This article proposes using communication pipelining to enhance the convergence speed of federated learning in mobile edge computing applications. Due to limited wireless subchannels, a subset of the total clients is scheduled in each iteration of federated learning algorithms. On the other hand, the scheduled clients wait for the slowest client to finish its computation. We propose to first cluster the clients based on the time they need per iteration to compute the local gradients of the federated learning model. Then, we schedule a mixture of clients from all clusters to send their local updates in a pipelined manner. In this way, instead of just waiting for the slower clients to finish their computations, more clients can participate in each iteration. While the time duration of a single iteration does not change, the proposed method can significantly reduce the number of required iterations to achieve a target accuracy. We provide a generic formulation for optimal client clustering under different settings, and we analytically derive an efficient algorithm for obtaining the optimal solution. We also provide numerical results to demonstrate the gains of the proposed method for different data sets and deep learning architectures.

In this study, the authors present and evaluate the performance of two mode selection schemes for device‐to‐device (D2D) enabled unmanned aerial vehicle‐based wireless networks. The proposed schemes are based on a threshold received signal strength and an average threshold D2D distance to select the D2D mode. The focus of the two schemes is either to enhance the quality of the signal or the connectivity in case of emergency situations. To evaluate the performances of the schemes, they derive the corresponding expressions of the probability of using D2D mode and ergodic capacity. Numerical results show the advantage of the presented schemes in off‐loading traffic from aerial platforms and shed lights on the effect of the environment on the performance of D2D enabled aerial networks.

In this paper, we propose a new mode selection scheme for D2D enabled long term evolution-advanced (LTE-A) systems. The proposed scheme is based on a predefined threshold parameter related to the LTE sidelink (SL) received signal strength (RSS), where above this threshold, the D2D mode outperforms the cellular communication mode. Using the stochastic geometry, the system model is described and the expression of the proposed threshold parameter is derived. To evaluate the performance of the proposed scheme, the expressions of the probability of using D2D mode and the average ergodic capacity are detailed and derived. Based on that, Monte Carlo simulations are used to confirm the derived expressions and the advantages of the proposed scheme over related work in the literature.

This paper analyzes the performance of hybrid control-access schemes for small cells (such as femtocells) in the context of two-tier overlaid cellular networks. The proposed hybrid access schemes allow for sharing the same downlink resources between the small-cell network and the original macrocell network, and their mode of operations are characterized considering post-processed signal-to-interference-plus-noise ratios (SINRs) or preprocessed interference-aware operation. This paper presents a detailed treatment of achieved performance of a desired user that benefits from MIMO arrays configuration through the use of transmit antenna selection (TAS) and maximal ratio combining (MRC) in the presence of Poisson field interference processes on spatial links. Furthermore, based on the interference awareness at the desired user, two TAS approaches are treated, which are the signal-to-noise-based selection and SINR-based selection. The analysis is generalized to address the cases of highly-correlated and un-correlated aggregated interference on different transmit channels. In addition, the effect of delayed TAS due to imperfect feedback and the impact of arbitrary TAS processing are investigated. The analytical results are validated by simulations, to clarify some of the main outcomes herein.

This paper presents new approaches to characterize the achieved performance of hybrid control-access small cells in the context of two-tier multi-input multi-output (MIMO) cellular networks with random interference distributions. The hybrid scheme at small cells (such as femtocells) allows for sharing radio resources between the two network tiers according to the densities of small cells and their associated users, as well as the observed interference power levels in the two network tiers. The analysis considers MIMO transceivers at all nodes, for which antenna arrays can be utilized to implement transmit antenna selection (TAS) and receive maximal ratio combining (MRC) under MIMO point-to-point channels. Moreover, it tar-gets network-level models of interference sources inside each tier and between the two tiers, which are assumed to follow Poisson field processes. To fully capture the occasions for Poisson field distribution on MIMO spatial domain. Two practical scenarios of interference sources are addressed including highly-correlated or uncorrelated transmit antenna arrays of the serving macrocell base station. The analysis presents new analytical approaches that can characterize the downlink outage probability performance in any tier. Furthermore, the outage performance in high signal-to-noise (SNR) regime is also obtained, which can be useful to deduce diversity and/or coding gains.

This paper presents an enhanced overlay spectrum sharing scheme in a cognitive radio network, which consists of two pairs; a primary transmitter- primary receiver pair and a secondary base station-secondary receiver pair. The developed scheme is based on a three time slots overlay spectrum sharing, which exploits the maximum diversity gain for both the primary and secondary users. The performance of the developed scheme is assessed through the derivation of new closed form expressions for the bit error rate for arbitrary signal-to-noise ratio (SNR). In addition, we carry out an asymptotic analysis in the high SNR regime, which proves that the proposed scheme achieves a full diversity order for the PU and SU. Results of analytical expressions and numerical simulations both verify that the proposed scheme achieves significant gains for error performance, for both the PU and SU, in comparison to other baseline schemes.

In classical information and communication theory, channel capacity and achievable rate over Gaussian channels, whether for constant or fading channels and whether for single user or multiple users, are characterized as functions of the power that is used for the transmission of the information. The associated circuit power that is needed in real communication transceivers to power up the electronics and to do all required signal processing is typically not involved in the capacity characterization. However, to achieve global energy efficiency of the communication systems, the adaptation of the communication schemes should be based on total power consumption. In this paper, we characterize the maximum achievable rates over many fundamental channels taking total power constraints into considerations.

We consider the design of space-time codes for the multiple-input multiple-output multicast communication systems with two classes of receivers. The first class comprises high-resolution (HR) receivers which have access to reliable channel state information (CSI) and can perform coherent detection, and the second class comprises low-resolution (LR) receivers which do not have access to CSI and can only perform non-coherent detection. We propose a layered encoding structure in which LR information available to both classes of receivers is encoded using Grassmannian constellations, and an incremental component, which is available only to the HR receivers, is encoded in the particular bases of the transmitted Grassmannian constellation points, thereby giving rise to constellations on the Stiefel manifold. The proposed structure enables reliable coherent communication of the HR information without compromising the reliability with which the basic LR information is non-coherently communicated. To effect rate-efficient communication of the incremental, HR layer, we use optimization methods on the Stiefel manifold to develop a novel technique for designing the unitary constellations directly. This approach alleviates the restriction imposed by the traditional techniques in which unitary space-time codes are constructed from scalar constellations. As such, this approach enables better control of the distance spectrum of the developed constellations and more effective utilization of the degrees of freedom that underlie the Stiefel manifold. For the LR receivers, we use maximum likelihood detection, whereas for the HR receivers, we develop a computationally-efficient two-step sequential detector which detects the LR information prior to detecting the incremental component superimposed on it. The detectors and the layered structure with the aforementioned constellations enable full diversity and maximum degrees of freedom to be achieved on the Grassmann and Stiefel manifolds.

In this paper, we provide a comprehensive analytical framework on the performance of multiple-input multiple-output (MIMO) mutlihop amplify-and-forward relay network employing transmit antenna selection (TAS) with receive maximal ratio combining (MRC) in the presence of randomly located interferers. The channel fading models per hop are assumed to be independent and follow exponential random distribution, and the interference is distributed according to the spatial Poisson point process PPP, which is further known as symmetric alpha stable distribution. The impact of spatial dependence across multiple antennas is facilitated by considering three different network interference models: 1) the isotropic model, where all receive antennas see interference belonging to the same Poisson point process (PPP); 2) the independent model, where each receive antenna sees interference belonging to a different independent PPP; and 3) the mixture model, where each receive antenna sees the superposition of the two previous models. The end-to-end transmission is performed using two TAS modes, which includes signal-to-noise ratio (SNR)-based selection, and signal-to-interference noise ratio (SINR)-based selection. The analysis provides new analytical results for the outage probability and symbol error rates (SER). In addition, the impact of feedback delay on the performance of the proposed system for SNR-based selection is investigated. We further study the optimum resource allocation strategies including power allocation, position allocation, and joint allocation for power and position. The obtained results are clarified through selected numerical and simulated examples.

In this paper, the long-term average achievable rate over block-fading buffer-aided relay channels is maximized using a hybrid scheme that combines three essential transmission strategies, which are decode-and-forward, compress-and-forward, and direct transmission. The proposed hybrid scheme is dynamically adapted based on the channel state information. The integration and optimization of these three strategies provide a more generic and fundamental solution and give better achievable rates than the known schemes in the literature. Despite the large number of optimization variables, the proposed hybrid scheme can be optimized using simple closed-form formulas that are easy to apply in practical relay systems. This includes adjusting the transmission rate and compression when compress-and-forward is the selected strategy based on the channel conditions. Furthermore, in this paper, the hybrid scheme is applied to three different models of the Gaussian block-fading buffer-aided relay channels, depending on whether the relay is half or full duplex and whether the source and the relay have orthogonal or non-orthogonal channel access. Several numerical examples are provided to demonstrate the achievable rate results and compare them to the upper bounds of the ergodic capacity for each one of the three channel models under consideration.

In this paper, we develop two approaches for designing unitarily-constrained space-time block codes, which are suitable for communicating high-resolution information in layered multiple-input multiple-output broadcast channels. Unlike existing space-time codes, which are usually synthesized from standard phase-shift keying (PSK) or quadrature amplitude modulation constellations, the space-time codes proposed herein are designed using direct optimization over the unitary group. In comparison with conventional unitary space-time block codes, including Alamouti code with PSK constellations, the space-time codes generated by the proposed approaches exhibit significantly better performance, more favorable distance spectra and more effective utilization of the degrees of freedom that underlie the unitary group.

This work proposes a suboptimal multi-user scheduling scheme for Gaussian broadcast channels which improves upon the classical single user selection, while considerably reducing complexity as compared to the optimal superposition coding with successful interference cancellation. The proposed scheme combines the two users with the maximum weighted instantaneous rate using superposition coding. The instantaneous rate and power allocation are derived in closed-form, while the long term rate of each user is derived in integral form for all channel distributions. Numerical results are then provided to characterize the prospected gains of the proposed scheme.

The aim of this work is to maximize the long-term average achievable rate region of a primary and a secondary source-destination pairs operating in an overlay setup over block-fading channels. To achieve this objective, we propose an opportunistic strategy to grant channel access to the primary and secondary sources based on the channel conditions in order to exploit the available multiple-link diversity gains in the system. The secondary source has causal knowledge of the primary messages and it acts as a relay of the primary source in return for getting access to the channel. To maximize the gains of relaying, the relay and destination are equipped with buffers to enable the use of channel-aware adaptive link selection. We propose and optimize different link selection policies and characterize their expected achievable rates. Also, we provide several numerical results to demonstrate the evident mutual benefits of buffer-aided cooperation and adaptive link selection to the primary and the secondary source-destination pairs.

We consider the design of space-time codes for multi-resolution multiple-input multiple-output (MIMO) broadcast communication systems. Two classes of receivers are considered: high-resolution (HR) receivers, which have access to reliable channel state information (CSI) and can perform coherent detection, and low-resolution (LR) receivers which do not have access to CSI and can only perform non-coherent detection. We propose a layered encoding structure, whereby, for the LR receivers, the transmitted codewords are chosen to be points on the Grassmann manifold whereas, for the HR receivers, incremental information is encoded in the particular bases of the transmitted codewords, thereby representing points on the Stiefel manifold. For the HR receivers, we develop a computationally-efficient two-step detector. Using this detector, we show that the proposed structure enables reliable coherent communication of the incremental HR information without compromising the reliability with which the basic LR information is non-coherently communicated. We also show that this structure enables full diversity to be achieved for both LR and HR receivers. Finally, we show that this structure achieves the maximum number of degrees of freedom for non-coherent LR channels and coherent HR channels with unitarily-constrained input signals.

In this paper, we consider a fading relay channel where the source uses two layers source coding with successive refinement. The two source layers are transmitted using superposition coding at the source and relay with optimal power allocation, and successive interference cancellation at the receivers (i.e. relay and destination). The power allocation for the two layers at the source and relay is subject to optimization in order to maximize the expected user satisfaction that is defined by a utility function of the total decoded rates at the destination. We assume that only the channel statistics are known. The relay is half-duplex and applies decode and forward. We characterize the expected utility function in terms of the channel statistics of the fading channels, and we solve the optimization problem using the numerical random search method. We provide many numerical examples to show the prospected gains of using the relay on the expected utility for different channel conditions. Furthermore, we obtain that for some conditions, it is optimal to send only one layer.

In this paper, we analyze the performance of a multiple-input multiple-output (MIMO) multi-hop relay network using transmit antenna selection and maximal ratio combining (TAS/MRC) employing amplify-and-forward relay mode over a Rayleigh fading channels in the presence of Poisson field of interferers, that is applied practically in current cellular networks like LTE (Long Term Evolution). The aggregated interference per hop is modeled as shot noise with α-stable distribution, which is assumed to be non-identical per receive antenna. In particular, we derive compact analytical expressions for exact, and upper bound outage probability of the system. To gain further insights on the proposed system behaviors, we derive asymptotic outage probability at high signal-to-ratio (SNR) which explicitly characterizes the diversity order and coding gain of the network. All derived analytical results are corroborated by simulation results to show the correctness of our derived results.

In this paper, we consider a novel layered coding approach with two layers. One of the two layers, denoted by the base-layer, can be received by any receiver even if it does not have reliable channel estimates. The other, refining-layer can only be received by any receiver that has channel state information. We propose signal constellations that allow the transmission of coherent and non-coherent information for the single-antenna transmitters. We derive upper bounds for the pairwise error probability for the coherent and non-coherent receivers and prove that our proposed signal constellations can achieve a diversity of order M for the 1 × M system, for both the coherent and non-coherent receivers.

In this paper, we obtain the optimal resource allocation scheme in order to maximize the achievable rate region in a dual-hop system that consists of two independent source-destination pairs sharing a single half-duplex relay. The relay decodes the received information and possesses buffers to enable storing the information temporarily before forwarding it to the respective destination. We consider both non-orthogonal transmission with successive interference cancellation at the receivers and orthogonal transmission. Also, we consider Gaussian block-fading channels and we assume that the channel state information is known and that no delay constraints are required. We show that, with the aid of buffering at the relay, joint user-and-hop scheduling is optimal and can enhance the achievable rate significantly. This is due to the joint exploitation of multiuser diversity and multihop diversity in the system. We provide closed-form expressions to characterize the average achievable rates in a generic form as functions of the statistical model of the channels. Furthermore, we consider sub-optimal schemes that exploit the diversity in the system partially and we provide numerical results to compare the different schemes and demonstrate the gains of the optimal one.

In this paper we consider joint optimization of rate and power for communication systems that use multilayer source coding with successive information refinement, accompanied with a broadcast approach at the physical layer of the system. We analyze the problem under the assumption of Rayleigh fading channels where rates and power ratios of the source layers are jointly optimized based on channel statistics information, with the objective of maximizing the expected user satisfaction which is usually defined by a differentiable concave increasing utility function of the total decoded rate. As special cases, we consider two utility functions; namely, the expected total decoded rate at the receiver and the expected rate distortion of a Gaussian source. We show that the optimal solution can be obtained using a two-dimensional bisection search for any number of layers. The outer bisection search is over the Lagrangian dual variable and the inner bisection search is over the decoding threshold of the layer. Moreover, we show that with a small number of layers, we can approach the performance upper bound that is achieved by transmitting an infinite number of layers.

In this paper we consider joint optimization of rate and power for communication systems that use multilayer superposition source coding with successive refinement of information. We assume a Rayleigh fading channel, where rates and power are jointly and optimally allocated between the source layers based on channel statistics information, with the objective of maximizing the expected total received rate at the end user. We show that the optimization problem possesses a strong duality, and hence we use the dual form to show that the optimal solution can be obtained using a two-dimensional bisection search for any number of layers. The outer bisection search is over the Lagrangian dual variable and the inner bisection search is over the decoding SNR threshold of the last layer. Moreover, we show that with a small number of layers, we can approach the performance upper bound, that is achieved by transmitting an infinite number of layers.

In this paper, we consider a two-way communication system in which two users communicate with each other through an intermediate relay over block-fading channels. We investigate the optimal opportunistic scheduling scheme in order to maximize the long-term average transmission rate in the system assuming symmetric information flow between the two users. Based on the channel state information, the scheduler decides that either one of the users transmits to the relay, or the relay transmits to a single user or broadcasts to both users a combined version of the two users' transmitted information by using linear network coding. We obtain the optimal scheduling scheme by using the Lagrangian dual problem. Furthermore, in order to characterize the gains of network coding and opportunistic scheduling, we compare the achievable rate of the system versus suboptimal schemes in which the gains of network coding and opportunistic scheduling are partially exploited.

We consider the problem of minimizing the expected distortion in the multilayer transmission of a Gaussian source using the broadcast approach with successive information enhancement. This minimization is contingent on the jointly optimal choice of the rates and power ratios of the different layers. This problem was tackled in the literature with the assumption that the fading channel has a finite number of states and the number of source layers matches the number of channel states. In this paper, we provide a more generic solution for a continuous Rayleigh fading channel, and for any predetermined number of layers. We prove that the primal optimization problem has a strong duality with the Lagrangian dual problem. Consequently, we propose a two-dimensional bisection search algorithm that can, for any number of layers, find the optimal solution of the dual problem which will be the same as the optimal solution of the primal problem. The complexity of the search algorithm has linear order with respect to the number of layers. We provide numerical results for the optimal rate and power allocation. Moreover, we show that with a small number of layers, we can approach the distortion lower bound that is achieved by transmitting an infinite number of layers.

In this paper, we maximize the achievable rate region of a dual-hop network with two sources serving two users independently through a single shared relay. We formulate the problem as maximizing the sum of the weighted long term average throughputs of the two users under stability constraints on the long term throughputs of the source-user pairs. In order to solve the problem, we propose a joint user-and-hop scheduling scheme, which schedules the first or second hop opportunistically based on instantaneous channel state information, in order to exploit multiuser diversity and multihop diversity gains. Numerical results show that the proposed joint scheduling scheme enhances the achievable rate region as compared to a scheme that employs multi-user scheduling on the second-hop alone.

We propose joint user-and-hop scheduling over dual-hop block-fading broadcast channels in order to exploit multi-user diversity gains and multi-hop diversity gains all together. To achieve this objective, the first and second hops are scheduled opportunistically based on the channel state information. The joint scheduling problem is formulated as maximizing the weighted sum of the long term achievable rates of the users under a stability constraint, which means that in the long term the rate received by the relay should equal the rate transmitted by it, in addition to power constraints. We show that this problem is equivalent to a single-hop broadcast channel by treating the source as a virtual user with an optimal weight that maintains the stability constraint. We show how to obtain the source weight either off-line based on channel statistics or on real-time based on channel measurements. Furthermore, we consider special cases including the maximum sum-rate scheduler and the proportional fair scheduler. We also show how to extend the scheme into one that allows multiple user scheduling via superposition coding with successive decoding. Numerical results demonstrate that our proposed joint scheduling scheme enlarges the rate region as compared to scheduling schemes that exploit the diversity gains partially.

Multiuser switched-diversity scheduling schemes were recently proposed in order to overcome the heavy feedback requirements of conventional opportunistic scheduling schemes by applying a threshold-based, distributed, and ordered scheduling mechanism. The main idea behind these schemes is that slight reduction in the prospected multiuser diversity gains is an acceptable trade-off for great savings in terms of required channel-state-information feedback messages. In this work, we characterize the achievable rate region of multiuser switched diversity systems and compare it with the rate region of full feedback multiuser diversity systems. We propose also a novel proportional fair multiuser switched-based scheduling scheme and we demonstrate that it can be optimized using a practical and distributed method to obtain the feedback thresholds. We finally demonstrate by numerical examples that switched-diversity scheduling schemes operate within 0.3 bits/sec/Hz from the ultimate network capacity of full feedback systems in Rayleigh fading conditions.

We investigate the power allocation problem in layered transmission with successive enhancement using the broadcast approach. This problem is particularly important for multimedia streaming applications and multicasting services over slowly fading channels with partial channel information at the transmitter. We consider the practical scenario of limited and pre-specified source layers, and we propose a generic and efficient solution to optimize the transmission power allocated to each layer in order to maximize the average user satisfaction defined by a utility function of the decoded layers. We assume a block-fading Rayleigh distributed channel model. However, our solution can be extended to a number of other channel distributions. Although the direct formulation of the problem is not convex, we propose change of optimization variable, and we use the dual problem and the perturbation analysis in order to obtain the optimal solution. The solution is based on a one-dimensional bisection search irrespective of the total number of layers. Our optimality analysis indicates that in some cases it may not be optimal to transmit all the layers.

A new multiuser scheduling scheme is proposed and analyzed in this paper. The proposed system combines features of conventional full-feedback selection-based diversity systems and reduced-feedback switch-based diversity systems. The new hybrid system provides flexibility in trading-off the channel information feedback overhead with the prospected multiuser diversity gains. The users are clustered into groups, and the users' groups are ordered into a sequence. Per-group feedback thresholds are used and optimized to maximize the system overall achievable rate. The proposed hybrid system applies switched diversity criterion to choose one of the groups, and a selection criterion to decide the user to be scheduled from the chosen group. Numerical results demonstrate that the system capacity increases as the number of users per group increases, but at the cost of more required feedback messages.

We provide the solution for optimizing the power and resource allocation over block-fading relay-assisted broadcast channels in order to maximize the long term average achievable rates region of the users. The problem formulation assumes regenerative (repetition coding) decode-and-forward (DF) relaying strategy, long-term average total transmitted power constraint, orthogonal multiplexing of the users messages within the channel blocks, possibility to use a direct transmission (DT) mode from the base station to the user terminal directly or a relaying (DF) transmission mode, and partial channel state information. We show that our optimization problem can be transformed into an equivalent "no-relaying" broadcast channel optimization problem with each actual user substituted by two virtual users having different channel qualities and multiplexing weights. The proposed power and resource allocation strategies are expressed in closed-form that can be applied practically in centralized relay-assisted wireless networks. Furthermore, we show by numerical examples that our scheme enlarges the achievable rates region significantly.

In this paper we consider multicast systems that use layered source coding with successive refinement in a Rayleigh fading environment. For this kind of systems, we optimally allocate the transmission power, and hence the decoding threshold of each layer, in order to maximize the expected satisfaction defined by a utility function of the decoded rates of the users. In particular, we use the Lagrangian dual problem in order to efficiently solve the primal non-convex problem for any number of layers. Furthermore, we show that for some channel statistics, it is not optimal to transmit all the layers, rather we should allocate zero power to some of the upper layers in order to maximize the expected user satisfaction.

A modification is proposed for the formula known from the literature that characterizes the boundary of the capacity region of Gaussian multiaccess fading channels. The modified version takes into account potentially negative arguments of the cumulative distribution function that would affect the accuracy of the numerical capacity results.

In this paper, we consider the design of channel-aware scheduling policies for centralized wireless networks with resource-sharing constraints. The work is motivated by the objective to utilize the wireless system resources efficiently in addition to meeting the resource-sharing constraints. Our solutions are based on scheduling policies achieving close-to-capacity performance. We suggest two algorithms - an offline and an online solution - to properly adjust the weighting factors of the scheduling policy in order to achieve the required resource-sharing constraints.

An algorithm for efficient encoding of LDPC codes is presented that does not impose any restrictions on the construction of the parity-check matrices. The algorithm modifies the parity check matrix, without changing the subspace spanned by its rows, by removing linear dependent rows and adding a small number of new rows such that the graph-based message-passing encoder will not get stuck in a stopping set. The added rows are designed by a new algorithm which is based on the notion of the "key set". The encoder exploits the sparseness of the parity-check matrix, and the encoding complexity grows almost linear with the blocksize, because the number of added rows, which may not be sparse, is relatively small.