ABSTRACT The robustness against timing offsets of filter bank multi‐carrier (FBMC) is appealing for grant‐free massive access scenarios that mainly adopt asynchronous transmissions. In this work, we propose a compressed sensing based algorithm for joint active user detection as well as timing offset and channel estimation in uplink communication under the combination of FBMC and grant‐free massive access systems, which is critical for subsequent decoding or other processes at receiver. The channel estimation part is based on generalized approximate massage passing (GAMP). The active user detection and timing offset estimation are based on loopy belief propagation (LBP) rules, where the expressions of message passing and belief distributions are derived. Besides, since the receiver may have no prior knowledge about some parameters such as noise variance and activity probability, we introduce the expectation maximization (EM) approach into the proposed algorithm. Moreover, we develop a preamble design method to improve the detection and estimation performance. Simulation results show that the proposed EM‐LBP‐GAMP algorithm can achieve satisfying performance in terms of missed activity detection probability, timing offset estimation error and normalized mean square error of channel estimation.

In this paper, we study filter bank multi-carrier (FBMC)-based uplink massive connectivity systems with asynchronous transmission and frequency-selective channels. We focus on the joint active device detection, delay and channel estimation (JADDCE) problem as they are critical for subsequent decoding or other processes. The problem can be put into the compressed sensing (CS) framework, where the structural sparsity is explored, and an efficient algorithm is proposed for JADDCE. Besides, since parameters such as noise variance and activity probability may not be perfectly known by the receiver, we introduce the expectation maximization (EM) method into the algorithm and derive the updating rules of the unknown parameters. Simulation results show satisfying performance of the proposed algorithm under frequency-selective channels.

This article studies the topic of probabilistic on-off division multiple access (ODMA) system design under multiple input-multiple output (MIMO) and massive MIMO with hybrid analog/digital architectures for unsourced random access (URA). Though probabilistic ODMA in Gaussian multiple access channel (GMAC) manifests desirable capacity, probabilistic ODMA extension towards MIMO and hybrid massive MIMO encounters problems have not been discussed. Specifically, a portion of data bits are utilized to select on-off patterns and not transmitted. These pattern-correlated bits and others are restored iteratively with pilot-free transmission in the so-called process of joint pattern and data detection. Discrepant to the state-of-arts, such as interleaving division multiple access (IDMA), where prior of patterns is an essential prerequest for the later detection, probabilistic ODMA detects pattern and data simultaneously with only linearly modulated signals. In this work, two probabilistic ODMA transceiver designs are proposed by different ranks of signal component matrix in MIMO and hybrid massive MIMO. Overall, our proposed designs retain the simplicity of the original ODMA and further exploits the sparsity of on-off patterns. For MIMO, on-off patterns are correlated with common codebook and a receiver is designed in the spirit of uncoupled URA. For hybrid massive MIMO, a novel probabilistic ODMA transceiver achieves probabilistic channel estimation via low-rank matrix completion aided by the sparsity of on-off patterns. Meanwhile, joint pattern and data detection is accomplished by iterative treating interference as noise (TIN) strategy. Extensive simulations are conducted under fair comparisons with state-of-the-arts to verify the validity of the proposed schemes.

The robustness of filter bank multi-carrier (FBMC) against delays is appealing for asynchronous grant-free massive connectivity systems. In this paper, we study the joint activity detection, delay and channel estimation (JADDCE) problem for filter bank multi-carrier (FBMC)-based uplink massive connectivity with asynchronous transmission and frequency-selective fading (FSF) channels. We formulate JADDCE as a compressed sensing (CS) problem to fully exploit the sparsity structure and propose an efficient algorithm based on generalized approximate message passing (GAMP) to solve it. Besides, since parameters such as noise variance and activity probability may not be perfectly known by the receiver, we introduce the expectation maximization (EM) method into the algorithm and derive the updating rules of the unknown parameters. We also utilize the analysis framework based on average mutual information (AMI) to find theoretical upper-bound of the channel estimation performance. Simulation results show satisfying detection and estimation performance of the proposed algorithm under FSF channels. Besides, the channel estimation performance can approach the theoretical upper-bound.

Filter bank multiple access (FBMA) without subbands orthogonality has been proposed as a new candidate waveform to better meet the requirements of future wireless communication systems and scenarios. It has the ability to process directly the complex symbols without any fancy preprocessing. Along with the usage of cyclic prefix (CP) and wide-banded subband design, CP-FBMA can further improve the peak-to-average power ratio and bit error rate performance while reducing the length of filters. However, the potential gain of removing the orthogonality constraint on the subband filters in the system has not been fully exploited from the perspective of waveform design, which inspires us to optimize the subband filters for CP-FBMA system to maximizing the achievable rate. Besides, we propose a joint optimization algorithm to optimize both the waveform and the covariance matrices iteratively. Furthermore, the joint optimization algorithm can meet the requirements of filter design in practical applications in which the available spectrum consists of several isolated bandwidth parts. Both general framework and detailed derivation of the algorithms are presented. Simulation results show that the algorithms converge after only a few iterations and can improve the sum rate dramatically while reducing the transmission delay of information symbols.