Bistatic Radio SLAM with Offline Shape Estimation

Abstract

The geometric connection between the propagation environment and millimeter wave (mmWave) signals can be leveraged for simultaneous localization and mapping (SLAM) in 5G and beyond networks. Conventional solutions either solve the SLAM problem for a single user equipment (UE) location or rely on Bayesian filtering techniques in which the environmental landmarks are modeled using a point object model. In this paper, we devise a labeled multi-model probability hypothesis density (PHD) filter which is able to track two different types of objects. In addition, we propose an offline shape estimation algorithm which utilizes labels of the PHD filter and the measurements for estimating the shape of reflecting surfaces and small scattering objects. The proposed method is validated using ray-tracing data as well as real-world 60 GHz experimental data. The results indicate that the proposed method outperforms a state-of-the-art benchmark algorithm and the offline shape estimation algorithm improves the extracted map of the environment.