The rapid development of deep space exploration missions has led to the continuous expansion of interplanetary networks (IPNs) for enhanced data transfer capacities and reliability. In this paper, we propose a novel deep reinforcement learning (DRL) framework for optimizing the scaling of IPN topologies (i.e., the placement of relay satellites), such that the routing and data scheduling of interplanetary data transfer (IP-DT) in the scaled IPN achieves maximized performance gain. Our proposal leverages graph neural networks (GNNs) to extract topological correlations within an IPN and hereby learns progressive local rewriting policies for approaching the optimal solution. Extensive simulations verify the effectiveness and robustness of our proposal, showing that the learned relay satellite placements facilitate higher reliability (in terms of data delivery ratios) and lower end-to-end latency for different IPN scenarios, when compared with the existing benchmarks.
Several 5G/6G applications are anticipated to require low latency and high service availability. Latency constraints are expected to be satisfied by utilizing edge computing servers termed multi-access edge computing (MEC) servers. MEC can provide many benefits, including improved network performance, increased reliability, and reduced costs. However, the location of MEC servers can have a significant impact on the reliability of the services they support. In this paper, we study the impact of combination of server location and the physical topology on service availability of connections between the user equipment (UE) and MEC. We concentrate on applications that utilize multi-connectivity between the UE and MEC. We focus on evaluating the availability of services supported by MEC servers located at different positions within the radio access network, such as at macro base stations, at the access sites, or at a central location as well as the effects of the backhaul topology deployed, e.g., ring or tree.
For better network performance and reliability, multi-connectivity (MC) paths are established to take advantage of path diversity for dependable packet delivery. However, MC can sometimes allocate excessive resources without considering the actual availability of the primary path. To address this inefficiency, we propose a risk-aware backup path allocation strategy that accounts for different risks that could impact the primary path's availability. Our approach includes estimating the number of users affected by specific risks and assigning riskdisjoint backup paths to user groups. This allocation method considers both the availability of primary and backup paths, as well as the shared use of backup resources among users. By grouping users according to disjoint risks and allocating backup paths accordingly, our method optimizes resource usage. The simulation results demonstrate that our approach incorporates the risks associated with the allocation of the backup path, enhances the resilience of the network.
In this paper, we focus on adaptive image processing and transmission for aerial imagery tasks. An unmanned aerial vehicle (UAV) engages in these tasks, namely capturing and processing aerial photographs, and transmitting the processed image data to the back-end system at a base station (BS) for further analysis. Specifically, task-oriented semantic features and wireless channel state information (CSI) are leveraged to adjust the resolutions and compression ratios in processing image blocks. The processed image data is then transmitted at optimal rates by selecting appropriate modulation schemes. To maximize task performance while minimizing latency, we propose an adaptive algorithm based on deep reinforcement learning (DRL) that intelligently selects the image processing modes and data modulation schemes. Experimental results on an aerial image classification task show that the proposed adaptive approach significantly improves the accuracy across various signal-to-noise ratios, outperforming traditional image processing and transmission methods.
Ensuring reliable audio and video communication during disaster events is a critical area of study for the scientific community. The effectiveness of first aid in such situations is critical and can be life-saving. In such a scenario, the use of Unmanned Aerial Vehicle (UAV) networks is able to guarantee the provisioning of connectivity in areas severely involved in emergency and disaster events and where the communication infrastructures are seriously compromised. Anyway, this type of network - known as Flying Ad-Hoc Networks (FANET) - has to take into account different issues related to resource constraints, especially in terms of bandwidth and energy. To address these challenges, in this paper, a recruiting strategy inspired by ant colony behavior is proposed. This strategy has been designed in order to consider multiple metrics such as distance, bandwidth, energy, and delay. Integrating such a resource-aware recruiting strategy into the FANET permits the displacement of additional UAVs when needed due to a surge in users' requests or an increasing burden imposed on the existing ones. This approach allows the network to adapt to user demands, guaranteeing effective task completion during critical missions.
Future smart factories are expected to deploy reliable applications over high-performance indoor wireless channels in the millimeter-wave (mmWave) bands. Since these bands are known to be susceptible to high path losses and Line-of-Sight (LoS) blockages, low-cost Reconfigurable Intelligent Surfaces (RISs) are used to enhance the reliability of wireless links. In this paper, we formulate a combinatorial optimization problem, solved with Integer Linear Programming (ILP) to optimally maintain reliable connectivity by solving the problem of allocating RIS to robots in an indoor wireless network. Our model exploits a new feature of the so-called nulling interference from RISs by tuning reflection coefficients. We further consider the system reliability by defining Quality-of-Service (QoS) at receivers in terms of Signal-to-Interference-plus-Noise Ratio (SINR) and connection outages due to insufficient transmission quality. Numerical results for the optimal solution and heuristics show the benefits of optimally deploying RISs by providing continuous connectivity through SINR. We also show that our method can significantly reduce outages due to link quality, while meeting the requirements on system and service reliability.
Unmanned Aerial Vehicles (UAVs) integrated with Mobile Edge Computing (MEC) servers are deployed to achieve rapid service coverage and deliver essential computational resources to end users. However, existing multi-UAV assisted MEC systems face challenges in balancing high-quality service with fairness. This paper investigates a fairness-based 3D trajectory optimization and task offloading strategy in multi-UAV assisted MEC, introducing a two-layer UAV model (fixed wing and rotary wing) to jointly optimize offloading strategies and flight trajectories. A Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MATD3) algorithm is developed to solve the inherently non-convex optimization problem in high-dimensional action spaces. Comparative results show up to 16 % reduction in system delay compared to MATD3-FH.
With the continuous evolution of 6G networks, a contradiction has formed between various and high-density concurrent services and limited resources, which aggravates the risk of node failures due to resource contention. Existing fault tolerance schemes are difficult to adapt to the dynamic 6G edge network due to problems such as high costs and insufficient scalability. In view of this, this paper proposes MigDiffusion: a preemptive migration fault-tolerant model based on Diffusion. This model uses the Graph Attention Networks (GAT) and Transformer for accurate fault prediction and classifies the prediction results through prototype networks. In addition, it innovatively utilizes Diffusion model to generate efficient migration decisions. Experimental results show that, on average, MigDiffusion improves by 14.34 % in energy consumption, 16.36 % in delay, and 34.29 % in SLO violation rate compared with other active faulttolerant benchmark algorithms.
This paper presents a comparative analysis of reinforcement learning (RL) and deep reinforcement learning (DRL) techniques for optimizing reliable electric vehicle (EV) routing and charging in dynamic urban environments. By modeling the EV routing problem as a Markov Decision Process, our study accounts for real-world constraints such as battery levels, charging station availability, traffic conditions, and range anxiety. We evaluate three approaches: traditional Q-Learning, heuristicaugmented Q-Learning (TQL), and Deep Q-Learning (DQL) across simulated networks of varying scales. The experimental results demonstrate that DQL achieves faster convergence and consistently outperforms the other methods in minimizing travel time, improving distance efficiency, and maximizing cumulative rewards while ensuring more reliable route planning. These findings highlight the potential of advanced RL frameworks to enhance EV operational efficiency and charging reliability, offering valuable insights for future smart and sustainable transportation systems.
In this paper, we address the challenges associated with heterogeneous non-independent and identically distributed (non-IID) data and dual jammers in terrestrial and nonterrestrial integrated sensing and communication (TNT-ISAC) networks. More specifically, we consider a multi-cell TNT-ISAC network whose communication and sensing are susceptible to interference from multiple dual jammers. We therefore introduce a resilient federated reinforcement learning (ReFRL) framework that is capable of detecting dual jammers in TNT-ISAC networks by leveraging spectral correlation function (SCF) features for communication signals and time-frequency distribution (TFD) features for sensing signals. We train our framework locally by integrating reinforcement learning (RL) with federated learning (FL) that uses convolutional neural network (CNN) models that have been tailored for SCF and TFD features. Our framework includes a Q-learning-based adaptive weight adjustment mechanism that optimizes local model combinations. It prioritizes the weights of the local models applied to TFD features for sensing devices and SCF features for communication devices. Our solution leverages FL to handle distributed data and RL to adapt to varying conditions. Numerical results significantly improve the handling of heterogeneous non-IID data, enhance the trade-off between sensing and communication, and increase the robustness of our network.
As cloud systems scale, ensuring service availability while minimizing backup resources has become a critical challenge. A reliable method for evaluating service availability is essential for building a robust cloud infrastructure, as accurately estimating backup failure probabilities is critical for effective resource allocation and maintaining overall reliability. The previous model enumerates all the failure patterns of the primary physical machine and the backup physical machine and accumulates the backup failure probability. When the number of physical machines increases, the computational complexity grows exponentially. This paper proposes models calculating backup failure probabilities for two common cloud infrastructure architectures: bare metal and virtual machine. For the bare metal case, a binomial distribution with quadratic time complexity calculates the backup failure probability. For the virtual machine case, integrating the first-fit decreasing algorithm with probabilistic analysis partially reduces the complexity associated with backup physical machine failure patterns. Numerical results confirm that our proposed models yield failure probabilities consistent with previous model results while decreasing computational complexity, particularly as the number of physical machines increases.
This paper focuses on the handover problem in indoor - outdoor hybrid communication scenarios. It adopts a centralized decision - making mode for user handover by the unmanned aerial vehicle base station (UAV - BS) central controller. Aiming at the issues of high dynamic unmanned aerial vehicles (UAV) network topology, frequent handovers caused by the variable link quality in indoor - outdoor hybrid communication scenarios, and the increased risk of interruption, a Markov decision process (MDP) model for the centralized handover problem is proposed. A reward function is defined based on service revenue, handover cost, and interruption cost to guide the handover decision - making agent to avoid the interruption risk. In response to the increase in the dimensions of handover decisions due to the introduction of network slicing technology, an intelligent handover algorithm based on deep reinforcement learning is proposed. The Double Deep Q Network (DDQN) deployed in the base - station controller serves as the intelligent agent for handover decision - making. Through simulation experiments, it is verified that the handover scheme based on deep reinforcement learning proposed in this paper can preferentially reduce the interruption probability according to the design tendency, balance the handover cost and service revenue, achieve the optimal cumulative reward performance, and outperform other traditional algorithms in comprehensive performance.
Telemedicine devices inherently possess dual attributes of healthcare and communications. This study proposes a dedicated testbed model employing a simulated experimental network architecture to evaluate the robustness boundaries of telemedicine service communication links We developed a specialized usability testing methodology leveraging this testbed to assess functional integrity and performance reliability of master-slave devices in controlled laboratory settings. Furthermore, we conducted a systematic meta-analysis of existing usability metrics, providing recommendations for subjective evaluation tools applicable to telemedicine devices across varied clinical scenarios.
Telecommunication networks are highly vulnerable to disasters, disrupting critical services and causing significant economic losses. This paper presents a constraintbased Integer Linear Programming (ILP) model for disaster recovery that jointly optimizes risk-aware routing and revenue maximization under bandwidth constraints. Unlike traditional heuristic approaches, our model dynamically prioritizes highvalue, low-risk services, ensuring optimal bandwidth allocation while mitigating network failures. The risk function accounts for link failure probabilities and service priority levels, influencing path selection in real time. By enforcing strict bandwidth and survivability constraints, the ILP model strategically schedules fewer but more profitable flows, optimizing economic efficiency without compromising resilience. Simulation results demonstrate that our approach reduces service disruptions, minimizes high-risk routing, and improves revenue efficiency compared to Dijkstra-based routing. These findings offer a scalable framework for network operators to enhance disaster resilience while maintaining economic viability.
Quantum Key Distribution allows the exchange of symmetric keys with information-theoretic security between adjacent nodes. Key relay techniques in Quantum Key Distribution Networks (QKDNs) address the need for symmetric keys between nodes that are either too distant for establishing a quantum link or separated by intermediate nodes. The Secure Key Rate is limited, which triggers a need to use buffers (Quantum Key Pools) to store keys for later use. The secure key relay uses One-Time Pad, thus the number of links in the route impacts the number of keys used to deliver a pair of symmetric keys. The need for efficient routing algorithms in QKDNs has led to different routing algorithms. This work revisits some routing approaches that have been improved here. Results show that using asymmetrical link costs reduces blocking, a Long Short-Term Memory approach is enhanced by including cross-validation, and finally some strategies to reduce management traffic are proposed.
In Network Functions Virtualization (NFV) applications, guaranteeing the reliability of mobile users' critical services is a core task for network service providers. Traditional Virtual Network Function (VNF) protection mechanisms employ uniform backup quantities across all VNF types, potentially leading to either insufficient reliability or inefficient resource utilization. This paper investigates the resource allocation problem in complex service function chaining (SFC) scenarios with parallelized VNFs, where the coexistence of different reliability requirements poses significant challenge for resource allocation. To address this challenge, this paper introduces a novel Flexible Backup (FB) mechanism to achieve reliable SFC operation with minimal resource overhead by differentially configuring the number of backups for different VNF types. We represent the problem as mixed-integer nonlinear programming (MINLP) and transform it into mixed-integer linear programming (MILP) using Taylor expansion and linear approximation fitting (TL) with reliability error below 0.15 %. For the complexity of solving multiple SFCs, this paper proposes a VNF protection algorithm based on column generation (CG_VP). Experiments show that CG_VP consumes the same amount of resources as MILP and significantly reduces the solution time (86.3 %).
With the emergence of advanced use cases in In-dustrial, Healthcare and Smart City loT networks, applications and devices exhibit a wide range of QoS requirements. In the realm of loT, WiFi is a widely adopted technology that employs Enhanced Distributed Channel Access (EDCA) for QoS provisioning. However, EDCA's reliance on just four Access Categories (ACs) falls short in meeting the diverse range of QoS requirements of loT networks. SG-EmPOWER, a Software Defined Networking (SDN) framework, has introduced Network Slicing to WiFi networks, offering a robust solution to address the QoS diversity challenges in loT networks. SG-EmPOWER ensures slice priorities through quantum assignments to the slices, however this approach faces limitations in achieving effective slice differentiation, particularly in light of WiFi's Distributed Coordination Function (DCF). To overcome this limitation, we have introduced channel access parameter such as Contention Window (CW) into the slice configuration in addition to quantum value. This not only improves slice QoS differentiation but also brings robust and dynamic control over network QoS. Our proposed slice control mechanism has demonstrated substantial improvements, with an average increase of 47.66 % in per-slice throughput and an average reduction in jitter in high-priority slices of 23 %.
This paper studies the problem of optimal and reliable routing for multicast sessions in wavelength-division-multiplexing (WDM) networks. The objective is to minimize the cost and the blocking probability when establishing several multicast sessions while guaranteeing protection against any single link failure. In the presence of static traffic, two different Integer Linear Programming (ILP) formulations, Joint ILP 1 and Joint ILP 2, find the working trees and the link-disjoint backup paths for the sessions simultaneously, with Joint ILP 2 choosing a specific number of links to route the working trees. The proposed ILPs reduce the cost by 25 – 30 % while coping with a larger number of demands per session compared to state-of-the-art solutions. This work also addresses the problem in a dynamic environment, with sequential multicast sessions of random demands, by proposing an ILP for the optimal solution and a heuristic called Demand-Aware Tree-Forming Optimal Path Pairs (DA- TF -OPP). Dealing with one multicast session at a time, DA-TF-OPP takes into account the sequence in which the demands should be routed and provides a working tree to prevent traffic loops, along with the backup paths. Compared to the state-of-the-art, DA-TF-OPP provides up to 5 % lower average cost and up to 3 % lower average blocking probability.
This paper introduces a novel approach to address the challenges of link adaptation in direct air-to-underwater optical communication systems, employing the deep deterministic policy gradient (DDPG) algorithm and the orthogonal frequency division multiplexing (OFDM) technique under noisy channel state information (CSI) conditions. The proposed method encom-passes dynamic adaptations of beam orientation, beam width, and power levels, treating these parameters as continuous variables to account for the inherent randomness in real-world scenarios. To evaluate the effectiveness of our approach, we compare the achieved success rates and consumed energy with our DDPG-based model against a Q-Iearning-based link adaptation strategy as well as two static benchmarks. Remarkably, the proposed model consistently achieved high success rates, nearly approaching 100%, while concurrently reducing energy consumption by a factor of ten compared to the Q-Iearning-based approach, all within the first 100 episodes.
This paper proposes an analytical model that deter-mines the unavailability of middle box functions, where multiple backup servers protect one or more functions, which we call a multiple-backup shared protection strategy. Middlebox functions as software are implemented on a general-purpose server through the utilization of network function virtualization. The metric of unavailability in middleboxes carries significant importance, underscoring the necessity for streamlined and efficient protection strategies. While existing approaches tackle this problem, they often become complex, or only exceptional cases are considered, prompting new model development. In our model, we enable each function to be protected by multiple backup servers to ensure function availability, and each backup server can recover at most one function simultaneously. We employ a Markov chain to an-alyze state transitions and establish equilibrium-state equations, providing an analytical foundation for assessing the performance of the multiple-backup shared protection strategy. Numerical results show that the multiple-backup shared protection strategy reduces the unavailability by 7.80-72.3 % compared to the single-backup shared protection strategy in our examined cases.
