Orthogonal AMP for Massive Access in Channels With Spatial and Temporal Correlations

We address the joint device activity detection and channel estimation (JACE) problem in a massive MIMO connectivity scenario in which a large number of mobile devices are connected to a base station (BS), while only a small portion are active at any given time. The main objective is to provide an ef...

Full description

Saved in:
Bibliographic Details
Published in:IEEE journal on selected areas in communications 2021-03, Vol.39 (3), p.726-740
Main Authors: Cheng, Yiyao, Liu, Lei, Ping, Li
Format: Article
Language:English
Subjects:
Algorithms
Antenna measurements
Antennas
Channel estimation
Channels
Compressed sensing
Correlation
Correlation analysis
device activity detection
Electronic devices
Massive connectivity
Message passing
orthogonal approximate message passing (OAMP)
Partial transmit sequences
Radio equipment
Sensors
Signal processing
spatial and temporal correlations
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We address the joint device activity detection and channel estimation (JACE) problem in a massive MIMO connectivity scenario in which a large number of mobile devices are connected to a base station (BS), while only a small portion are active at any given time. The main objective is to provide an efficient transmission and detection scheme with both spatial and temporal correlations. We formulate JACE as a multiple measurement vector (MMV) problem with correlated entries in the vectors to be estimated. We propose an MMV form of the orthogonal approximate message passing algorithm (OAMP-MMV). We derive a group Gram-Schmidt orthogonalization (GGSO) procedure for the realization of OAMP-MMV. We outline a state evolution (SE) procedure for OAMP-MMV and examine its accuracy using numerical results. We also compare OAMP-MMV with existing alternatives, including AMP-MMV and GTurbo-MMV. We show that OAMP-MMV outperforms AMP-MMV when pilot sequences are generated using Hadamard pilot matrices. Such a pilot design is attractive due to the low-cost signal processing technique using the fast Hadamard transform (FHT). We also show that OAMP-MMV outperforms GTurbo-MMV in correlated channels.
ISSN:0733-8716
1558-0008
DOI:10.1109/JSAC.2020.3018799