Deep Learning-Driven Landmark Mapping with Channel Impulse Responses

Abstract

In this paper, we propose a deep learning-driven landmark mapping algorithm using channel impulse responses (CIRs). Existing radio simultaneous localization and mapping (SLAM) utilize less accurate signal channel information and has a high computational complexity in processing. To address these challenges, we leverage raw data, CIRs, instead of angle and distance information. Furthermore, we replace mapping filters in existing radio SLAM algorithms with deep learning to enhance mapping performance. Through the simulation results, the effectiveness of the proposed algorithm is verified by comparing the benchmarks.