Channel estimation, user activity identification and signal detection in grant-free multiple access systems

Abstract

The 5th and beyond wireless systems are expected to be device-centric. It has been widely recognized that in the device-centric systems, the traditional grant-based multiple access (GMA) methods are low efficiency due to the hand-shaking procedure resulted resource consumption and latency. Hence, the grant-free multiple access (GFMA) techniques have been proposed and studied, in order to solve the problems experienced by the GMA. This thesis focuses on the GFMA for massive machine-type communications (mMTC), with emphasis on the physical layer techniques, including channel estimation, user equipment (UE) activity identification (UAI) and information detection. To these objectives, we first provide a literature review in terms of the research background and various methods for achieving GFMA. Then, the performance of the dynamic direct sequence code-division multiple-access (DyDS-CDMA) and multicarrier CDMA (DyMC-CDMA) with minimum mean-square error (MMSE) and MMSE-assisted successive interference cancellation (MMSE-SIC) detection schemes are studied. In our studies, we assume that each base station (BS) or access point (AP) with limited degrees-of-freedom is capable to support a massive number of potential UEs, while each UE becomes active with a small probability. Hence, the active UEs and the number of them are highly dynamic, making the number of active UEs possibly higher than the degrees-of-freedom of the system. Assuming ideal UAI and ideal channel estimation of active UEs, we study the potential performance achievable by the DyDS-CDMA and DyMC-CDMA systems employing the MMSE-assisted successive interference cancellation detection (MMSE-SICD), to demonstrate the feasibility of MMSE-SICD for operation in massive GFMA (mGFMA) systems, where the number of active UEs may be much higher than the systems’ degree-of-freedom. Our studies reveal that the DyDS-CDMA and DyMC-CDMA systems aided by the MMSE-SICD are highly efficient for operation in mGFMA environments. Near single-user performance is achievable, even when the average number of active UEs per time-slot reaches two times of the system’s degrees-of-freedom, and hence, the number of active UEs of a time-slot may be much higher than two times of the system’s degrees-of-freedom. Following the above preparation work, then, we investigate the channel estimation and propose UAI algorithms for mGFMA systems. Specifically, channel estimation is studied from several aspects by assuming different levels of knowledge to the AP, and based on which five UAI approaches are proposed. We study the performance of channel estimation, the statistics of estimated channels, and the performance of UAI algorithms. Our studies show that the proposed approaches are capable of circumventing some of the shortcomings of the existing techniques designed based on compressive sensing (CS) and message passing algorithms (MPAs). They are robust for operation in the mGFMA systems where the active UEs and the number of them are highly dynamic. Then, we investigate a multicarrier mGFMA system with an assumption that a big number of highly dynamic UEs are monitored by an AP with multiple receive antennas (MRA), which is referred to as the MRA/MC-mGFMA system. The channel estimation, UAI and information detection are separately or jointly addressed. To be more specific, firstly, the channels of both active and inactive UEs are estimated in the principle of MMSE. Then, based on the estimated channels, a low-complexity threshold-based UAI (TB-UAI) is proposed to detect the activities of UEs. Finally, information of active UEs is detected in the principle of MMSE-SIC. Furthermore, a joint algorithm, referred to as SIC-MMSE-JCUD, is proposed for joint channel estimation, UAI and data detection in the principle of MMSESIC. Additionally, considering that for a practically limited bandwidth, no set of the welldesigned signature sequences can support a big number of UEs in a mGFMA system, we propose a class of sequences by combining the Gold-sequences with the Zadoff-Chu (ZC) sequences. Our studies show that deploying multiple receive antennas at AP is beneficial to channel estimation, UAI and information detection. Aided by the multiple receive antennas of AP, a low-complexity TB-UAI algorithm is highly efficient for UAI. Furthermore, our proposed class of signature sequences allows to attain much better performance than the random sequences. Then, we extend our studies to the massive distributed grant-free multiple-access (MDRGFMA) systems, to investigate the mGFMA with the cell-free scenario. In our studies, we assume a MDR-GFMA system where remote radio heads (RRHs) or APs, or simply distributed antennas (DAs) are randomly distributed in a given area based on the point Poisson (PP) distribution, while UEs are uniformly distributed. We assume that signals transmitted by UEs experience both the large-scale fading of propagation path-loss and shadowing as well as the small-scale Rayleigh fading. Signals received by different RRHs/APs are forwarded to a so-called signal processing central unit (SPCU), where channel estimation, UAI and data detection are carried out. In terms of signal processing at SPCU, channel estimation is achieved in the principle of MMSE. Following channel estimation, an orthogonal matching pursuit (OMP) relied algorithm is implemented to attain initial UAI, which is enhanced with the aid of the pilot detection of each initially identified active UE. Finally, the data sent by active UEs is detected using either MMSE detection or the MMSE-SIC detection. Our studies show that the proposed algorithms are effective, and achieve expected performance in the MDR-GFMA systems with various dynamics, including active UEs and the number of them, locations of DAs and the number of them serving different UEs, geographically resulted large-scale fading of propagation path-loss and shadowing

More information

Identifiers

ORCID for Lie-Liang Yang: orcid.org/0000-0002-2032-9327

Catalogue record

Export record