基于信道状态信息的无监督定位与环境建图的神经射频 SLAM
信息论
2022-03-17 v1 机器学习
信号处理
math.IT
摘要
我们提出了一种神经网络架构,用于从信道状态信息(CSI)值中以无监督方式联合学习用户位置与环境建图(可达等距变换),且无需位置信息。该模型基于编码器-解码器架构。编码器网络将 CSI 值映射至用户位置。解码器网络通过使用虚拟锚点对环境进行参数化来建模传播物理过程,旨在根据编码器输出与虚拟锚点位置,重构由超分辨率方法从 CSI 中提取的飞行时间(ToF)集合。该神经网络的任务为集合预测,并据此进行端到端训练。所提模型仅通过施加基于物理的解码器,即可学习到可解释的潜变量(即用户位置)。结果表明,在基于合成光线追踪的数据集上,所提模型在单锚点 SISO 设置下达到亚米级精度,同时在二维环境中以 4cm 中位误差、三维环境中以 15cm 中位误差恢复环境地图。
引用
@article{arxiv.2203.08264,
title = {Neural RF SLAM for unsupervised positioning and mapping with channel state information},
author = {Shreya Kadambi and Arash Behboodi and Joseph B. Soriaga and Max Welling and Roohollah Amiri and Srinivas Yerramalli and Taesang Yoo},
journal= {arXiv preprint arXiv:2203.08264},
year = {2022}
}
备注
Accepted at IEEE International Conference on Communications 2022. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other work