Featured by centralized processing and cloud based infrastructure, Cloud Radio Access Network (C-RAN) is a promising solution to achieve an unprecedented system capacity in future wireless cellular networks. The huge capacity gain mainly comes from the centralized and coordinated signal processing at the cloud server.However, full-scale coordination in a large-scale C-RAN requires the processing of very large channel matrices, leading to high computational complexity and channel estimation overhead. To resolve this challenge, we exploit the near-sparsity of large C-RAN channel matrices, and derive a unified theoretical framework for clustering and parallel processing. Based on the framework, we propose a dynamic nested clustering (DNC) algorithm that not only greatly improves the system scalability in terms of baseband-processing and channel-estimation complexity, but also is amenable to various parallel processing strategies for different data center architectures. With the proposed algorithm, we show that the computation time for the optimal linear detector is greatly reduced from O(N 3 ) to no higher than O(N 42 23 ), where N is the number of RRHs in C-RAN.

In this paper, we endeavour to seek a fundamental understanding of the potentials and limitations of training-based multiuser multiple-input multiple-output (MIMO) systems. In a multiuser MIMO system, users are geographically separated. So, the near-far effect plays an indispensable role in channel fading. The existing optimal training design for convenitional MIMO does not take the near-far effect into account, and thus is not applicable to a multiuser MIMO system. In this work, we use the majorization theory as a basic tool to study the tradeoff between the channel estimation quality and the information throughput. We establish tight upper and lower bounds of the throughput, and prove that the derived lower bound is asymptotically optimal for throughput maximization at high signal-to-noise ratio. Our analysis shows that the optimal training sequences for throughput maximization in a multiuser MIMO system are in general not orthogonal to each other. Futhermore, due to the near-far effect, the optimal training design for throughput maximization is to deactivate a portion of users with the weakest channels in transmission. These observations shed light on the practical design of training-based multiuser MIMO systems. Index Terms-Training-based multiuser MIMO, throughput maximization, massive MIMO

Abstract-Cloud Radio Access Network (C-RAN) is a promising architecture for unprecedented capacity enhancement in nextgeneration wireless networks thanks to the centralization and virtualization of base station processing. However, centralized signal processing in C-RANs involves high computational complexity that quickly becomes unaffordable when the network grows to a huge size. Among the first, this paper endeavours to design a scalable uplink signal detection algorithm, in the sense that both the complexity per unit network area and the total computation time remain constant when the network size grows. To this end, we formulate the signal detection in C-RAN as an inference problem over a bipartite random geometric graph. By passing messages among neighboring nodes, message passing (a.k.a. belief propagation) provides an efficient way to solve the inference problem over a sparse graph. However, the traditional message-passing algorithm is not guaranteed to converge, because the corresponding bipartite random geometric graph is locally dense and contains many short loops. As a major contribution of this paper, we propose a randomized Gaussian message passing (RGMP) algorithm to improve the convergence. Instead of exchanging messages simultaneously or in a fixed order, we propose to exchange messages asynchronously in a random order. The proposed RGMP algorithm demonstrates significantly better convergence performance than conventional message passing. The randomness of the message updating schedule also simplifies the analysis, and allows the derivation of the convergence conditions for the RGMP algorithm. In addition, we generalize the RGMP algorithm to a blockwise RGMP (B-RGMP) algorithm, which allows parallel implementation. The average computation time of B-RGMP remains constant when the network size increases.

Banded linear systems arise in many communication scenarios, e.g., those involving inter-carrier interference and inter-symbol interference. Motivated by recent advances in deep learning, we propose to design a high-accuracy low-complexity signal detector for banded linear systems based on convolutional neural networks (CNNs). We develop a novel CNN-based detector by utilizing the banded structure of the channel matrix. Specifically, the proposed CNN-based detector consists of three modules: the input preprocessing module, the CNN module, and the output postprocessing module. With such an architecture, the proposed CNN-based detector is adaptive to different system sizes, and can overcome the curse of dimensionality, which is a ubiquitous challenge in deep learning. Through extensive numerical experiments, we demonstrate that the proposed CNN-based detector outperforms conventional deep neural networks and existing model-based detectors in both accuracy and computational time. Moreover, we show that CNN is flexible for systems with large sizes or wide bands. We also show that the proposed CNN-based detector can be easily extended to near-banded systems such as doubly selective orthogonal frequency division multiplexing (OFDM) systems and 2-D magnetic recording (TDMR) systems, in which the channel matrices do not have a strictly banded structure.The work in this paper will be partially presented in IEEE Globecom 2018 [1].

No abstract