Millimeter Wave Communication SLAM: Framework and RFSoC Prototype

In this paper, we establish a framework and realworld prototype based on Xilinx RFSoC for mmWave communication SLAM (C-SLAM). Users perform mmWave communication with the base stations employing beam management, based on which the users localize themselves and reconstruct the map of the environment. Specifically, we design a line-of-sight (LOS) and first-order non-line-of-sight (NLOS) path extraction algorithm and an algorithm that fuses angle features of LOS paths with inertia measurement unit (IMU) and wheel encoder measurements for precise self-localization. We introduce radar cross section based (RCS-based) grid map to represent the environment, and we derive reflection planes with firstorder NLOS paths to build the map. Unlike most traditional work, we implement the C-SLAM prototype on Xilinx RFSoC. Empirical experiments show that the localization with extended kalman filter (EKF) achieves an average accuracy of 0.0347 m, and the map constructed encompasses both contour information and reflection ability of the environment. During C-SLAM, realtime data transmission is maintained at over 1 Gbps, and thus the proposed C-SLAM demonstrates high practicality without interfering with mmWave communications.