Radio Point Cloud-Based Positioning and Mapping

Abstract

This paper proposes a point cloud-based radio positioning and mapping algorithm that enables geometry-aware reconstruction from channel impulse responses (CIRs). Traditional radio simultaneous localization and mapping (SLAM) algorithms rely on predefined virtual anchors or scattering models, which restrict their ability to represent arbitrary geometries. The proposed algorithm directly infers reflection points from multiantenna CIR measurements, aligns them via a UE-to-global transform, and aggregates their global positions over time to construct a flexible radio point cloud map. Simulation results in a simplified urban scenario show 0.4 m UE localization RMSE and 0.25 m mapping Chamfer distance, confirming that the point cloud-based approach enables geometry-aware radio sensing.