based radio simultaneous localization and mapping (SLAM) algorithm enabling mapping arbitrary structures (ASs) as well as the standard point landmarks. The ASs cannot be characterized by a low-dimensional state, in contrast to the standard point landmarks, leading to incorrect mapping results in the existing radio SLAM methods. To tackle the incorrect mapping issue, we develop a DP-based data association method, where the landmarks are maintained by the clusters, and each birth point by the measurement is assigned to the existing or a new cluster. Compared to the well-known state-of-the-art method, we evaluate the performance of the proposed algorithm under the scenario with multiple landmarks deployed. This validation represents that radio SLAM is possible in an environment where objects of ASs exist through the proposed method.