发表机构
Communications and Information Theory Chair (CommIT), Faculty of Electrical Engineering and Computer Science, Technische Universität Berlin（柏林工业大学电气与计算机科学学院通信与信息论讲席）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文针对无小区用户中心网络，提出联合随机接入与定位的统一框架，采用位置相关分区码本和近似消息传递方法，在频率选择性信道下实现优于传统方案的前导码检测与高精度定位。
AI 中文摘要
我们研究了面向无小区（CF）用户中心网络的随机接入（RACH）方案，以处理大量地理分布的用户，这些用户具有零星流量和间歇性活动。RACH必须允许系统：1）检测由（尚未知的）随机接入用户在RACH时隙中发送的前导码；2）对其进行定位，以便快速分配用户中心无线单元（RU）簇。大多数先前的工作使用简化模型，忽略了帧同步但码片异步的传输、某些用户-RU对可能存在的视距（LoS）传播，以及产生频率选择性信道的多径非视距（NLoS）传播。基于我们之前的工作，我们考虑了位置相关的分区随机接入码本，其中地理区域（位置）中的用户使用相应的随机接入前导码子集。我们提出了一个统一框架，用于在空间一致的全网络信道模型上进行联合检测和定位，该模型包含了这些被忽略的方面。我们评估了两种方案：1）一种使用Zadoff-Chu（ZC）序列的“传统”方案，将3GPP两步RACH扩展到CF情况；2）我们扩展至多径频率选择性衰落的多源近似消息传递（AMP）方法。对于这两种方案，我们开发了新颖的近似GLRT前导码检测和具有超分辨率细化的最大似然位置估计器，隐式利用嵌入在LoS分量中的接收信号强度、到达角和到达时间差信息。数值结果表明，基于AMP的方案实现了更优的前导码检测，而两种方案都具有相似且出色的定位能力。
英文摘要
We study random access (RACH) schemes for cell-free (CF) user-centric networks to handle many geographically distributed users with sporadic traffic and intermittent activity. The RACH must allow the system to: 1) detect preambles sent by the (yet unknown) random access users in the RACH slot; 2) localize them for fast allocation of user-centric radio-unit (RU) clusters. Most prior work uses simplified models, neglecting frame-synchronous but chip-asynchronous transmission, possible line-of-sight (LoS) propagation for certain user-RU pairs, and multipath non-line-of-sight (NLoS) propagation yielding frequency-selective channels. Building on our previous work, we consider location-dependent partitioned random access codebooks where users in a geographic area (location) use the corresponding subset of random access preambles. We present a unified framework for joint detection and localization over a spatially consistent network-wide channel model, incorporating these neglected aspects. We evaluate two schemes: 1) a ``legacy'' scheme using Zadoff-Chu (ZC) sequences, extending the 3GPP 2-step RACH to the CF case; 2) our multisource approximate message passing (AMP) approach extended to multipath frequency-selective fading. For both schemes, we develop novel approximated GLRT preamble detection and Maximum-Likelihood position estimators with super-resolution refinement, implicitly exploiting received signal strength, angle of arrival, and time-difference of arrival information embedded into LoS components. Numerical results show that the AMP-based scheme achieves superior preamble detection, while both schemes have similar and excellent localization capability.
