In this article, we provide a set of federated learning (FL) protocols for future Internet architectures, which integrate the edge computing with the Internet of Things (IoT) known as “IoT edge computing.” The proposed protocols aim to the efficient implementation of the FL, i.e., distributed intelligence, in future IoT networks, where edge computing will leverage the overall procedure at the edge of the network. We first provide a list of application requirements for such an FL implementation, which result in the architecture of constrained and nonconstrained IoT devices based on a set of Internet engineering task force (IETF) standards. The FL protocols consist of three stages as follows: 1) initial configuration; 2) distributed training; and 3) cloud updates. The specified FL protocols are tested using an experimental IoT platform, which is used to obtain experimental results that provide the performance evaluation of the FL protocols in terms of accuracy, time, and latency. We propose those FL protocols for the next-generation Internet (NGI), where IoT, edge computing, and FL will be blended efficiently for future Internet applications.

This paper proposes the cooperative cognitive net-work slicing virtualization solution for smart Internet of things (IoT) applications. To this end, we deploy virtualized small base stations (vSBSs) in SDR devices that offer network-slicing virtualization option. The proposed virtualized solution relies on Fed4Fire wireless experimental platform. In particular, we assume that multiple IoT devices can have access to different vSBSs, which coordinate their resources in a cooperative manner using machine learning (ML). To this end, a proactive resource management is deployed in the unlicensed band, where a cooperative solution is facilitated using the licensed band. The cooperative network slicing is managed and orchestrated using small cell virtualization offered by the Fed4Fire . Experimental trials are carried out for certain number of users and results are obtained that highlight the benefit of employing cooperative cognitive network slicing in future virtualized wireless networks.

Industrial Internet of Things (IIoT) networks are considered the large-scale deployment of IoT devices for industrial applications such as smart manufacturing, harvesting and supply chain management. The Internet of Things (IoT) devices are typically connected over a wireless medium, given the large geographical distribution area and the increasing demand for flexible installations. In some cases, a combination of wired and wireless connectivity can be assumed as common practice. In both scenarios, wireless communications for IIoT networks is a fundamental component of the system architecture that needs to satisfy stringent requirements such as reliable connectivity and minimal delays. Therefore, the dependability of wireless communications for IIoT networks should be carefully studied to provide new solutions, which can guarantee that applications can meet their real-time and reliability requirements while optimizing the control capability of the overall network. This paper focuses on the dependable wireless communications in the IIoT networks, where wireless control and monitoring tasks need to meet stringent real-time and reliability constraints. After reviewing recent solutions and discussing their suitability for IIoT networks, we highlight the yet open challenges that needs to be tackled by both academia and industry.

In this paper, the problem of sub-channel selection for carrier aggregation (CA) systems is examined. CA enables the achievement of high data rate links via simultaneous transmissions over multiple component carriers. A CA system usually occupies only a limited number of sub-channels M of these components due to limitations on the maximum permitted number of sub-channels per system. From an information theoretic point of view, a CA system should detect and employ the M-best sub-channels out of the N available ones. To that end, such a system probes a subset of sub-channels during each coherence time via pilot transmission. Then, for the best M sub-channels, one-bit feedback information is transmitted in order to prohibit the transmission through sub-channels with gain below a threshold. The aim is to derive tractable forms via employing the extreme value theory for the sum rate (lower bound on ergodic capacity) achieved by the system under Rayleigh fading and then, to optimize jointly the training length and power, the number of the probed sub-channels (probing bandwidth size) and the feedback threshold such that the sum rate is maximized by considering the sub-channel estimation error. The accuracy of the theoretical analysis is verified by numerical results.

In this paper, we study distributed power allocation for multiflow carrier aggregation in cognitive cellular networks. Our approach differs from the conventional water filling (WF) algorithm since we deal with heterogeneous fading channels, wherein all of the Lagrange multipliers are not handled equally over the heterogeneous cells. The distributed power control solution is carried out over heterogeneous fading channels that are considered nonidentically distributed Nakagami-m fading channels. We first formulate the optimization problem, and we then solve it using the alternating direction method of multipliers (ADMM), which allows the required decomposition for each channel and the required statistical learning among the different subproblem solutions. We also provide comparison with the dual decomposition method and WF solutions considered as solutions without cognition. Simulation results highlight the performance gain of ADMM in terms of the number of iterations. Furthermore, we provide a multiuser application scenario, where the analysis on the fading channel model with Nakagami-m distribution is carried out. We distinguish into the synchronous and the asynchronous ADMM implementations tackling the asynchronous user updates that can be found in a heterogeneous network deployment. The simulation results are obtained to highlight the impact of the partial barrier and the bounded delay in the asynchronous use case.

Current application of carrier aggregation (CA), a key enabling technology for wideband communications over fragmented spectrum, is limited to only a small number of component carriers (CCs). However, wireless broadband connectivity of future networks requires the exploitation of multiple, contiguous, and non-contiguous CCs (NC-CCs). By considering a large number of aggregated NC-CCs, the concept of massive non-contiguous CA (NC-CA) is realized, which offers the highest flexibility in spectrum aggregation. In order to fully exploit the system-level benefits of massive NC-CA, transceivers capable of efficiently accommodating parallel transmissions over all of the available NC-CCs are required. This is a challenging task in which current NC-CA transceiver architectures, including the straightforward parallel chains implementation, fail to accomplish, as they suffer from scalability and efficiency issues. Focusing on the digital baseband processing part of the transceiver, this paper proposes an efficient filter-bank transceiver architecture able to accommodate a large number of NC-CCs with a small complexity increase compared with the ideal case of the transceiver operating on a single CC with the same aggregate bandwidth. Specification of this new architecture is cast as a filter design problem with minimal assumptions on transmitted waveform format, channel statistics, and decoding scheme, allowing for universally applicable designs. Both heuristic and optimal design methodologies are provided, the latter obtained using the notion of capacity mean square error as an appropriate performance/design metric. The performance of the proposed architecture is evaluated analytically, and a case study where it is applied for OFDM transmissions is provided.

This paper considers the problem of analytically assessing the maximum achievable rates (capacity) of orthogonal frequency division multiplexing (OFDM) transmissions when the receiver, for complexity reasons, only accounts for the common phase error (CPE) effect due to phase noise impairments with inter-carrier interference (ICI) treated as noise. By recognizing that the functional form of the CPE with respect to the phase noise realization is actually a free design parameter, determination of the capacity is posed as a functional optimization problem with respect to the, so-called, CPE function. A simple lower bound of the capacity is obtained, revealing the performance degradation due to the unknown CPE at the receiver, as well as the suboptimal performance achieved in severe phase noise conditions by the conventional CPE function that is routinely employed in previous works. The existence of an optimal number of subcarriers that balances the effects of the (unknown) CPE and ICI is highlighted and critical system design/operation issues, such as selection of the CPE function and effect of unknown channel on the achievable rate, are discussed. The analysis in this paper can be employed for determining the suitability of OFDM in phase noise channels and provides a tractable utility function for resource allocation purposes.

The extension of long term evolution (LTE) networks in unlicensed spectrum areas under the licensed assisted access concept aims at achieving higher transmission rates via the aggregation of the aforementioned bands along with the licensed ones within the 3G Partnership Project framework. A prospect carrier aggregation (CA) scheme should handle efficiently the coexistence of the LTE systems that compete for the same unlicensed spectrum areas along with their incumbent users (i.e., Wi-Fi). In this paper, a novel CA scheme is proposed for licensed/unlicensed MIMO LTE systems that allocates optimally the resources (power and resource blocks) of an evolved Node B to user equipments. Furthermore, the proposed approach handles the coexistence matters within the unlicensed bands with an efficient decentralized way. The new scheme involves the solution to a mixed integer nonlinear programming problem and thus, an optimal low complexity method is proposed based on the Lagrange dual decomposition. Furthermore, the proposed technique is extended to the imperfect channel state information (CSI) case. To that end, a novel listen-before-talk scheme is developed via which the required unlicensed bands CSI are estimated in a blind manner. The performance of all of the proposed techniques is verified via indicative simulations.

Inter-band carrier aggregation (CA) is considered as the key enabling technology towards a wideband communication system that can exploit and aggregate non-contiguous spectrum from heterogeneous bands. However, the commonly suggested transceiver design using independent baseband (signal processing) modules for each band (component carrier) is not efficient and scalable. This paper addresses this challenge, by proposing a new filter-bank based transceiver design, able to achieve CA of two component carriers using a single baseband module. Appropriate information-theoretic performance metrics are employed in order to obtain universal designs that are independent of modulation format, channel statistics and decoding scheme. Various design methodologies are considered, whose performance is investigated numerically and analytically. It is demonstrated that there exist transceiver designs achieving two-component CA with small additional complexity compared to the conventional single component carrier transmission.

In this paper a novel low complexity scheme that achieves also reduced feedback overhead is studied for reporting the channel state information in Carrier aggregation MIMO systems. Such systems usually involve a large number of carriers and users, so the calculation and reporting of the required RI/PMI/CQI indexes could result in high computational complexity and feedback overhead. Given that the user scheduling and rate is based on these reported values, efficient computation and feedback schemes are crucial for achieving performances that can accommodate the wireless systems requirements. To that end, at first we propose a novel scheme for the RI/PMI/CQI calculation that exploits the adjacent channels correlation in order to reduce the complexity. Then, a novel distributed scheduling scheme is proposed based on a stable matching approach that requires reduced feedback information. The performance of the proposed approach is evaluated by indicative simulations.

The Energy Efficiency (EE) of Carrier Aggregation (CA) enabled communications systems is studied. In more detail, the problem of the optimal power allocation for CA enabled systems exploiting multiple Parallel Frequency Bands (PFBs) is addressed based on an EE maximization criterion. Both the case of full Channel State Information at the Transmitter (CSIT) as well as the case of quantized CSIT are studied and novel EE-optimal power allocation algorithms are derived. The derived algorithms are used in order to investigate the relation between the system's EE and the system design parameters, e.g. the number of PFBs, the number of feedback bits and the fading conditions. Focusing on the problem of optimal feedback allocation, it is shown by means of simulations that the optimal allocation of feedback to the PFBs is related to the fading conditions. Moreover, focusing on the problem of the selection of the number of PFBs, it becomes evident that the optimal selection of the number of PFBs is strongly related to the number of available feedback bits. Thereby, the interplay between EE-optimal feedback and power allocation design and system parameters is explored.

In this paper a novel Carrier Aggregation (CA) scheme is proposed for downlink MIMO LTE-A Systems. The proposed approach achieves increased transmission rates by establishing the communication links via both licensed and unlicensed bands without generating or experiencing interference to/from the users of the latter bands. To that end, a rate optimization problem is defined and solved subject to the previous zero interference constraints, a total power constraint and a maximum number of aggregated bands constraint. It turns out that the previous problem is a Mixed Integer Non Linear Programming (MINLP) one that requires an exhaustive search procedure in order to be solved. To tackle this, an optimal low complexity method is proposed based on the Lagrange dual decomposition. The performance of the original (MINLP) and the low-complexity proposed techniques is verified via indicative simulations.

One of the core technologies being standardized by 3GPP for LTE-A is the introduction of Relay Nodes (RNs). RNs are intended for ensuring coverage at cell edges as well as for the provision of enhanced capacity at hot spot areas. An extension to this concept is the Mobile Relay (MR). MRs can be mounted on vehicles and the original idea is to serve users inside high speed trains thus counter fighting the inherent severe fading and vehicle penetration loss. In this work we present a framework for exploiting Mobile Relay (MRs) even at low speeds in urban environments for bypassing traffic from nearby users, either within or outside the vehicles. In particular we show that apart from increased capacity and good quality coverage this approach achieves important energy savings for the mobile terminals.

In this paper, we investigate the performance of multi-user diversity (MUD) with optimal power allocation (OPA) in spectrum sharing (SS) under average interference power (AIP) constraint. In particular, OPA through average transmit power constraint in conjunction with the AIP constraint is assumed to maximize the ergodic secondary capacity. The solution of this problem requires the calculation of two Lagrange multipliers instead of one as obtained for the peak interference power (PIP) constraint and calculated using the well known water-filling algorithm. To this end, an algorithm based on bisection method is devised in order to calculate both Lagrange multipliers iteratively. Moreover, Rayleigh and Nakagami-m fading channels with one and multiple primary users are considered to derive the required end-to-end SNR analysis. Numerical results are depicted to corroborate our performance analysis and compare it with the PIP case highlighting hence, the impact of the AIP constraint compared to the PIP constraint application.

Cognitive relaying has been introduced for opportunistic spectrum access systems by which a secondary node forwards primary packets whenever the primary link faces an outage condition. For spectrum sharing systems, cognitive relaying is parametrized by an interference power constraint level imposed on the transmit power of the secondary user. For sensing-based spectrum sharing, the probability of detection is also involved in packet relaying control. This paper considers the choice of these two parameters so as to maximize the secondary nodes' throughput under certain constraints. The analysis leads to a Markov decision process using dynamic programming approach. The problem is solved using value iteration. Finally, the structural properties of the resulting optimal control are highlighted.

The development, delivery and management of mobile services are the subject of many research activities in both the academia and industry. The ultimate goal of these efforts is a dynamic environment that enables the delivery of situation-aware, personalized multimedia services over heterogeneous, ubiquitous infrastructures, commonly termed as systems beyond 3rd generation (3G). Reconfigurability and adaptability are key aspects of the mobile systems beyond 3G. Reconfigurable mobile systems and networks introduce additional requirements and complexity. Using the existing network control plane is inadequate for the realization of the reconfigurability process. Introducing a reconfiguration management plane is very important for the deployment of network/system wide reconfigurability. We intend to discuss the basic functionality of RMP and respective interrelations.