The satellite-aerial-terrestrial network (SATN) is a promising architecture to achieve seamless global coverage and meet diverse quality-of-service (QoS) requirements in next-generation wireless communications. To support such multi-layered connectivity, we consider a mixed radio frequency (RF) and free-space optical (FSO) architecture, where heterogeneous users access a high-altitude platform (HAP) via RF links, and the HAP, acting as an aerial relay, forwards the aggregated traffic to a satellite through an FSO backhaul. Existing transmission schemes for such mixed RF-FSO SATNs, however, are not well suited to providing differentiated statistical QoS guarantees for heterogeneous users. To address this limitation, we propose a mixed RF-FSO QoS-aware uplink transmission (MRQ-UT) scheme. Specifically, we impose statistical delay-QoS constraints at both the user and HAP buffers, thereby explicitly capturing heterogeneous constraints on queueing delay and buffer overflow. On this basis, we derive the system effective capacity using a two-stage tandem queue model, which captures the sequential queuing behavior over the RF access and FSO backhaul links. Building upon this model, we develop a tractable effective-capacity-based optimization framework and propose a statistical channel-aware joint power and beamforming algorithm that enhances QoS provisioning under imperfect channel state information. Simulation results demonstrate that the proposed MRQ-UT scheme significantly outperforms benchmark schemes in terms of effective capacity and statistical QoS performance.

Federated learning (FL) has been recognized as a viable solution for local-privacy-aware collaborative model training in wireless edge networks, but its practical deployment is hindered by the high communication overhead caused by frequent and costly server-device synchronization. Notably, most existing communication-efficient FL algorithms fail to reduce the significant inter-device variance resulting from the prevalent issue of device heterogeneity. This variance severely decelerates algorithm convergence, increasing communication overhead and making it more challenging to achieve a well-performed model. In this paper, we propose a novel communication-efficient FL algorithm, named FedQVR, which relies on a sophisticated variance-reduced scheme to achieve heterogeneity-robustness in the presence of quantized transmission and heterogeneous local updates among active edge devices. Comprehensive theoretical analysis justifies that FedQVR is inherently resilient to device heterogeneity and has a comparable convergence rate even with a small number of quantization bits, yielding significant communication savings. Besides, considering non-ideal wireless channels, we propose FedQVR-E which enhances the convergence of FedQVR by performing joint allocation of bandwidth and quantization bits across devices under constrained transmission delays. Extensive experimental results are also presented to demonstrate the superior performance of the proposed algorithms over their counterparts in terms of both communication efficiency and application performance.

Time-sensitive networking (TSN) is proposed to satisfy the increasingly stringent demands of Industrial 4.0 for deterministic transmission. This is achieved by generating a series of centrally configured gate control lists to strictly restrict the forwarding time of arriving flows. However, such a centralized scheme requires prior information of all flows, severely impeding TSN from providing an online response to dynamic industrial applications. To solve this problem, we innovatively propose to use admission control (AC) to realize deterministic transmission in the virtualized TSN network. In this approach, AC is distributively executed on each node and link, whereby flows of applications are served by passing through a series of virtual network functions (VNFs). This distributed AC execution is regarded as a VNF embedding (VNE) process. Specifically, we propose a two-stage online framework, Smart Admission Control (SmartAC), to cater to dynamic applications. The first stage, referred to as thestatic stage, obtains a deterministic VNE solution by synthesizing AC decisions of individual TSN nodes and links. The second stage, referred to as thedynamic stage, fine-tunes VNE solutions obtained from thestatic stageto adapt to the harsh environment with insufficient resources or limited VNF migration budgets. Simulation results demonstrate the effectiveness of SmartAC in improving response rate and resource utilization ratio. Notably, SmartAC reduces runtime by 90% compared to existing algorithms and exhibits robustness across different network topologies.

Satellites equipped with computing capabilities serve as invaluable access platforms for 5G and beyond (NextG) non-terrestrial networks (NTNs). They facilitate the continuous execution of resource-intensive edge-assisted deep learning (DL) tasks that are offloaded from Internet-of-Things (IoT) user equipment (UEs) in remote areas. To this end, satellite access network (SAN) resources need to be carefully “sliced”, consid-ering both the constrained energy availability and the scarcity of SAN resources. Existing SAN slicing approaches tend to treat offloaded tasks conventionally, overlooking the intricate semantics associated with DL tasks. In this paper, we propose semantic SAN (SemSAN), the first semantic SAN slicing algorithm for NextG AI-native NTNs. Our keen observations reveal that various DL tasks (i) can tolerate different degrees of image compression, and (ii) may yield equivalent model accuracy when employing DNN models with different sizes. These observations inspire us to further exploit the computation capability of a SAN to support more tasks while still minimizing overall energy consumption. After analyzing the characteristics of this optimization problem, we propose an online greedy SemSAN slicing algorithm to approximate its optimal solution. Extensive experiments verify the effectiveness of SemSAN in energy saving and its ability to support a substantial number of tasks, compared with other baselines.

Collaborative Edge Computing (CEC) is a new edge computing paradigm that enables neighboring edge servers to share computational resources with each other. Although CEC can enhance the utilization of computational resources, it still suffers from resource waste. The primary reason is that end-users from the same area are likely to offload similar tasks to edge servers, thereby leading to duplicate computations. To improve system efficiency, the computation results of previously executed tasks can be cached and then reused by subsequent tasks. However, most existing computation reuse algorithms only consider one edge server, which significantly limits the effectiveness of computation reuse. To address this issue, this paper applies computation reuse in CEC networks to exploit the collaboration among edge servers. We formulate an optimization problem that aims to minimize the overall task response time and decompose it into a caching subproblem and a scheduling subproblem. By analyzing the properties of optimal solutions, we show that the optimal caching decisions can be efficiently searched using the bisection method. For the scheduling subproblem, we utilize projected gradient descent and backtracking to find a local minimum. Numerical results show that our algorithm significantly reduces the response time in various situations.

With the rapid development of Mobile Edge Computing (MEC), various real-time applications have been deployed to benefit people's daily lives. The performance of these applications relies heavily on the freshness of collected environmental information, which can be quantified by its Age of Information (AoI). In the traditional definition of AoI, it is assumed that the status information can be actively sampled and directly used. However, for many MEC-enabled applications, the desired status information is updated in an event-driven manner and necessitates data processing. To better serve these applications, we propose a new definition of AoI and, based on the redefined AoI, we formulate an online AoI minimization problem for MEC systems. Notably, the problem can be interpreted as a Markov Decision Process (MDP), thus enabling its solution through Reinforcement Learning (RL) algorithms. Nevertheless, the traditional RL algorithms are designed for MDPs with completely unknown system dynamics and hence usually suffer long convergence times. To accelerate the learning process, we introduce Post-Decision States (PDSs) to exploit the partial knowledge of the system's dynamics. We also combine PDSs with deep RL to further improve the algorithm's applicability, scalability, and robustness. Numerical results demonstrate that our algorithm outperforms the benchmarks under various scenarios.

Energy saving plays an important role in designing AI-native 6G networks. Radio Access Network (RAN) slicing is a fundamental tool to save energy through resource multiplexing. However, as the AI services required by users become more heterogenous than ever in 6G network, service-oriented RAN slicing naturally consumes a lot of energy, leading to a tradeoff between QoS guarantees and energy saving for the network scheduler to decide. In this paper, we propose sustainable service-oriented (SSO) RAN slicing scheduler for 6G networks to jointly optimize workload distribution and resource allocation. The target is to minimize the long-term average energy consumption using the meta reinforcement learning (MRL) method. To be specific, each type of services is treated as an independent optimization problem, where the workload distribution is solved by convex optimization and the resource allocation is solve by Q-learning policy. Numerical results show that SSO effectively reduces the system energy consumption while satifying QoS requirements, as compared with benchmarks.

Federated learning (FL) can be used in mobile-edge networks to train machine learning models in a distributed manner. Recently, FL has been interpreted within a model-agnostic meta-learning (MAML) framework, which brings FL significant advantages in fast adaptation and convergence over heterogeneous data sets. However, existing research simply combines MAML and FL without explicitly addressing how much benefit MAML brings to FL and how to maximize such benefit over mobile-edge networks. In this article, we quantify the benefit from two aspects: 1) optimizing FL hyperparameters (i.e., sampled data size and the number of communication rounds) and 2) resource allocation (i.e., transmit power) in mobile-edge networks. Specifically, we formulate the MAML-based FL design as an overall learning time minimization problem, under the constraints of model accuracy and energy consumption. Facilitated by the convergence analysis of MAML-based FL, we decompose the formulated problem and then solve it using analytical solutions and the coordinate descent method. With the obtained FL hyperparameters and resource allocation, we design an MAML-based FL algorithm, called automated FL (AutoFL), that is able to conduct fast adaptation and convergence. Extensive experimental results verify that AutoFL outperforms other benchmark algorithms regarding the learning time and convergence performance.

Various middleboxes are ubiquitously deployed in networks to perform packet processing functions, such as firewalling, proxy, scheduling, etc., for the flows passing through them. With the explosion of network traffic and the demand for multiple types of network resources, it has never been more challenging on a middlebox to provide Quality-of-Service (QoS) guarantees to grouped flows. Unfortunately, all currently existing fair queueing algorithms fail in supporting hierarchical scheduling, which is necessary to provide QoS guarantee to the grouped flows of multiple service classes. In this paper, we present two new multi-resource fair queueing algorithms to support hierarchical scheduling, collapsed Hierarchical Dominant Resource Fair Queueing (collapsed H-DRFQ) and dove-tailing H-DRFQ. Particularly, collapsed H-DRFQ transforms the hierarchy of grouped flows into a flat structure for flat scheduling while dove-tailing H-DRFQ iteratively performs flat scheduling to sibling nodes on the original hierarchy. Through rigorous theoretical analysis, we find that both algorithms can provide hierarchical share guarantees to individual flows, while the upper bound of packet delay in dove-tailing H-DRFQ is smaller than that of collapsed H-DRFQ. We implement the proposed algorithms on Click modular router and the experimental results verify our analytical results.

Personalized Federated Learning (PFL) is a new Federated Learning (FL) approach to address the heterogeneity issue of the datasets generated by distributed user equipments (UEs). However, most existing PFL implementations rely on synchronous training to ensure good convergence performances, which may lead to a serious straggler problem, where the training time is heavily prolonged by the slowest UE. To address this issue, we propose a semi-synchronous PFL algorithm, termed as Semi-Synchronous Personalized FederatedAveraging (PerFedS2), over mobile edge networks. By jointly optimizing the wireless bandwidth allocation and UE scheduling policy, it not only mitigates the straggler problem but also provides convergent training loss guarantees. We derive an upper bound of the convergence rate of PerFedS2 in terms of the number of participants per global round and the number of rounds. On this basis, the bandwidth allocation problem can be solved using analytical solutions and the UE scheduling policy can be obtained by a greedy algorithm. Experimental results verify the effectiveness of PerFedS2 in saving the training time as well as guaranteeing the convergence of training loss, in contrast to synchronous and asynchronous PFL algorithms.

Personalized Federated Learning (PFL) is a new Federated Learning (FL) paradigm, particularly tackling the heterogeneity issues brought by various mobile user equipments (UEs) in mobile edge computing (MEC) networks. However, due to the ever-increasing number of UEs and the complicated administrative work it brings, it is desirable to switch the PFL algorithm from its conventional two-layer framework to a multiple-layer one. In this paper, we propose hierarchical PFL (HPFL), an algorithm for deploying PFL over massive MEC networks. The UEs in HPFL are divided into multiple clusters, and the UEs in each cluster forward their local updates to the edge server (ES) synchronously for edge model aggregation, while the ESs forward their edge models to the cloud server semi-asynchronously for global model aggregation. The above training manner leads to a tradeoff between the training loss in each round and the round latency. HPFL combines the objectives of training loss minimization and round latency minimization while jointly determining the optimal bandwidth allocation as well as the ES scheduling policy in the hierarchical learning framework. Extensive experiments verify that HPFL not only guarantees convergence in hierarchical aggregation frameworks but also has advantages in round training loss maximization and round latency minimization.

This paper investigates the problem of providing ultra-reliable and power-efficient virtual reality (VR) experiences for wireless mobile users. To ensure reliable ultra-high-definition (UHD) video frame delivery to mobile users and enhance their immersive visual experiences, a coordinated multipoint (CoMP) transmission technique and millimeter wave (mmWave) communications are exploited. Owing to user movement and time-varying wireless channels, the wireless VR experience enhancement problem is formulated as a sequence-dependent and mixed-integer problem with a goal of maximizing users’ feeling of presence (FoP) in the virtual world, subject to power consumption constraints on access points (APs) and users’ head-mounted displays (HMDs). The problem, however, is hard to be directly solved due to the lack of users’ accurate tracking information and the sequence-dependent and mixed-integer characteristics. To overcome this challenge, we develop a parallel echo state network (ESN) learning method to predict users’ tracking information by training fresh and historical tracking samples separately collected by APs. With the learnt results, we propose a deep reinforcement learning (DRL) based optimization algorithm to solve the formulated problem. In this algorithm, we implement deep neural networks (DNNs) as a scalable solution to produce integer decision variables and solve a continuous power control problem to criticize the integer decision variables. Finally, the performance of the proposed algorithm is compared with various benchmark algorithms, and the impact of different design parameters is also discussed. Simulation results demonstrate that the proposed algorithm is more 4.14% power-efficient than the benchmark algorithms.

The Virtual Network Function (VNF) deployment problem in Network Function Virtualization (NFV) is of broad theoretical and practical interests. As the traffic surges in data centers, VNF deployment requires load balancing across the servers to avoid possible congestions caused by the uneven distribution of VNFs. In addition, VNFs always specify placement constraints, restricting them to run only on part of the servers. Therefore, we study the load balancing problem for VNF deployment with placement constraints. Despite the rich bodies of recent work on the constrained load balancing problem using the network flow algorithms, they all suffer exponential complexities, leading to unbearable running time in practical executions. In this paper, we propose a new load balancing policy termed Constrained Min-max Placement (CMMP) that schedules VNFs in a way similar to the max-min allocation, where we try to assign the most possible VNFs to the poorest loaded server. The online scheduler for CMMP has a logarithmic time complexity and is simple enough to implement in practice. Trace-driven simulations show that the online CMMP speeds up at least two orders of magnitude of running time comparing to other network flow algorithms.

As the volume of traffic flows surges, providing Quality-of-Service (QoS) guarantees to flows by fair queueing has never been more challenging in Network Function Virtualization (NFV). There has been a recent effort in both industry and academia to develop fair queueing algorithms across multiple resources in NFV. However, all existing works fail to support hierarchical scheduling, a crucial feature that also provides QoS guarantees to grouped flows on tenant boundaries. In this paper, we present two new multi-resource fair queueing algorithms that support hierarchies, collapsed Hierarchical Dominant Resource Fair Queueing (collapsed H-DRFQ) and dove-tailing H-DRFQ, both of which provide hierarchical share guarantees. Through formal analysis, we find that the dove-tailing H-DRFQ outper-forms collapsed H-DRFQ by providing a smaller delay bound. However, according to the simulation results, both algorithms have their pros and cons. Dove-tailing H-DRFQ benefits to the flows with more complex hierarchies, while collapsed H-DRFQ is better for the flows with simpler attribution structures. Meanwhile, our simulation shows that both H-DRFQ algorithms can achieve near-perfect fairness.