The decoupling of data and control planes in the forthcoming 5G wireless networks enables the efficient implementation of multitier architectures, where coverage and connectivity are guaranteed by top-tier macro-cells, and at the same time, high throughput and low latency are achieved locally by lower tiers in the hierarchy. This paper considers a new architecture for such lower tiers, dubbed fog massive multi-in multi-out (MIMO), where the user equipment (UE) nodes can connect to a “fog” of the densely deployed remote radio heads (RRHs) in a seamless, user-centric, and opportunistic manner. In the case of dense small cells, traditional cellular architectures inherently give rise to frequent handovers and pilot sequence re-assignments, which typically incur excessive protocol overhead and latency. In the proposed fog massive MIMO architecture, the UEs implicitly associate themselves with the most convenient RRHs in a completely autonomous manner. Each UE makes use of the uplink pilot sequences that contain equal-weight codewords and enable a novel “on-the-fly” pilot contamination control mechanism. We analyze the spectral efficiency and the outage probability of the proposed architecture via stochastic geometry, using some recent results on unique coverage in Boolean models, and provide a detailed comparison with the benchmark represented by massive MIMO cellular system with a genie-aided minimum-distance user-cell association. Our analysis, corroborated by extensive system simulation, reveals that there exists a “sweet spot” of the per pilot user load (number of users per pilot), such that the proposed system achieves a spectral efficiency close to that of the genie-aided cellular system. In these conditions, it is possible to achieve the low protocol overhead and low user RRH association latency promised by the fog massive MIMO at virtually no significant performance cost in terms of system spectral efficiency with respect to the cellular benchmark.

We investigate the viability of using machine-learning techniques for estimating user-channel features at a large-array base station (BS). In the scenario we consider, user-pilot broadcasts are observed and processed by the BS to extract angle-of-arrival (AoA) specific information about propagation-channel features, such as received signal strength and relative path delay. The problem of interest involves using this information to predict the angle-of-departure (AoD) of the dominant propagation paths in the user channels, i.e., channel features not directly observable at the BS. To accomplish this task, the data collected in the same propagation environment are used to train neural networks. Our studies rely on ray-tracing channel data that have been calibrated against measurements from Shinjuku Square, a famous hotspot in Tokyo, Japan. We demonstrate that the observed features at the BS side are correlated with the angular features at the user side. We train neural networks that exploit different combinations of measured features at the BS to infer the unknown parameters at the users. The evaluation based on standard statistical performance metrics suggests that such data-driven methods have the potential to predict unobserved channel features from observed ones.

We propose a novel scheme for estimating the large- scale gains of the channels between user terminals (UTs) and base stations (BSs) in a cellular system. The scheme leverages TDD operation, uplink (UL) training by means of properly designed non-orthogonal pilot codes, and massive antenna arrays at the BSs. Subject to Q resource elements allocated for UL training and using the new scheme, a BS is able to estimate the large- scale channel gains of K users transmitting UL pilots in its cell and in nearby cells, provided K ≤Q2. Such knowledge of the large-scale channel gains of nearby out-of-cells users can be exploited at the BS to mitigate interference to the out-of-cell users that experience the highest levels of interference from the BS. We investigate the large-scale gain estimation performance provided by a variety of non-orthogonal pilot codebook designs. Our simulations suggest that among all the code designs considered, Grassmannian line-packing type codes yield the best large-scale channel gain estimation performance.

This paper considers a new architecture underpinned by on-the-fly pilot contamination control, termed Fog massive MIMO, where the users are able to establish high-throughput and low-latency data links in a seamless and opportunistic manner, as they travel through a dense fog of high capacity Remote Radio Heads (RRHs). Utilizing some recent results on unique coverage in Boolean models, we analyze the spectral efficiency and outage probability of the proposed architecture via stochastic geometry. Our analysis, supported by extensive system simulation, reveals that there exists a “sweet spot” of the per-pilot user load (number of users per pilot), such that the proposed system achieves spectral efficiency close to that of the ideal cellular massive MIMO baseline, while exhibiting the simplicity and low-latency of completely user-centric operations.

We consider a single-cell scenario involving a single base station (BS) with a massive array serving multi-antenna terminals in the downlink of a mmWave channel. We present a class of multiuser MIMO schemes, which rely on uplink training from the user terminals, and on uplink/downlink channel reciprocity. The BS employs virtual sector-based processing according to which, user-channel estimation and data transmission are performed in parallel over non-overlapping angular sectors. The uplink training schemes we consider are non-orthogonal, that is, we allow multiple users to transmit pilots on the same pilot dimension (thereby potentially interfering with one another). Elementary processing allows each sector to determine the subset of user channels that can be resolved on the sector (effectively pilot contamination free) and, thus, the subset of users that can be served by the sector. This allows resolving multiple users on the same pilot dimension at different sectors, thereby increasing the overall multiplexing gains of the system. Our analysis and simulations reveal that, by using appropriately designed directional training beams at the user terminals, the sector-based transmission schemes we present can yield substantial spatial multiplexing and ergodic user-rates improvements with respect to their orthogonal-training counterparts.

Hybrid preprocessing, where an analogue preprocessing matrix is used to feed a large antenna array to fewer up/down-conversion chains, helps reduce hardware cost of massive Multiple-Input-Multiple-Output systems. Here, we analyze a variant of hybrid preprocessing, namely hybrid preprocessing with selection (HPwS), as an attractive solution to reduce this hardware cost while retaining good performance. In HPwS, the preprocessing matrix, built from radio frequency (RF) hardware, has a larger number of input ports L than the number of up/down-conversion chains K. A bank of RF switches connects the instantaneously best K input ports to the up/down-conversion chains. The preprocessor is designed based on average channel statistics and therefore needs to be updated only infrequently. This allows for a higher diversity-order and/or simpler RF hardware than some conventional hybrid preprocessing systems. In this paper, we propose a generic architecture of HPwS with a possibly non-unitary rectangular preprocessing matrix. A novel preprocessor is designed that maximizes a capacity lower bound for channels with isotropic scattering. In addition, we study how L, the number of users, and the use of low complexity RF hardware, such as discrete phase-shifters, impact system performance. We also present a method to extend the preprocessor design to anisotropic channels.

Technologies for millimeter wave (mmWave) communication are at the forefront of investigations in both industry and academia, as the mmWave band offers the promise of orders of magnitude additional available bandwidths to what has already been allocated to cellular networks. The much larger number of antennas that can be supported in a small footprint at mmWave bands can be leveraged to harvest massive-MIMO type beamforming and spatial multiplexing gains. Similar to Long-Term Evolution (LTE) systems, two prerequisites for harvesting these benefits are detecting users and acquiring user channel state information (CSI) in the training phase. However, due to the fact that mmWave channels encounter much harsher propagation and decorrelate much faster, the tasks of user detection and CSI acquisition are both imperative and much more challenging than in LTE bands. In this paper, we investigate the problem of fast user detection and CSI acquisition in the downlink of small cell mmWave networks. We assume Time Division Duplex (TDD) operation and channel-reciprocity based CSI acquisition. To achieve densification benefits we propose pilot designs and channel estimators that leverage a combination of aggressive pilot reuse with fast user detection at the base station and compressed sensing channel estimation. As our simulations show, the number of users that can be simultaneously served by the entire mmWave-band network with the proposed schemes increases substantially with respect to traditional compressed sensing based approaches with conventional pilot reuse.

Dense large-scale antenna deployments are one of the most promising technologies for delivering very large throughputs per unit area in the downlink (DL) of cellular networks. We consider such a dense deployment involving a distributed system formed by multi-antenna remote radio head (RRH) units connected to the same fronthaul serving a geographical area. Knowledge of the DL channel between each active user and its nearby RRH antennas is most efficiently obtained at the RRHs via reciprocity based training, that is, by estimating a user's channel using uplink (UL) pilots transmitted by the user, and exploiting the UL/DL channel reciprocity. We consider aggressive pilot reuse across an RRH system, whereby a single pilot dimension is simultaneously assigned to multiple active users. We introduce a novel coded pilot approach, which allows each RRH unit to detect pilot collisions, i.e., when more than a single user in its proximity uses the same pilot dimensions. Thanks to the proposed coded pilot approach, pilot contamination can be substantially avoided. As shown, such a strategy can yield densification benefits in the form of increased multiplexing gains per UL pilot dimension with respect to conventional reuse schemes and some recent approaches assigning pseudorandom pilot vectors to the active users.

Two key traits of 5G cellular networks are much higher base station (BS) densities-especially in the case of low-power BSs-and the use of massive MIMO at these BSs. This paper explores how massive MIMO can be used to jointly maximize the offloading gains and minimize the interference challenges arising from adding small cells. We consider two interference management approaches: joint transmission (JT) with local precoding, where users are served simultaneously by multiple BSs without requiring channel state information exchanges among cooperating BSs, and resource blanking, where some macro BS resources are left blank to reduce the interference in the small cell downlink. A key advantage offered by massive MIMO is channel hardening, which enables to predict instantaneous rates a priori. This allows us to develop a unified framework, where resource allocation is cast as a network utility maximization (NUM) problem, and to demonstrate large gains in cell-edge rates based on the NUM solution. We propose an efficient dual subgradient based algorithm, which converges towards the NUM solution. A scheduling scheme is also proposed to approach the NUM solution. Simulations illustrate more than 2x rate gain for 10th percentile users vs. an optimal association without interference management.

Massive MIMO is one of the most promising approaches for coping with the predicted wireless data traffic explosion. Future deployment scenarios will involve dense heterogeneous networks, comprised of massive MIMO base stations with different powers, numbers of antennas and multiplexing gain capabilities, and possibly highly nonhomogeneous user density (hot-spots). In such dense irregularly deployed networks, it will be important to have mechanisms for associating users to base stations so that the available wireless infrastructure is efficiently used. In this paper, we consider the optimal user-cell association problem for massive MIMO heterogeneous networks and illustrate how massive MIMO can also provide nontrivial advantages at the system level. Unlike previous treatments that rely on integer program problem formulations and their convex relaxations, the user-cell association problem is formulated directly as a convex network utility maximization and solved efficiently by a centralized subgradient algorithm. As we show, the globally optimal solution is physically realizable, in that there exists a sequence of integer-valued associations approaching arbitrarily closely the optimal fractional association. We also consider simple decentralized user-centric association schemes, where each user individually and selfishly connects to the base station with the highest promised throughput. Such user-centric schemes where users make local association decisions in a probabilistic manner can be viewed as games and are known to converge to Nash equilibria. Surprisingly, as we show, under certain conditions, the globally optimal solution is close to these Nash equilibria. Such decentralized approaches are, therefore, attractive not only for their simplicity, but also because they operate near the system social optimum. Our theoretical results are confirmed by extensive simulations with realistic LTE-like network parameters.

Multi-tier networks with large-array base stations (BSs) that are able to operate in the "massive MIMO" regime are envisioned to play a key role in meeting the exploding wireless traffic demands. Operated over small cells with reciprocity-based training, massive MIMO promises large spectral efficiencies per unit area with low overheads. Also, near-optimal user-BS association and resource allocation are possible in cellular massive MIMO HetNets using simple admission control mechanisms and rudimentary BS schedulers, since scheduled user rates can be predicted a priori with massive MIMO. Reciprocity-based training naturally enables coordinated multi-point transmission (CoMP), as each uplink pilot inherently trains antenna arrays at all nearby BSs. In this paper we consider a distributed-MIMO form of CoMP, which improves cell-edge performance without requiring channel state information exchanges among cooperating BSs. We present methods for harmonized operation of distributed and cellular massive MIMO in the downlink that optimize resource allocation at a coarser time scale across the network. We also present scheduling policies at the resource block level which target approaching the optimal allocations. Simulations reveal that the proposed methods can significantly outperform the network-optimized cellular-only massive MIMO operation (i.e., operation without CoMP), especially at the cell edge.

Large-scale distributed Multiuser MIMO (MU-MIMO) is a promising wireless network architecture that combines the advantages of "massive MIMO" and "small cells." It consists of several Access Points (APs) connected to a central server via a wired backhaul network and acting as a large distributed antenna system. We focus on the downlink, which is both more demanding in terms of traffic and more challenging in terms of implementation than the uplink. In order to enable multiuser joint precoding of the downlink signals, channel state information at the transmitter side is required. We consider Time Division Duplex (TDD), where the downlink channels can be learned from the user uplink pilot signals, thanks to channel reciprocity. Furthermore, coherent multiuser joint precoding is possible only if the APs maintain a sufficiently accurate relative timing and phase synchronization. AP synchronization and TDD reciprocity calibration are two key problems to be solved in order to enable distributed MU-MIMO downlink. In this paper, we propose novel over-the-air synchronization and calibration protocols that scale well with the network size. The proposed schemes can be applied to networks formed by a large number of APs, each of which is driven by an inexpensive 802.11-grade clock and has a standard RF front-end, not explicitly designed to be reciprocal. Our protocols can incorporate, as a building block, any suitable timing and frequency estimator. Here we revisit the problem of joint ML timing and frequency estimation and use the corresponding Cramer-Rao bound to evaluate the performance of the synchronization protocol. Overall, the proposed synchronization and calibration schemes are shown to achieve sufficient accuracy for satisfactory distributed MU-MIMO performance.

Recently, a novel class of high-performance multi-round MU- MIMO schemes were developed by Yang et al. [1] for the practically attractive scenario in which the transmitter has available partial channel state information (CSI) regarding the current state of the channel, but also full delayed CSI. We study methods for combining user scheduling with these multi-round transmission schemes, in order to realize some of their multiplexing gains in the form of high-rates in the SNR regime of practical interest. In these multi-round schemes several differently encoded signals are transmitted to each user over many time-dispersed slots. The challenge in designing schedulers is that certain signals intended for a given user are transmitted in different forms and combinations with other-user signals over multiple slots, occurring at distinctly different times with distinctly different channels. We develop a simple and novel scheduling framework based on appropriate resource division between transmission rounds and round-specific scheduling criteria. Simulation-based performance analysis of the proposed scheduling algorithms, operated on the 2-user transmission scheme of Yang et al. [1] and a 3-user extension suggest that these schemes can harvest significant scheduling benefits in some scenarios.

A new coding scheme for image transmission over noisy channel is proposed. Similar to standard image compression, the scheme includes a linear transform followed by successive refinement scalar quantization. Unlike conventional schemes, though, in the proposed system the quantized transform coefficients are linearly mapped into channel symbols using systematic linear encoders. This fixed-to-fixed length "linear index coding" approach avoids the use of an explicit entropy coding stage (e.g., arithmetic or Huffman coding), which is typically fragile to channel post-decoding residual errors. We use linear codes over GF(4), which are particularly suited for this application, since they are matched to the dead-zone quantizer symbol alphabet and to the QPSK modulation used on the deep-space communication channel. We optimize the proposed system where the linear codes are systematic Raptor codes over GF(4). The rateless property of Raptor encoders allows to implement a "continuum" of coding rates, in order to accurately match the channel coding rate to the transmission channel capacity and to the quantized source entropy rate for each transform subband and refinement level. Comparisons are provided with respect to the concatenation of state-of-the-art image coding and channel coding schemes used by Jet Propulsion Laboratories (JPL) for the Mars Exploration Rover (MER) Mission.

When transmitting source-encoded data, not all information bits are equally important, due to the different sensitivity of the source decoder to errors. Unequal Error Protection (UEP) consists of allocating coding redundancy depending on the importance of the information bits. We consider progressive transmission of source-encoded data under three different packet formats where either number of source bits per packet is fixed (fixed-k approach), or packet block length is fixed (fixed-n approach) or both parameters allowed to vary for each block (variable-(n, k) approach). Most existing results are based on some chosen family of channel codes and consider a single-user setting. Thanks to the recent finite length error probability results by Polyanskiy et al., in this work we investigate the UEP concept using the new finite-length random coding bounds. In the single-user case, we show that when codes meeting Polyanskiy achievability bounds are used, UEP does not obtain significant advantages over Equal-Error Protection (EEP) (advantages disappear for the variable-(n, k) case). Based on these results, a low complexity optimization algorithm is proposed for the multiuser (multicast) scenario.

The multiple description coding of an independent and identically distributed Gaussian source over bit error channels is considered. A novel practical multiple description coding scheme is proposed based on scalar embedded quantizers and rateless codes. A peer-to-peer network model is studied where users can receive different number of descriptions and an arbitrary large number of descriptions can be created and transmitted between peers. For this network model convex system optimization problems are defined and solved. The proposed scheme can be regarded as an efficient implementation of "channel optimized quantization", where the index assignment is obtained via a low-complexity linear function. The scheme is shown to be robust to bit errors and to achieve higher reconstruction signal-to-noise ratios when compared to previously reported designs.

The multicasting of an independent and identically distributed Gaussian source over a binary erasure broadcast channel is considered. This model applies to a one-to-many transmission scenario in which some mechanism at the physical layer delivers information packets with losses represented by erasures, and users are subject to different erasure probabilities. The reconstruction signal-to-noise ratio (SNR) region achieved by concatenating a multiresolution source code with a broadcast channel code is characterized and four convex optimization problems corresponding to different performance criteria are solved. Each problem defines a particular operating point on the dominant face of the SNR region. Layered joint source-channel codes are constructed based on the concatenation of embedded scalar quantizers with binary raptor encoders. The proposed schemes are shown to operate very close to the theoretical optimum.

In this paper the multicasting of independent parallel Gaussian sources over a binary erasure broadcast channel is considered. Multiresolution embedded quantization and layered joint source-channel coding schemes are used in order to serve simultaneously several users at different channel capacities. The convex nature of the rate-distortion function, computed by means of reverse water-filling, allows us to solve relevant convex optimization problems corresponding to different performance criteria. Then, layered joint source-channel codes are constructed based on the concatenation of embedded scalar quantizers with binary rateless encoders.

In this paper we present reduced-complexity high- performance receivers for single-user MIMO systems that employ coded OFDM transmission with bit-interleaved coded modulation (BICM). High data-rate versions of these systems have been proposed for 4G. The near-optimum inner-outer decoder structures for these MIMO/BICM/OFDM/QAM systems have complexity that grows exponentially fast with the number of multiplexed streams (number of transmit antennas) and the number of bits represented by the QAM constellations. We focus on reduced-complexity inner-outer decoder structures that exploit adaptive soft-output M algorithms for soft inner decoding. Our simulations show that these receivers can be systematically optimized to yield substantial complexity reductions without significant loss in performance.

We present a novel distributed video coding algorithm based on transform coding of distributed sources and exploiting the geometrical relationships between the location of the sensors. The geometry is used to align the video sequences and distributed quantization of transform coefficients is used to eliminate spatial and inter-sensor redundancy. In contrast with most of the current video compression standards which only exploit spatial and temporal dundancy within each video sequence, we also consider the significant redundancy between the sequences. Results demonstrate that our algorithm yields a significant saving in bit rate on the overlapping portion of multiple views.

Two different aspects of distributed source coding is discussed. First, a previously developed distributed uniform scalar quantization method is improved by adopting non-uniform quantization. It is observed that compared to uniform quantization, the non-uniform scheme further approaches to the distortion-rate bound by 0.5 bit. Turning then to sources with memory, the tradeoff between exploitation of time and space correlation is exposed. It is shown by example that optimal transforms deviate from their traditional counterparts. Specifically, optimal transforms in the distributed framework do not necessarily fully decorrelate each sequence in time