Reconfigurable Holographic Surface Aided Collaborative Wireless SLAM Using Federated Learning for Autonomous Driving

Abstract

Simultaneous Localization and Mapping (SLAM) utilizing millimeter-wave (mmWave) radars is widely recognized as an essential component for autonomous driving applications. In this paper, we present a Reconfigurable Holographic Surface (RHS)-aided SLAM system, incorporating federated learning. The hardware cost of autonomous driving systems can be significantly reduced by replacing the expensive phased array antennas, traditionally used in mmWave radars, with the low-cost RHS metasurface antenna. Furthermore, multiple vehicles can collaborate through the federated learning framework, obtaining additional sensed data to enhance SLAM performance. However, the distinctive radiation structure of the RHS and the information exchange within the federated learning framework introduce complexities to the overall SLAM system design. To address these challenges, we propose a multi-vehicle SLAM protocol that regulates RHS-based radar sensing and data processing across multiple vehicles. Additionally, we design algorithms for RHS radiation optimization and federated learning-based localization and mapping. Simulation results demonstrate the efficacy of the proposed approach when compared to existing phased array-based and non-cooperative schemes.