From Radio Signals to Spatial Maps: End-to-End Learning for Virtual Anchor Mapping

This paper proposes an end-to-end learning algorithm for virtual anchor (VA) mapping with channel impulse responses (CIRs). Existing probabilistic simultaneous localization and mapping (SLAM) algorithms are evaluated in specific environments, raising concerns about their ability to generalize to broader and more complex scenarios. To address this issue, the proposed algorithm is trained in a wide variety of simulation environments and tested in previously unseen scenarios. Additionally, the algorithm begins by determining the number of VAs and subsequently refines their positions through an iterative process. The simulation results show 99.17% accuracy and 3m generalized optimal subpattern assignment (GOSPA) performance.