Federated learning (FL) for large language models (LLMs) offers a privacy-preserving scheme, enabling clients to collaboratively fine-tune locally deployed LLMs or smaller language models (SLMs) without exchanging raw data. While parameter-sharing methods in traditional FL models solves number of technical challenges, they still incur high communication overhead and struggle with adapting to heterogeneous model architectures. Federated distillation, a framework for mutual knowledge transfer via shared logits, typically offers lower communication overhead than parameter-sharing methods. However, transmitting logits from LLMs remains challenging for bandwidth-limited clients due to their high dimensionality. In this work, we focus on a federated LLM distillation with efficient communication overhead. To achieve this, we first propose an adaptive Top-k logit selection mechanism, dynamically sparsifying logits according to real-time communication conditions. Then to tackle the dimensional inconsistency introduced by the adaptive sparsification, we design an adaptive logits aggregation scheme, effectively alleviating the artificial and uninformative inputs introduced by conventional zero-padding methods. Finally, to enhance the distillation effect, we incorporate LoRA-adapted hidden-layer projection from LLM into the distillation loss, reducing the communication overhead further while providing richer representation. Experimental results demonstrate that our scheme achieves superior performance compared to baseline methods while effectively reducing communication overhead by approximately 50%.

The growing complexity of next-generation telecommunication networks demands more intelligent and adaptive management approaches to address the diverse requirements of emerging services. This paper proposes a novel network slicing framework that integrates the Multi-Access Technology Real-Time Intelligent Controller (mATRIC) within the O-RAN system, enabling resource unification of 5G and Wi-Fi networks and efficient allocation. The framework dynamically adapts to unify the 5G and Wi-Fi access technologies. Specifically, we propose slice deployment policies based on two Deep Reinforcement Learning (DRL) algorithms: the on-policy Trust Region Policy Optimization (TRPO) and the off-policy Soft Actor-Critic (SAC) approaches, which aims to optimize slice deployments across varying service requirements. To evaluate the effectiveness of the proposed policies, we compare them against the on-policy Advantage Actor-Critic (A2C) and off-policy Deep Deterministic Policy Gradient (DDPG) methods. Extensive simulations demonstrate that the on-policy algorithms achieve superior performance in terms of slice acceptance ratios and resource utilization, highlighting their potential to enhance slice deployment in future heterogeneous networks beyond 5G and into 6G.

The integration of sensing and communication would enable wireless networks to monitor the surrounding environment in addition to communications tasks. In other words, the sensing capability, traditionally used in radar systems, would be integrated into the communication system to build a symbiotic framework known as integrated sensing and communication (ISAC). Since radar sensing and wireless communication share similar characteristics, this integration would lead to spectrum efficiency and reduction in hardware cost compared to two separate systems. However, the co-design of sensing and communication poses several challenges and complexities in the physical and network layers, as well as security and privacy aspects of wireless systems. While there is a wealth of research addressing the above-mentioned challenges, a gap still remained between current research activities and the requirements of ISAC. Recently, 3GPP Rel-19 introduced 32 ISAC use cases along with their requirements. This paper reviews the 3GPP use cases to identify the required technologies and expected sensing outcomes, highlighting the gap between current research activities and ISAC requirements. The paper further explores concepts required to support the use cases, such as positioning, and different sensing sources (e.g., ambient radio frequency and radar signals). Following this, we explore the mutual benefits of integrated sensing with communication, security, radio access networks, digital twin, advanced antenna technologies, and multiple physical dimension transmission. Finally, the paper details the current progress of 3GPP technical specification groups and studies open challenges, available tools, and datasets in ISAC.

The recently introduced QUIC protocol has greatly increased the flexibility of end-to-end transmissions on the Internet, surpassing the design limits of the most popular transport protocols: TCP and UDP. However, some of TCP’s main design principles were carried onto QUIC, which may not be suitable for real-time use-cases; primarily, full reliability, as it requires every packet to be retransmitted until acknowledged by the receiver. In this work, we present Dynamic Reliability (DRel), a partial reliability framework that allows for granular alteration of the reliability per packet at the transport layer. The framework is housed by QUIC and its multipath extension, yet offering no-ack and no-retransmit for the true meaning of unreliable packet transmission (congestion control does not impact and is not influenced by unreliable packets). The “dynamic” in DRel refers to interchangeable reliable and unreliable transmission, in one session and across multiple paths, depending on the volatility of the communication system; guided by reliability policies. Fluidly altering packet reliability may offer a means in meeting stringent 5G and Beyond transmission requirements, especially for xURLLC use-cases. We examine the performance of DRel in single- and multiple-path architectures through system-level simulation using Mininet. The results illustrate comparable performance to vital QoE metrics for the dynamic reliability policies compared to the original (MP)QUIC. Alternatively, the enhancements at the transport layer stem from a reduction in communication congestion by up to 80% for single- and multiple-path connections compared to the original (MP)QUIC. As a result, the amount of backlogged and out-of-order packets is reduced, downsizing intermediate and end-to-end buffer occupancies.

The promises of multipath transport are to aggregate bandwidth, improve resource utilisation and enhance reliability. In this article, we demonstrate that the way multipath coupled congestion control is defined today leads to a suboptimal resource utilisation when network paths are disjoint, i.e., they do not share a bottleneck link. With growing interest in standardising Multipath QUIC (MPQUIC), we have implemented the practical Shared Bottleneck Detection (SBD) algorithm from RFC8382 in MPQUIC (MPQUIC-SBD). Through extensive experiments, we evaluate MPQUIC-SBD in the context of video streaming with various Adaptive Bitrate (ABR) algorithms, addressing both ABR classes of rule- and learning-based solutions. We demonstrate that MPQUIC-SBD accurately detects shared bottlenecks over 90% of the time, depending on the ABR algorithm, as the size of the video segments increases. In non-shared bottleneck scenarios, when MPQUIC-SBD detects that its QUIC subflows do not share the same network resources, it decouples their congestion windows accordingly, enabling video throughput gains of up to 37% compared to MPQUIC. These gains translate directly into improved video quality metrics, including higher bitrate, better resolution and reduced buffering, resulting in an enhanced quality of experience for users.

Time-triggered federated learning, in contrast to conventional event-based federated learning, organizes users into tiers based on fixed time intervals. However, this network still faces challenges due to a growing number of devices and limited wireless bandwidth, increasing issues like stragglers and communication overhead. In this paper, we apply model pruning to wireless Time-triggered systems and jointly study the problem of optimizing the pruning ratio and bandwidth allocation to minimize training loss under communication latency constraints. To solve this joint optimization problem, we perform a convergence analysis on the gradient l2-norm of the asynchronous multi-tier federated learning (FL) model with adaptive model pruning. The convergence upper bound is derived and a joint optimization problem of pruning ratio and wireless bandwidth is defined to minimize the model training loss under a given communication latency constraint. The closed-form solutions for wireless bandwidth and pruning ratio by using KKT conditions are then formulated. As indicated in the simulation experiments, our proposed TT-Prune demonstrates a 40% reduction in communication cost, compared with the asynchronous multi-tier FL without model pruning, while maintaining the model convergence at the same level.

Multi-Access Edge Computing (MEC) and Open Radio Access Network (O-RAN) have emerged as promising paradigms to address diverse applications and service require-ments. MEC brings computational power to the network edge, enabling efficient offloading of compute-intensive tasks, while O-RAN facilitates network disaggregation for centralized deployment. The inherent conflict between MEC's user-centric proximity and O-RAN's disaggregation goals poses challenges in their integration. In this paper, we introduce a comprehensive analytical framework to optimize the integration of RAN dis-aggregation and MEC services. This framework models RAN functions and considers both full and partial task offloading strategies. This collaborative approach aims to minimize costs related to bandwidth usage and computing resource utilization. The simulation results indicate that the integration of MEC can significantly increase network costs. However, this impact can be mitigated by enhancing the centralization of computing resources. Additionally, doubling the weight of the link cost leads to a significant increase in the number of full compared to partial offloading, highlighting the tradeoff between these two strategies regarding computing resources and bandwidth utilization.

The use of multipath transport protocols in explored in Beyond 5G services given the various capabilities it could add to the service delivery. One of such is the possibility of exploiting multiple access connectivity, i.e. using multiple interface of the device concurrently, as well as dynamic adaptation of reliability at the transport layer. In this work, we present traffic classification that will then allow encrypted data to be classified for different level of reliability, hence enabling the delivery of dynamic reliability at the multipath transport protocol. The results show a significant improvement of up to 45 % in data overhead and of up to 20% reduction in end-to-end application-level latency, in comparison to the standard MP-QUIC implementation and intelligent state of the art reliability policies.

While the concept of AI4Net has been widely discussed in the past decade and adopted in 5G, its counterpart, Net4AI, has not gained that much attention so far. This is mostly due to the absence of solutions for the network to support AI applications beyond providing the communication infrastructure. In-Network Computing (INC) is a promising paradigm, potentially being integrated into 6G, which opens new solutions for realizing Net4AI. This paper focuses on the specific case of INC-assisted Split-AI. A Neural Network (NN) is split vertically and the executions of some layers of the split NN are offloaded to the entities of an Intelligent 6G User Plane. With the example of INC-assisted Split-AI, we elaborate on the challenges of Net4AI and discuss key requirements for the 6G architecture in terms of novel capabilities and information exchange.

Although edge intelligence (EI) propels the development of Internet of Things (IoT) applications to a new stage, the distributed nature of the end-users in EI networks greatly hinders its practical deployment. First, the resources of distributed end devices are limited, including computing and transmission resources, while the intelligent model typically necessitates intensive computation and substantial data from the network end. Secondly, the resources of end devices also exhibit heterogeneity, further complicating the learning in EI. Specifically, each device varies in computational capabilities, making it challenging to synchronise updates in collaborative learning approaches. Additionally, owing to the dispersed locations, each device encounters diverse wireless conditions, impeding effective communication with the edge server. Therefore, addressing the communication and computation constraints is necessary to foster practical EI applications. While novel distributed learning (DL) algorithms and machine learning (ML)-related techniques exhibit great potential, related review work lacks. Motivated by the literature gap, we provide a comprehensive review of the latest research endeavours on facilitating efficient EI deployment via examining novel DL algorithms and ML-related techniques. We also demonstrate the interplay of computation and communication efficiency in the resource-constrained EI landscape.

Although federated learning has achieved many breakthroughs recently, the heterogeneous nature of the learning environment greatly limits its performance and hinders its real-world applications. The heterogeneous data, time-varying wireless conditions and computing-limited devices are three main challenges, which often result in an unstable training process and degraded accuracy. Herein, we propose strategies to address these challenges. Targeting the heterogeneous data distribution, we propose a novel adaptive mixing aggregation (AMA) scheme that mixes the model updates from previous rounds with current rounds to avoid large model shifts and thus, maintain training stability. We further propose a novel staleness-based weighting scheme for the asynchronous model updates caused by the dynamic wireless environment. Lastly, we propose a novel CPU-friendly computation-reduction scheme based on transfer learning by sharing the feature extractor (FES) and letting the computing-limited devices update only the classifier. The simulation results show that the proposed framework outperforms existing state-of-the-art solutions and increases the test accuracy, and training stability by up to 2.38%, 93.10% respectively. Additionally, the proposed framework can tolerate communication delay of up to 15 rounds under a moderate delay environment without significant accuracy degradation.

Driven by the development of programmable networking hardware, In-network Computing (INC) has gained a considerable amount of attention in recent years. However, INC has so far barely been studied in the context of mobile networks, despite the vast advantages shown for fixed networks, such as latency or traffic reduction. Motivated by an Augmented Reality (AR) use-case, our work envisions an INC-enabled Intelligent User Plane (IUP) for 6G networks, which allows offloading computational tasks to UP entities having enhanced computational capabilities. The 6G IUP thus helps to keep mobile end-devices lighter and supports meeting the stringent delay requirements of novel applications, such as AR. Besides elaborating on the involved prospects and challenges, we identify key enablers for realizing the INC-enabled IUP. We show that embedding INC into the 6G system entails major changes in the architecture, as compared to the current 5G design.

In this paper, we propose an opportunistic scheme for the transmission of model updates from Federated Learning (FL) clients to the server, where clients are wireless mobile users. This proposal aims to opportunistically take advantage of the proximity of users to the base station or the general condition of the wireless transmission channel, rather than traditional synchronous transmission. In this scheme, during the training, intermediate model parameters are uploaded to the server, opportunistically and based on the wireless channel condition. Then, the proactively-transmitted model updates are used for the global aggregation if the final local model updates are delayed. We apply this novel model transmission scheme to one of our previous work, which is a hybrid split and federated learning (HSFL) framework for UAVs. Simulation results confirm the superiority of using proactive transmission over the conventional asynchronous aggregation scheme for the staled model by obtaining higher accuracy and more stable training performance. Test accuracy increases by up to 13.47% with just one round of extra transmission.

This article proposes a novel mathematical model for teleoperation over communication networks. For teleoperation over a communication network, a high packet rate can result in inefficient data transmission and cross-traffic problems, leading to extra delay and jitter. This article proposes an long short-term memory (LSTM)-based mathematical model which focuses on kinesthetic data reduction without loss of transparency during the transmission process through joint training combined with haptic data and perceptual deadband. Since the LSTM network can deal with a time series of haptic data, we further test the system performance through practically collected data. We investigate the packet rate and perceptual transparency of the proposed mathematical model by comparing with the conventional deadband. Additionally, we compare the proposed mathematical model with the perceptual deadband-based codecs. Simulation results show that the proposed solution further reduces the packet rate when dealing with haptic data without noticeable distortion. Also, comparing with the current just noticeable difference perceptual threshold, the proposed mathematical model helps improve the practicality of the bilateral teleoperation system without losing transparency.

Cellular-connected unmanned aerial vehicle (UAV) with flexible deployment is foreseen to be a major part of the sixth generation (6G) networks. The UAVs connected to the base station (BS), as aerial users (UEs), could exploit machine learning (ML) algorithms to provide a wide range of advanced applications, like object detection and video tracking. Conventionally, the ML model training is performed at the BS, known as centralized learning (CL), which causes high communication overhead due to the transmission of large datasets, and potential concerns about UE privacy. To address this, distributed learning algorithms, including federated learning (FL) and split learning (SL), were proposed to train the ML models in a distributed manner via only sharing model parameters. FL requires higher computational resource on the UE side than SL, while SL has larger communication overhead when the local dataset is large. To effectively train an ML model considering the diversity of UEs with different computational capabilities and channel conditions, we first propose a novel distributed learning architecture, a hybrid split and federated learning (HSFL) algorithm by reaping the parallel model training mechanism of FL and the model splitting structure of SL. We then provide its convergence analysis under non-independent and identically distributed (non-IID) data with random UE selection scheme. By conducting experiments on training two ML models, Net and AlexNet, in wireless UAV networks, our results demonstrate that the HSFL algorithm achieves higher learning accuracy than FL and less communication overhead than SL under IID and non-IID data, and the learning accuracy of HSFL algorithm increases with the increasing number of the split training UEs. We further propose a Multi-Arm Bandit (MAB) based best channel (BC) and best 2-norm (BN2) (MAB-BC-BN2) UE selection scheme to select the UEs with better wireless channel quality and larger local model updates for model training in each round. Numerical results demonstrate it achieves higher learning accuracy than BC, MAB-BC and MAB-BN2 UE selection scheme under non-IID, Dirichlet-nonIID and Dirichlet-Imbalanced data.

One of the most challenging services fifth-generation (5G) mobile network is designed to support, is the critical services in-need of very low latency, and/or high reliability. It is now clear that such critical services will also be at the core of beyond 5G (B5G) networks. While 5G radio design accommodates such supports by introducing more flexibility in timing, how efficiently those services could be scheduled over a shared network with other broadband services remains as a challenge. In this paper, we use network slicing as an enabler for network sharing and propose an optimization framework to schedule resources to critical services via puncturing technique with minimal impact on the regular broadband services. We then thoroughly examine the performance of the framework in terms of throughput and reliability through simulation.

The ever-growing use of unmanned aerial vehicles (UAVs) as aerial users is becoming a major part of the sixth generation (6G) networks, which could provide various applications, like object detection and video surveillance, by exploiting machine learning (ML) algorithms. However, the training of conventional centralized ML algorithms causes high communication overhead due to the transmission of large datasets and may reveal user privacy. Hence, distributed learning algorithms, including federated learning (FL) and split learning (SL), are proposed to train ML models in a distributed manner via sharing model parameters rather than raw data. Due to the different learning structures, they have different communication and learning efficiency. We propose a new distributed learning architecture, namely hybrid split and federated learning (HSFL), by adopting the parallel model training mechanism of FL and the network splitting structure of SL. Through the simulations in wireless UAV networks, the HSFL algorithm is demonstrated to have higher learning accuracy than FL and less communication overhead than SL under non-IID data. We further propose a Multi-Arm Bandit (MAB) based best channel (BC) and best 2-norm (BN2) (MAB-BC-BN2) UE selection scheme to select the UEs with better channel quality and larger local model updates in each round. Numerical results demonstrate it achieves higher learning accuracy than the benchmark schemes, BC, MAB-BC, and MAB-BN2 UE selection schemes.

The emergence of Network Functions Virtualization (NFV) is being heralded as an enabler of the recent technologies such as 5G/6G, IoT and heterogeneous networks. Existing NFV monitoring frameworks either do not have the capabilities to express the range of telemetry items needed to perform management or do not scale to large traffic volumes and rates. We present IntOpt, a scalable and expressive telemetry system designed for flexible NFV monitoring using active probing and P4. IntOpt allows us to specify monitoring requirements for individual service chain, which are mapped to telemetry item collection jobs that fetch the required telemetry items from P4 programmable data-plane elements. We propose mixed integer linear program (MILP) as well as a simulated annealing based random greedy (SARG) meta-heuristic approach to minimize the overhead due to active probing and collection of telemetry items. Using P4-FPGA, we benchmark the overhead for telemetry collection. Our numerical evaluation shows that the proposed approach can reduce monitoring overheads by 39% and monitoring delays by 57%. Such optimization may as well enable existing expressive monitoring frameworks to scale for larger real-time networks.

With evolution of network function virtualization (NFV), network services can be provided as service function chains (SCs), each consisting of multiple virtual network functions (VNFs). The deployment of SCs including placement of VNF instances and virtual links connecting these functions, onto the substrate physical network is a critical issue which significantly affects the performance of the offered network services. Due to the unpredictable traffic and network state variations, as well as diverse quality of service (QoS) requirements, an online SCs deployment approach is needed to cope with different service requests and real-time network traffics. In this paper, we employ edge intelligence using a distributed deep reinforcement learning approach to deploy SCs in order to jointly balance the load on the physical nodes and links in the edge environments. The evaluation results show that the proposed approach outperforms state-of-the-art algorithms in terms of minimizing the drop rate of the incoming service chain requests. In addition, the proposed approach is able to rapidly deploy service flows even in the large real-world network typologies.

5G is the fifth generation of cellular networks. It enables billions of connected devices to gather and share information in real time; a key facilitator in Industrial Internet of Things (IoT) applications. It has more capabilities in terms of bandwidth, latency/delay, processing powers and flexibility to utilize either edge or cloud resources. Furthermore, 6G is expected to be equipped with the new capability to converge ubiquitous communication, computation, sensing and controlling for a variety of sectors, which heightens the complexity in a more heterogeneous environment This increased complexity, combined with energy efficiency and Service Level Agreement (SLA) requirements makes application of Machine Learning (ML) and distributed ML necessary. A decentralized approach stemming from distributed learning is a very attractive option compared with a centralized architecture for model learning and inference. Distributed ML exploits recent Artificial Intelligence (AI) technology advancements to allow collaborated ML, whilst safeguarding private data, minimizing both communication and computation overhead along with addressing ultra-low latency requirements. In this paper, we review a number of distributed ML architectures and designs, that focus on optimizing communication, computation and resource distribution. Privacy, information security and compute frameworks, are also analyzed and compared with respect to different distributed ML approaches. We summarize the major contributions and trends in this area and highlight the potential of distributed ML to help researchers and practitioners make informed decisions on selecting the right ML approach for 5G and Beyond related AI applications. To enable distributed ML for 5G and Beyond, communication, security, and computing platform often counter balance each other, thus, consideration and optimization of these aspects at an overall system level is crucial to realize the full potential of AI for 5G and Beyond. These different aspects do not only pertain to 5G, but will also enable careful design of distributed machine learning architectures to circumvent the same hurdles that will inevitably burden 5G and Beyond network generations. This is the first survey paper that brings together all these aspects for distributed ML.

This article presents a data-driven framework for trajectory recommendation in automated and cooperative driving. The considered cooperative driving maneuver is lane-merge coordination, and while the trajectory recommendation can only be communicated to the connected vehicles, in computation of those recommendations both connected and unconnected vehicles are taken into account. The data-driven framework is implemented centrally, comprising of two main components of a traffic orchestrator (TO) and data fusion (DF). The TO predicts the safest trajectories for connected vehicles involved in the lane-merge maneuver. The DF incorporates camera detected vehicles in order to map all vehicles, including connected and unconnected. To this end, the recommendations are built using various state-of-the-art machine learning (ML) techniques, including deep reinforcement learning and dueling deep Q-network. Our evaluations are conducted using the real-system deployed in the test track, with a mix of connected and unconnected vehicles. The results demonstrate the precision of predicted trajectories, and the percentage of successful lane merge achieved deploying different ML techniques.

In this paper, we study the role that machine learning can play in cooperative driving. Given the increasing rate of connectivity in modern vehicles, and road infrastructure, cooperative driving is a promising first step in automated driving. The example scenario we explored in this paper, is coordinated lane merge, with data collection, test and evaluation all conducted in an automotive test track. The assumption is that vehicles are a mix of those equipped with communication units on board, i.e. connected vehicles, and those that are not connected. However, roadside cameras are connected and can capture all vehicles including those without connectivity. We develop a Traffic Orchestrator that suggests trajectories based on these two sources of information, i.e. connected vehicles, and connected roadside cameras. Recommended trajectories are built, which are then communicated back to the connected vehicles. We explore the use of different machine learning techniques in accurately and timely prediction of trajectories.

In this paper, a framework for lane merge coordination is presented utilising a centralised system, for connected vehicles. The delivery of trajectory recommendations to the connected vehicles on the road is based on a Traffic Orchestrator and a Data Fusion as the main components. Deep Reinforcement Learning and data analysis is used to predict trajectory recommendations for connected vehicles, taking into account unconnected vehicles for those suggestions. The results highlight the adaptability of the Traffic Orchestrator, when employing Dueling Deep Q-Network in an unseen real world merging scenario. A performance comparison of different reinforcement learning models and evaluation against Key Performance Indicator (KPI) are also presented.

The emerging Cloud-RAN architecture within the fifth generation (5G) of wireless networks plays a vital role in enabling higher flexibility and granularity. On the other hand, Cloud-RAN architecture introduces an additional link between the central, cloudified unit and the distributed radio unit, namely fronthaul (FH). Therefore, the foreseen reliability and latency for 5G services should also be provisioned over the FH link. In this paper, focusing on Ethernet as FH, we present a reliable packet-based FH communication and demonstrate the upper and lower bounds of latency that can be offered. These bounds yield insights into the trade-off between reliability and latency, and enable the architecture design through choice of splitting point, focusing on high layer split between PDCP and RLC and low layer split between MAC and PHY, under different FH bandwidth and traffic properties. Presented model is then analyzed both numerically and through simulation, with two classes of 5G services that are ultra reliable low latency (URLL) and enhanced mobile broadband (eMBB).

The Tactile Internet will enable users to physically explore remote environments and to make their skills available across distances. An important technological aspect in this context is the acquisition, compression, transmission, and display of haptic information. In this paper, we present the fundamentals and state of the art in haptic codec design for the Tactile Internet. The discussion covers both kinesthetic data reduction and tactile signal compression approaches. We put a special focus on how limitations of the human haptic perception system can be exploited for efficient perceptual coding of kinesthetic and tactile information. Further aspects addressed in this paper are the multiplexing of audio and video with haptic information and the quality evaluation of haptic communication solutions. Finally, we describe the current status of the ongoing IEEE standardization activity P1918.1.1 which has the ambition to standardize the first set of codecs for kinesthetic and tactile information exchange across communication networks.

Within the ongoing activities devoted to the definition of 5G networks, massive Internet of Things (mIoT) is regarded as a compelling use case, both for its relevance from business perspective, and for the technical challenges it poses to network design. With their envisaged massive deployment of devices requiring sporadic connectivity and small data transmission, yet QoS constrained, mIoT services will require adhoc end- to-end (E2E) solutions, i.e., featuring access and core network enhanced Control and User planes (CP/UP) mechanisms. This paper presents and evaluates a novel connectivity solution to manage massive number of devices. The paper presents an analytical model developed to evaluate the performance of the proposed solution. Quantitative results derived from the model demonstrate the effectiveness of the solution proposed in this paper, compared to 4G systems, and its ability to reduce CP signaling and optimize UP resource utilization for massive device deployment.

This paper studies the application of non-orthogonal multiple access (NOMA) to cache-enabled cloud radio access network (CRAN) with mixed multicast and unicast transmission. Users requesting the same content are grouped together and served with a cluster of remote radio heads (RRHs) using distributed beamforming. In addition, the user with better channel condition in each group is allowed to request an extra unicast content via the NOMA protocol. Each RRH has a local cache which enables it to acquire the requested contents either from the local cache or from the central processor via the fronthaul link. Taking the maximum fronthaul capacity into consideration, we investigate the subchannel (SC) allocation problem to both RRHs and multicast groups to improve the weighted network sum rate. The optimal solution requires exhaustive search, which become prohibitively complicated as the number of RRHs and groups increases. To tackle this problem effectively, we formulate this problem as a three-sided matching problem among SCs, RRHs and multicast groups, and propose a novel low-complexity matching algorithm. We prove mathematically that the proposed algorithm converges to a stable matching within limited number of iterations. Numerical results unveil that the proposed algorithm closely approaches the optimal solution and outperforms the conventional orthogonal multiple access (OMA)-based CRAN.

So far, research on Smart Cities and self-organizing networking techniques for fifth-generation (5G) cellular systems has been one-sided: a Smart City relies on 5G to support massive machine-to-machine (M2M) communications, but the actual network is unaware of the information flowing through it. However, a greater synergy between the two would make the relationship mutual, since the insights provided by the massive amount of data gathered by sensors can be exploited to improve the communication performance. In this paper, we concentrate on self-organization techniques to improve handover efficiency using vehicular traffic data gathered in London. Our algorithms exploit mobility patterns between cell coverage areas and road traffic congestion levels to optimize the handover bias in heterogeneous networks and dynamically manage mobility management entity (MME) loads to reduce handover completion times.

Narrowband-Internet of Things (NB-IoT) has been released by 3GPP to provide extended coverage and low energy consumption for low-cost machine-type devices. Requiring only a reasonably low-cost hardware update to the already deployed long term evolution base stations and being compatible with current core network and enhanced core solutions that aim to reduce the battery consumption and minimize the signaling, NB-IoT deployments are quickly increasing, making NB-IoT a dominating technology for low-power wide area networks. To this aim, in this paper, we focus on group communications (i.e., multicast) in NB-IoT to efficiently support the transmission of firmware, software, task updates, or commands toward a large set of devices. We discuss the architectural and procedural enhancements needed to support the unique features of group communications in machine-type environments, such as customer-driven group formation. We also extend the NBIoT frame to include a channel for multicast transmissions. Finally, we propose two transmission strategies for multicast content delivery and evaluate their performance considering the impact on the downlink background traffic and the channel occupancy.

With the emergence of Cloud-RAN as one of the dominant architectural solutions for the next-generation mobile networks, the reliability and latency on the fronthaul (FH) segment become critical performance metrics for applications such as the Tactile Internet. Ensuring FH performance is further complicated by the switch from point-to-point dedicated FH links to packet-based multi-hop FH networks. This change is largely justified by the fact that packet-based fronthauling allows the deployment of FH networks on the existing Ethernet infrastructure. This paper proposes to improve the reliability and latency of packet-based fronthauling by means of multi-path diversity and erasure coding of the MAC frames transported by the FH network. Under a probabilistic model that assumes a single service, the average latency required to obtain reliable FH transport and the reliability-latency tradeoff is first investigated. The analytical results are then validated and complemented by a numerical study that accounts for the coexistence of the enhanced mobile broadband and ultra-reliable low-latency services in fifth-generation networks by comparing orthogonal and non-orthogonal sharing of FH resources.

In future wireless networks, network functions virtualization lays the foundations for establishing a new dynamic resource management framework to efficiently utilize network resources. In this paper, a network service can be viewed as a chain of virtual network functions (VNFs), called a service function chain (SFC), served via placement, admission control (AC), and embedding into network infrastructure, based on the resource management objectives and the state of network. To fully exploit such a potential and reach higher network performance, resource management stages should be jointly performed. To this end, two main challenges are: how to present a system model that formulates the desired resource allocation problem for different types of SFCs as well as different features, and how to tackle the computational complexity of the problem and solve it in a tractable manner. In this paper, we address these two issues and solve the joint problem of AC and SFC embedding. We introduce a comprehensive system model, and formulate the joint task as a mixed integer linear programming. This formulation encompasses splittable VNF and multi-path routing scenarios. We employ relaxation, reformulation, and successive convex approximation methods to solve the problem. Simulation results demonstrate that the proposed schemes outperform the earlier works.

The 5G mobile network is expected to meet the diverse demands from multiple types of business services. At the same time, some of the 5G use cases come with hard, and often expensive to meet, requirements in terms of latency and bandwidth. It is a common understanding that one system can not fit all and there is a need for customizing network according to the requirements of specific business use cases. Network slicing is introduced to partition the physical network to different slices to be configured for providing different quality of service as requested by the slice' operator and required by the slice' users. Since these slices will be used by the businesses, e.g. verticals, allocating physical resources to the network slices, is not anymore only a matter of performance but also a matter of revenue and business model. In this paper, we address a joint resource and revenue optimization a novel auction based model. Through extensive simulation study, we demonstrate our proposed auction model can allocate network resources to network slices for providing (i) higher satisfaction of requirements per network slice, and (ii) increased network revenue1.

With Network Function Virtualization (NFV), network functions are deployed as modular software components on the commodity hardware, and can be further chained to provide services, offering much greater flexibility and lower cost of the service deployment for the network operators. At the same time, replacing the network functions implemented in purpose built hardware with software modules poses a great challenge for the operator to maintain the same level of performance. The grade of service promised to the end users is formalized in the Service Level Agreement (SLA) that typically contains the QoS parameters, such as minimum guaranteed data rate, maximum end to end latency, port availability and packet loss. State of the art solutions can guarantee only data rate and latency requirements, while service availability, which is an important service differentiator is mostly neglected. This paper focuses on the placement of virtualized network functions, aiming to support service differentiation between the users, while minimizing the associated service deployment cost for the operator. Two QoS-aware placement strategies are presented, an optimal solution based on the Integer Linear Programming (ILP) problem formulation and an efficient heuristic to obtain near optimal solution. Considering a national core network case study, we show the cost overhead of availability-awareness, as well as the risk of SLA violation when availability constraint is neglected. We also compare the proposed function placement heuristic to the optimal solution in terms of cost efficiency and execution time, and demonstrate that it can provide a good estimation of the deployment cost in much shorter time.

Vehicular CrowdSensing (VCS) is an emerging solution designed to remotely collect data from smart vehicles. It enables a dynamic and large-scale phenomena monitoring just by exploring the variety of technologies which have been embedded in modern cars. However, VCS applications might generate a huge amount of data traffic between vehicles and the remote monitoring center, which tends to overload the LTE networks. In this paper, we describe and analyze a gEo-clUstering approaCh for Lte vehIcular crowDsEnsing dAta offloadiNg (EUCLIDEAN). It takes advantage of opportunistic vehicle-to-vehicle (V2V) communications to support the VCS data upload process, preserving, as much as possible, the cellular network resources. In general, it is shown from the presented results that our proposal is a feasible and an effective scheme to reduce up to 92.98% of the global demand for LTE transmissions while performing vehicle-based sensing tasks in urban areas. The most encouraging results were perceived mainly under high-density conditions (i.e., above 125 vehicles/km2), where our solution provides the best benefits in terms of cellular network data offloading.

Splitting functionalities of radio access network (RAN) and cloudification of such functionalities is considered as one of the key enablers of the next generation mobile and wireless networking, i.e. 5G, and is often referred to as software-defined RAN, virtualized RAN or Cloud RAN. Defining the splitting point, and maintaining the tight interaction between different functionalities in the RAN is, however, critical. Success of such cloudification depends on the availability of high speed fronthaul, while high speed fronthauling is costly. In this paper we experiment splitting MAC and PHY layer with fronthauling through Ethernet that allows using commodity and low-cost industry standard equipment. We examine the effect of packetization on latency, and study the pros and cons of splitting MAC and PHY layer, within a hardware-based testbed.

Mobile data offloading on smaller cells such as Wi-Fi comes as a natural solution to boost cellular networks capacity and keep up with the rapid increase of mobile data traffic demand. In this paper, we propose an offloading mechanism through the abstraction of Software-defined Networking (SDN) in the mobile backhaul to provide programmable offloading policy derivation that are aware of users and applications as well as the condition of wireless network. The proposed mechanism considers the real-time network condition to derive the offloading policies and efficiently accommodate the traffic in both LTE and Wi-Fi networks. Numerical results prove that the proposed approach can significantly improves dropping rate of the incoming traffic with using more real-time and dynamic decisions for offloading.

This paper's central aim is to address the issue of resource management at the Local Mobility Anchors (LMA) in the Proxy Mobile IPv6 (PMIPv6) networks. A class-based admission control is proposed to improve the bottleneck effect caused by triangular routing in PMIPv6, where resource units are rationed amongst different classes of traffic according to their QoS requirements. The PMIPv6 network is modeled as an M/M/m/m tandem queuing network with two types (classes) of arrival process and an analytical model is presented. Performance of our proposed admission control scheme is evaluated through simulation and results are compared to the case where no distinction in terms of resource unit allocation between classes of traffic was considered.

Abstract-In this paper, we propose a dynamic OFDMA based subcarrier/power allocation scheme, which aims to balance the aggregate rate and the achieved TCP throughput of competing TCP flows. The proposed allocation utilizes the theoretical TCP throughput which can be accomplished in the end-to-end path. In doing so, the TCP aware scheme attempts to minimize the gap between the allocated rate and the theoretical upper bound under the system constraints. Such a technique can be of significant importance since due to its popularity, TCP is commonly used for streaming video or other ultimedia applications. Numerical investigations reveal that the proposed approach, provides more balance towards the TCP throughput, and under some considerations significantly increase the fairness among competing TCP flows over end-to-end paths of different characteristics. In addition to that, it also manages to avoid starvation of TCP flows with poor channel conditions.

A plethora of modifications to transmission control protocol (TCP) have recently been proposed, a major aim being to improve its performance over wireless links. Two schools of thought have emerged: the first investigates changes to the transport-layer protocol, whereas the second explores the potential to enhance the characteristics of lower layers to improve the end-to-end performance of TCP. This study focuses on the latter, and, in contrast to most research in this area, which thus-far has concentrated on a single TCP flavour, examines the case where different TCP flavours are competing over a wireless link. To this end, the authors present and assess a cross-layer solution to adapt the coding rate at the link-layer based on the detected TCP flavour, to maximise fairness among TCP flows. Through both analysis and simulation, the authors show that the proposed scheme considerably improves the fairness among different TCP flavours that compete over a wireless link. Furthermore, the proposed approach has minimal detrimental effect on the aggregate throughput of TCP flows.

The IEEE 802.15.4 specification has generated a lot of interest in recent times, especially within the Wireless Sensor Network (WSN) research community, primarily because energy efficiency is one of the specifications' design cornerstone. As this specification is relatively new, it is incumbent that its operational mechanisms are well understood, to allow for efficient cross layer interactions between the different processes that make up the specification and existing or new higher layer protocols. In this paper, we present a deterministic Petri-Net model of the IEEE 802.15.4 CSMA-CA process, that is timer driven and operates within the bounds of the contention access period (CAP). Using this model, we are able to analyze the performance characteristics of the CSMA-CA process, especially in terms of channel throughput and energy consumption. We also verify the extracted system indices by comparing them to those gotten from a full model of the specification, implemented using the OPNET network simulation platform.

A significant body of recent research has analyzed the problematic behavior of TCP over wireless links, and a plethora of modifications to TCP have been proposed in order to increase its performance in such contexts. Two schools of thought have emerged: the first proposes changes to the end-to- end protocol, while the second explores the potential to enhance lower layers as a means to improve the end-to-end performance of TCP. This paper focuses on the latter, and in contrast to most research in this area, which thus-far has concentrated on a single TCP flavor, examines the case where different TCP flavors are competing over a wireless link. To this end, we present and assess a cross-layer solution that involves the adaptation of lower layer characteristics (i.e., the coding rate) based on the detected TCP flavor, in order to maximize the fairness among TCP flows. Through extensive numerical investigations, we show that the proposed scheme considerably improves the fairness over wireless links among different TCP flavors. Our approach also has a minimal effect on the aggregate throughput of the TCP flows, and in cases where the packet error rate is very low, has a small positive effect on throughput.

A range of flavors of TCP are already in existence, and further flavors are being introduced in order to, for example, cope with the packet loss characteristics of wireless links. Moreover, the proliferation new wireless standards and the relative performance differences among them have been mushrooming in recent years. Given the increasingly heterogeneous nature of the Internet, mechanisms do not usually exist for a server to specifically select an appropriate TCP flavor for each individual download. In this paper, we therefore present and assess a cross-layer solution for a node (e.g. a base-station) to quickly adapt lower-layer characteristics (the coding rate and local ARQ retransmissions threshold) based on the detected TCP flavor, in order to optimize the end-to-end performance of the download for that utilized flavor. We demonstrate that the proposed scheme has considerable potential to improve the overall download throughput, while placing no burden on the server and requiring no changes to existing TCP implementations.

In this paper, we discuss the development of a Software Download Management (SDM) module, envisioned to handle the implications and associated operations of downloads for mobile terminal reconfiguration. We highlight the SDM operations in the pre-, during-, and post-download phases, and some emphasis is also placed on the transport-layer decision making capabilities of the SDM. An important emerging capability is mass-reconfiguration, involving the alteration of common functionalities in many terminals concurrently. Relevant scenarios here might include the adaptation of a protocol or codec, an upgrade of an operating system capability, or indeed a change in network configuration which requires a large number of terminals to be adapted accordingly. A particular emphasis is also therefore on one-to-many download functionalities.

Transmission control protocol (TCP), the almost universally used reliable transport protocol in the Internet, has been engineered to perform well in wired networks where packet loss is mainly due to congestion. TCP throughput, however, degrades over wireless links, which are characterized by a high and greatly varying bit error rate and by intermittent connectivity. Over such wireless links, the performance achieved by TCP can be improved through the use of cross-layer algorithms at the link-level, which interact with the TCP state machine. In this paper, a TCP-aware dynamic ARQ algorithm is therefore proposed, which utilizes TCP timing information to prioritize ARQ packet retransmissions. Numerical investigation of the proposed algorithm demonstrates the performance improvements that can be attained through this approach, in comparison with TCP-agnostic link-layer approaches.