In this paper, we discuss a handover management scheme for Next Generation Self-Organized Networks. We propose to extract experience from full protocol stack data, to make smart handover decisions in a multi-cell scenario, where users move and are challenged by deep zones of an outage. Traditional handover schemes have the drawback of taking into account only the signal strength from the serving, and the target cell, before the handover. However, we believe that the expected Quality of Experience (QoE) resulting from the decision of target cell to handover to, should be the driving principle of the handover decision. In particular, we propose two models based on multi-layer many-to-one LSTM architecture, and a multi-layer LSTM AutoEncoder (AE) in conjunction with a MultiLayer Perceptron (MLP) neural network. We show that using experience extracted from data, we can improve the number of users finalizing the download by 18 %, and we can reduce the time to download, with respect to a standard event-based handover benchmark scheme. Moreover, for the sake of generalization, we test the LSTM Autoencoder in a different scenario, where it maintains its performance improvements with a slight degradation, compared to the original scenario.

Traffic prediction plays a vital role in efficient planning and usage of network resources in wireless networks. While traffic prediction in wired networks is an established field, there is a lack of research on the analysis of traffic in cellular networks, especially in a content-blind manner at the user level. Here, we shed light into this problem by designing traffic prediction tools that employ either statistical, rule-based, or deep machine learning methods. First, we present an extensive experimental evaluation of the designed tools over a real traffic dataset. Within this analysis, the impact of different parameters, such as length of prediction, feature set used in analyses, and granularity of data, on accuracy of prediction are investigated. Second, regarding the coupling observed between behavior of traffic and its generating application, we extend our analysis to the blind classification of applications generating the traffic based on the statistics of traffic arrival/departure. The results demonstrate presence of a threshold number of previous observations, beyond which, deep machine learning can outperform linear statistical learning, and before which, statistical learning outperforms deep learning approaches. Further analysis of this threshold value represents a strong coupling between this threshold, the length of future prediction, and the feature set in use. Finally, through a case study, we present how the experienced delay could be decreased by traffic arrival prediction.

The connectivity properties of ad hoc networks have been extensively studied over the past few years, from local observables, to global network properties. In this paper we introduce a novel layer of network dynamics which lives and evolves on top of the ad hoc network. Nodes are assumed selfish and a snow-drift type game is defined dictating the way nodes decide to allocate their cooperative resource efforts towards other nodes in the network. The dynamics are strongly coupled with the physical network causing the cooperation network topology to converge towards a stable equilibrium state, a global maximum of the total pay-off. We study this convergence from a connectivity perspective and analyse the inherent parameter dependence. Moreover, we show that direct reciprocity can be an efficient incentive to promote cooperation within the network and discuss the analogies between our simple yet tractable framework with D2D proximity based services such as LTE-Direct. We argue that cooperative network dynamics have many application in ICT, not just ad hoc networks, and similar models as the one described herein can be devised and studied in their own right.

Selfishness and altruism are two distinctive node behaviors in wireless networks. Users of network nodes, e.g. mobile phones, would like their individual devices to be operational for as long as possible. On the other hand, users and operators of wireless networks may wish for an overall long network lifetime via collaborative node operations. In contrast to the most of the work done till now, in this paper, we consider a node-centric scenario, where a node can balance its energy consumption according to its needs to prolong the node lifetime. More specifically, the energy balancing problem is defined as an inventory control problem, where the node may choose to be selfish or unselfish with various degrees having in mind a range of parameters, such as its battery state and the existence and rate of energy harvesting. Each node can use the proposed solution to allocate the available energy between internal (node itself) and external (collaborative) tasks subject to dependencies among application and communication layers and the node operating system. We also discuss a number of other possible extensions of the method to include interactions among different nodes in the network. To illustrate one such an extension, a case study with a large number of nodes among two separate but co-located wireless mesh networks is presented. By applying the inventory balancing algorithm, considerable increase in the network lifetime is shown for both networks.

How to choose an optimal number of parity bits at each transmission instant is one of the main problems in incremental redundancy (IR) coding. When the channel state information (CSI) is unknown, a consequence is that IR code throughput may be degraded considerably compared to the complete CSI case. For a communication subject to the unknown CSI distribution, this paper establishes a theoretical basis for a parity bit selection by means of a risk-sensitive optimal control. A set of optimal coding policies are obtained by the dynamic programming which provide a more stable throughput compared to the non-optimized IR coding. These policies have different levels of robustness which are measured by the so-called risk-sensitive parameter contained in the optimization objective function.

The linear minimum mean-squared error (MSE) channel estimator for systems employing per-subcarrier transmit antenna selection is developed. Frequency domain correlations after the selection process are shown to be approximated well using a simple function, which makes near-optimal channel estimation practically possible. However, the resultant estimators are not robust to errors in the assumed model, in terms of the antenna-to-subcarrier assignment used at transmission, as we motivate both analytically and via simulations. This is an issue when the channels for conveying this information to the receiver are severely constrained. We present windowed channel estimations and new robust estimation algorithms, either individually or in combination, as solutions to counter this sensitivity. The robust estimators are developed based on the principles of 1) minimising the worst case MSE over the set of possible models and 2) minimising the expected MSE over the set of possible models. The latter estimator is preferred due to the lower implementation complexity and better MSE performance. We also prove theoretical asymptotic (in signal-to-noise ratio) performances of the two estimators used with the proposed correlation model. Simulation results illustrate the performance gains and improved robustness offered by the developed estimators.

As the CMOS semiconductor technology enters nanometer regime, interconnect processes must be compatible with device roadmaps and meet manufacturing targets at the specified wafer size. The resulting ubiquitous process variations cause errors in data delivering through interconnects. This paper proposes an Information Theory based design method to accommodate process variations. Different from the traditional delay based design metric, the current approach uses achievable rate to relate interconnect designs directly to communication applications. More specifically, the data communication over a typical interconnect, a bus, subject to process variations (“uncertain” bus), is defined as a communication problem under uncertainty. A data rate, called the achievable rate, is computed for such a bus, which represents the lower bound on the maximal data rate attainable over the bus. When a data rate applied over the bus is smaller than the achievable data rate, a reliable communication can be guaranteed regardless of process variations, i.e., a bit error rate arbitrarily close to zero is achievable. A single communication strategy to combat the process variations is proposed whose code rate is equal to the computed achievable rate. The simulations show that the variations in the interconnect resistivity could have the most harmful effect regarding the achievable rate reduction. Also, the simulations illustrate the importance of taking into account bus parasitic parameters correlations when measuring the influence of the process variations on the achievable rates.

The linear minimum mean-squared error (MSE) channel estimator for systems employing per-subcarrier transmit antenna selection is developed. It is shown that the frequency domain correlations after the selection process is approximated well using a simple explicit function. Performance of resultant estimators, which assume a certain correlation model, deteriorates quickly at high signal-to-noise ratios (SNR) when model errors exist. To overcome this deficiency, robust estimators based on the principles of minimising the worst case and expected MSEs over the possible model set are developed. Solution to the latter is given in closed form, and is seen to achieve MSE optimal performance at low SNR while the high SNR performance is the same as that of the least-squares (LS) estimator. The optimal estimator with the proposed correlation function and correct model knowledge performs about 3 dB better than the LS estimator at a packet-error-rate of 0.01 for a channel coded system employing 16-QAM modulations, whereas in the presence of model errors, the latter robust estimator is performing about 2 dB better than the LS estimator.

This paper is concerned with multiple-input multiple-output (MIMO) wireless channel capacity, when the probability distribution of the channel matrix p(H) is not completely known to the transmitter and the receiver. The partial knowledge of a true probability distribution of the channel matrix p(H) is modelled by a relative entropyD(middot||middot) such thatD(p||pnom) lesd,dges 0, wheredis the distance from the so-called nominal channel matrix distribution pnom(H). The capacity of this compound channel is equal to the maximin of the mutual information, where the minimum is with respect to the channel matrix distribution, and the maximum is with respect to the covariance matrix of a transmitted signal. The existence of a minimizing probability distribution is proved, and the explicit formula for the minimizing distribution is derived in terms of the nominal distribution pnom(H) and parameterd. A number of properties of the mutual information, minimized over the set of channel distributions, are derived. Specifically, upper and lower bounds are derived for the minimized mutual information, while its convexity with respect todis shown. In the case of the Rayleigh fading, an explicit formula for the capacity and the optimal transmit covariance matrix are derived.

In this paper, achievable rates for compound Gaussian multiple-input–multiple-output (MIMO) channels are derived. Two types of channels, modeled in the frequency domain, are considered when: 1) the channel frequency response matrix$H$belongs to a subset of$H^{\infty}$normed linear space, and 2) the power spectral density (PSD) matrix of the Gaussian noise belongs to a subset of$L_1$space. The achievable rates of these two compound channels are related to the maximin of the mutual information rate. The minimum is with respect to the set of all possible$H$matrices or all possible PSD matrices of the noise. The maximum is with respect to all possible PSD matrices of the transmitted signal with bounded power. For the compound channel modeled by the set of$H$matrices, it is shown, under certain conditions, that the code for the worst case channel can be used for the whole class of channels. For the same model, the water-filling argument implies that the larger the set of matrices$H$, the smaller the bandwidth of the transmitted signal will be. For the second compound channel, the explicit relation between the maximizing PSD matrix of the transmitted signal and the minimizing PSD matrix of the noise is found. Two PSD matrices are related through a Riccati equation, which is always present in Kalman filtering and liner-quadratic Gaussian control problems.

We describe a multiple optical sources - multiple detectors scheme, based on either repetition MIMO or space-time coding and low-density parity-check (LDPC) codes. The proposed scheme is able to operate under strong atmospheric turbulence and provides excellent coding gains. The LDPC codes are designed using the concept of pairwise-balanced designs. Bit-error rates and achievable information rates are reported assuming non-ideal photodetection. To improve the spectral efficiency we employ the concept of bit-interleaved LDPC-coded modulation based on pulse-amplitude modulation.

The control of partially observed jump linear systems over jump communication channels is investigated when the Plant State Information (PSI) and the Channel State Information (CSI) are present at the transmitter and the receiver side of the communication channel. The PSI refers to the state of a Markov chain driving the plant dynamics, while the CSI is the state of a Markov chain driving the communication channel. Necessary conditions for observability and stabilizability in probability and r-th mean are derived by using information transmission theorem and Shannon's lower bound on the rate distortion. Based on the source-channel matching principle, a transmission technique is designed that achieves observability and stabilizability of the controlled jump system in mean-square sense.

This paper addresses the issue of multiple-input multiple-output (MIMO) wireless channel capacity, when the probability distribution of the channel matrix p(H) is not completely known to the transmitter and the receiver. The partial knowledge of a true probability distribution of the channel matrix is modelled by using a relative entropy D(pparq). All possible channel matrix distributions p(H) satisfy D(pparpnom) les d, d ges 0, i.e., they lie within d distance from the so-called nominal channel matrix distribution pnom(H). The information channel capacity is defined as a maximin optimization problem, where the mutual information is a pay-off function. The minimum is with respect to the channel matrix distribution, and the maximum is with respect to the covariance matrix of a transmitted signal. Based on the derived characteristics of the pay-off function, the formula for the channel matrix distribution, which minimizes the mutual information, is derived. In the case of the Rayleigh fading, the formula for the information capacity and the optimal transmit covariance matrix are obtained. In addition, the existence of the saddle point of the maximin optimization problem is established for this particular case

The paper gives a survey of spectrum shaping codes used for digital recording systems. This class of codes belongs to the broader class of modulation codes, which are widely used in recording systems for adjusting the source characteristics to the characteristics of the recording channel. The Shannon noiseless capacities of recording channels are considered, as well as the spectra of maxentropic sequences of M-ary recording constraints. In addition, some practical encoding and decoding schemes are discussed.

The power control of wireless networks is formulated using a stochastic optimal control framework, in which the evolution of the channel is described by stochastic differential equations (SDEs). The latter capture the spatio-temporal variations of the communication link, as well as the randomness. This class of models is more realistic than the static models usually encountered in the literature. Under this scenario, average and probabilistic Quality of Service (QoS) measures are introduced to evaluate the performance of any control strategy by using Chernoff bounds. Moreover, the Chernoff bound is computed explicitly, while the solution of the stochastic optimal power control is obtained through pathwise optimization. The pathwise optimization can be solved using linear programming if predictable control strategies are introduced. Finally, if predictable control strategies do not hold, it is shown that the proposed power control problem reduces to particular convex optimizations.