The evolution of 6G networks introduces heterogeneous services with stringent computing and latency demands. However, constrained edge resources, intricate task dependencies, and dynamic network fluctuations intensify resource contention, increasing the risk of node faults and service interruption. Current fault-tolerant methodologies lack the necessary adaptability to handle the coupled complexity of task interdependencies and volatile resource states, leading to sub-optimal decisions or excessive system overhead. To address these challenges, this paper innovatively proposes TransDiffuse—an intelligent preemptive service migration framework for 6G edge networks. First, the framework employs a Transformer-GAT hybrid model to capture long-range temporal load dynamics and spatial topological constraints, enabling accurate failure prediction. Second, to navigate the trade-off between migration overhead and service robustness, we devise a diffusion-based decision module. This module efficiently explores the discrete combinatorial solution space to synthesize near-optimal service orchestration. Furthermore, a comprehensive evaluation system is constructed to validate the effectiveness of TransDiffuse. Experiments demonstrate that TransDiffuse reduces energy consumption by 32.4%, decreases task completion time by 25.6%, and improves resource balance by 18.7%, while keeping service violations below 5%. This work achieves joint optimization of energy, delay, and resource efficiency, offering a robust solution for resilient service orchestration in 6G edge networks.

Device-to-device (D2D)-aided mobile edge computing (MEC) has emerged as a key enabling technology for future sixth-generation (6G) wireless networks. The goal of D2D-MEC is to reduce system latency for edge user equipments (UEs) by enabling access to cloud computing capabilities at the network edge, thereby supporting high transmission rates. To address the vulnerability of communication signals to physical obstructions, we employ relay techniques to enhance system performance and extend coverage. However, relay nodes and base station (BS) are typically equipped with large-scale antenna arrays, which lead to significant implementation costs and limiting practical deployment. To address this issue in a cost-efficient manner without sacrificing system performance, movable antenna (MA) technology is introduced. The key idea of MA technology lies in dynamically optimizing antenna positions to improve system capacity. Therefore, we propose a novel resource allocation framework for an movable relay-aided D2D-MEC system. The proposed scheme jointly optimizes the MA positions at UEs, relays, and the BS, along with the associated beamforming vectors, MEC server resource allocation, and computational task offloading rates. The objective is to minimize the maximum system latency while satisfying both computation and communication rate constraints. Furthermore, considering that current MA control mechanisms primarily rely on mechanical actuation, MA movement delay is incorporated into the latency model to capture the trade-off between antenna mobility and system delay. The resulting optimization problem is non-convex and involves multiple coupled variables. To solve this problem, we develop a parallel and distributed algorithm based on the penalty dual decomposition (PDD) framework, which is further integrated with the successive convex approximation (SCA) method to obtain a suboptimal solution. Simulation results demonstrate that the proposed algorithm significantly reduces system latency and enhances overall efficiency compared to benchmark schemes employing conventional fixed-position antennas (FPAs) at the relays and BS.

As emerging services such as AI datacenter interconnection, immersive AR/VR, and autonomous driving demand greater transport flexibility and resource efficiency, Metro Transport Networks (MTNs) have become the critical infrastructure for next-generation metro backbones providing flexible and efficient resource orchestration. Nevertheless, the Routing and Slot Allocation (RSA) problem in MTNs remains challenging due to combinatorial complexity, high-dimensional coupling in the action space, and resource fragmentation. We propose EFP-RSA, a novel entropy-regularized reinforcement learning framework that addresses the RSA problem through factorized policy. It factorizes the joint optimization into coordinated path selection and slot assignment, alongside entropy regularization annealing to balance exploration with convergence. Compared with baselines, EFP-RSA reduces blocking probability by $\mathbf{3 5. 3 \% - 6 0. 7 \%}$ and increases slot utilization by $\mathbf{1 9. 4 \% - 5 8. 6 \%}$, while maintaining low load variance and high solution efficiency, demonstrating strong generalization and scalability in complex MTN scenarios.

With the growing demand for highly reliable and resource-efficient multicast services in next-generation networks, deploying multicast service function chains (SFCs) poses significant challenges due to complex interdependencies among VNF placement, routing, and link protection. Unlike traditional unicast scenarios, multicast SFCs must simultaneously serve multiple destinations while ensuring reliability and minimizing resource overhead. This paper presents a reliability-aware multicast SFC deployment framework that integrates centralized optimization with distributed decision-making. A unified reliability-constrained model is formulated to jointly determine VNF placement, multicast routing, and link backup configuration. To enable tractable optimization, the inherently nonlinear reliability-constrained formulation is transformed into an equivalent linear model, achieving coordinated reliability assurance and resource efficiency.To enhance scalability, a game-theoretic distributed algorithm is designed to model domain interactions and iteratively converge to near-optimal solutions with lower complexity. Simulation results show that the proposed framework achieves an effective reliability–cost trade-off, reducing resource consumption by about 15–23% compared with heuristic schemes. The proposed approach provides a flexible and scalable solution for reliable multicast SFC deployment in 6G-oriented networks.

In the realm of the Internet of Vehicles (IoV), the concept of grouping autonomous vehicles into platoons stands out as a promising driving scenario. A platoon comprises interconnected vehicles, with the foremost vehicle designated as the Platoon Leader (PL), while each of those trailing behind is a Platoon Member (PM). In such contexts, information freshness quantified using the Age of Information (AoI) critically ensures road traffic safety. This paper explores the joint packet transmission scheduling and power allocation problem with the objective of minimizing AoI in multi-platoon vehicular networks; these latter exhibiting high dynamics incurring notable uncertainty and complexity. To alleviate this optimization problem’s complexity a decentralized partially observable Markov Decision Process (Dec-POMDP) formulation is adopted. Then, an AoI-aware joint scheduling and power control scheme based on Multi-Agent Twin Delayed Deep Deterministic policy gradient (MATD3) algorithm is proposed. In addition, in order to improve the efficiency of the MATD3’s learning phase, the algorithm has been augmented with Priority Experience Replay (PER). Simulation results show that this approach outperforms the baseline MATD3 method by 17.3% in terms of the achieved mean AoI.

With the development of Unmanned Aerial Vehicles (UAVs), Low Earth Orbit (LEO) satellites, and Mobile Edge Computing (MEC), integrating MEC into Space–Air–Ground Integrated Networks (SAGINs) is crucial for Internet of Remote Things (IoRT). However, limited UAV resources and VM I/O interference in MEC hinder efficient task offloading and resource allocation. To address this, we propose an MEC-enabled SAGIN model and formulate a joint optimization problem of UAV trajectory, task offloading, and MEC resource allocation to minimize delay and energy consumption. The problem is modeled as a Markov Decision Process (MDP) and solved using a multi-agent twin delayed deep deterministic policy gradient (MATD3) algorithm. Simulation results show that MATD3 reduces task delay and energy consumption by 14% and 9%, respectively, compared with MADDPG, and achieves an additional 5% delay reduction over the case without dynamic VM allocation.

Network Function Virtualization (NFV) enables flexible deployment of Service Function Chains (SFCs) by hosting Virtual Network Functions (VNFs) on commodity servers. However, physical node and link failures can disrupt an operational SFC and lead to service interruption. This paper investigates post-failure SFC recovery via VNF migration and virtual-link rerouting under reliability and resource constraints. We model the recovery decision as selecting a feasible migration plan that remaps VNFs to available physical nodes and reroutes virtual links over surviving physical paths, while satisfying computing and bandwidth capacities as well as end-to-end delay limits. The migration overhead is primarily dominated by the number of migrated VNFs and is jointly optimized with bandwidth consumption. To address the large-scale combinatorial nature of the problem, we propose a column generation framework in which the Restricted Master Problem (RMP) selects recovery plans from a growing column pool and the Pricing Problem (PP) generates new plans with negative reduced cost. A simulated annealing (SA) procedure is embedded into the PP to improve exploration over the neighborhood induced by VNF remapping and path re-optimization. Simulation results show that, compared with Greedy, the proposed CG-SA reduces bandwidth consumption by 35.3%–57.9% (50.0% on average) and improves the median SFC reliability from 0.909 to 0.947 under random failures.

Unmanned aerial vehicle (UAV)-assisted mobile edge computing (MEC) has emerged as a promising paradigm to support computation-intensive and latency-sensitive applications in urban environments. However, building-induced blockages and complex urban geometries severely degrade air-to-ground (A2G) links, causing unstable connectivity and undermining quality-ofservice (QoS) guarantees. To address these challenges, this paper proposes OA-LyDROO, an obstacle-aware and dynamic optimization framework that jointly handles computation offloading, resource allocation, and UAV trajectory design. A hybrid A2G channel model is developed to capture obstacle-induced occlusion and time-correlated fading. By integrating Lyapunov optimization with deep reinforcement learning, OA-LyDROO enables real-time joint optimization under temporal coupling, stochastic task arrivals, and probabilistic delay constraints. Theoretical analysis proves queue stability and establishes an $[O(1 / V), O(V)]$ performance-delay tradeoff. Extensive simulations have verified the effectiveness of OA-LyDROO. Compared with the benchmark scheme, it significantly reduces queue length without sacrificing throughput, while ensuring an energy efficiency improvement of $\mathbf{1 3. 6 - 4 6 \%}$. The obstacle-aware channel modeling ensures robust link quality and mitigates deep fading, while both theoretical and experimental results confirm the superiority of the proposed framework.

The emergence of intelligent mobile technologies and the widespread adoption of 5G wireless networks have made Federated Learning (FL) a promising method for protecting privacy during distributed model training. However, traditional FL frameworks rely on static aggregators such as base stations, encountering obstacles such as increased energy demands, frequent disconnections, and poor model performance. To address these issues, this paper investigates an innovative aUtonomous Aerial Vehicle (UAV)-assisted FL framework, aiming to utilize UAVs as mobile model aggregators to collaborate with devices in training models, while minimizing the total energy consumption of devices and ensuring that FL can achieve the target model accuracy. By adopting the Distributed Approximate NEwton (DANE) method for local optimization, we analyze the convergence of FL and derive device scheduling constraints that aid in convergence. Accordingly, we formulate a problem of minimizing the total energy consumption of devices, integrating a constraint on global model accuracy, and jointly optimizing the UAV trajectory, device scheduling, bandwidth allocation, time slot lengths, as well as the uplink transmission power, CPU frequency, and local convergence accuracy. Then, we decompose this non-convex optimization problem into three subproblems and propose an iterative algorithm based on Block Coordinate Descent (BCD) with convergence guarantee. Simulation results indicate that, compared with various benchmark methods, our proposed UAV-assisted FL framework significantly reduces the total energy consumption of devices and achieves an improved trade-off between energy and convergence accuracy.

In the context of multicast Network Function Virtual(NFV)-enabled Internet of Things (IoT), numerous sensor devices must efficiently transmit data to multiple data centers for real-time monitoring and analysis. Certain data centers require aggregated data from multiple sensors for informed decision-making. Outdated information lead to incorrect decisions, resulting in economic losses. Therefore, ensuring the timely and effective delivery of information is of paramount importance. However, the issue of information freshness in multicast networks has received limited attention. The deployment Virtual Network Functions (VNFs), data scheduling, and the multitude of routing possibilities pose significant challenges to studying information freshness. We introduce Mixed-Age of Aggregated Information (MAoAI) to quantify information freshness in multicast networks, integrating Age of Information (AoI) and Age of Aggregated Information (AoAI). To address this, we propose an optimization framework for coordinating multicast service requests. This framework takes into account VNF deployment and sharing, multicast routing, transmission scheduling, and data aggregation, mathematically formulated as a complex Integer Linear Programming (ILP) model. To tackle the scalability issue, we develop a Nash equilibrium-based Multicast and Scheduling Game (MSGame) approach, reducing CPU runtime by an average of 98.02% compared to ILP. Comprehensive simulations show improved solution quality and approximate optimal solutions with fewer iterations.

Uncrewed Aerial Vehicle (UAV) is extensively employed as a mobile base station in areas with inadequate cellular infrastructure to enhance the freshness of vehicle sensors. The Age of Information (AoI) is a metric utilized to characterize the freshness of information produced by vehicle sensors. This paper investigates the use of Uncrewed Aerial Vehicles (UAVs) as mobile base stations to enhance the freshness of vehicle sensor information in areas with inadequate cellular infrastructure. We focus on minimizing the Age of Information (AoI) and UAV energy consumption in a Vehicle-to-UAV (V2U) network within the Manhattan scenario. The challenge lies in jointly optimizing UAV trajectories and vehicle data packet scheduling amidst high vehicle mobility and limited communication range. To address this issue, we employ Reinforcement Learning (RL) to formulate the problem as a Markov Decision Process (MDP), proposing a Dueling Double Deep Q-Network (D3QN) method for trajectory and scheduling optimization. We also introduce Priority Experience Replay (PER) to improve reward acquisition for the UAV, addressing the issue of sparse rewards due to the expansive space for vehicle movement. Simulation results provide empirical evidence supporting the efficacy of the proposed algorithm in comparison to baseline policies.

Virtual reality (VR) demands extensive computation while imposing strict requirements for ultralow latency, placing a significant burden on wireless communication systems. In recent years, there has been a growing interest in leveraging uncrewed aerial vehicle (UAV)-assisted mobile-edge computing (MEC) as a promising technology to provide flexible computing resources at the edge of wireless networks. To meet the computational demands of VR, we propose a collaborative three-layer edge computing framework assisted by multiple UAVs. This framework enables VR rendering tasks to be executed locally on user devices or offloaded to UAVs and base station (BS) for execution. By jointly optimizing the flight trajectories of UAVs and the rendering modes of users, we aim to maximize the average rendering completion rate, defined as the ratio of successfully completed VR rendering tasks within the specified delay constraints, while minimizing the average energy consumption of UAVs. To enhance adaptability, we adopt a multiagent twin delayed deep deterministic policy gradient (MATD3) approach that provides an efficient strategy for multi-UAV-assisted VR rendering, even in partially observable scenarios. Simulation results validate our proposed approach and demonstrate that the MATD3 algorithm surpasses the classical multiagent deep deterministic policy gradient (MADDPG) algorithm in terms of convergence speed and the average rendering completion rate.

Multiaccess edge computing (MEC) plays a crucial role in providing low-latency and high-data transmission services to Internet of Things (IoT) devices. However, in remote areas where deploying edge devices is challenging, optimizing delay remains a significant research focus. To address this issue, our research investigates a multidrones-assisted IoT task offloading model. In this model, tasks generated by IoT devices equipped with energy harvesting (EH) capabilities are offloaded to MEC servers with the assistance of multiple drones. In order to monitor and manage the energy consumption of IoT devices and the task backlog of edge servers, the energy consumption and task update queues are established. We formulate a mixed integer nonlinear programming (MINLP) problem, which aims to optimize the allocation of communication and computation resources to minimize the execution latency of IoT devices. To ensure the stability of each queue, we employ the weighted perturbation method within the Lyapunov optimization framework to decompose the original problem. And a low complexity multidrones assisted offloading (MUAO) algorithm is designed. Simulation results show that MUAO consistently exhibits lower latency and energy consumption compared to the baseline scheme and other existing algorithms, while maintaining a low packet loss rate of only 5%.

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.

In the realm of complex IoT-based smart city advancements, the real-time reception, processing, and maintenance of up-to-date multi-sourced data is essential for ensuring efficient urban infrastructure operations and functionality. Beyond the typical Age of Information (AoI), such applications vociferate the urgent need for a new metric, capable of capturing and accounting for the age of the aggregated data; namely, the Age of Aggregated Information (AoAI). This paper addresses an AoAI minimization problem for mixed-paths IoT networks. This problem is formulated as a Mixed Integer Linear Program (MILP) that jointly considers data packet scheduling and routing as well as nodal energy and power constraints. To overcome this problem’s notable complexity, the Column Generation Algorithm (CGA) is used to break it down into a Relaxation Master Problem (RMP) and a Pricing Problem (PP) with the objective of identifying optimal scheduling and aggregation strategies. Experimental results demonstrate the potency of the proposed CGA-based algorithm in generating accurate sub-optimal solutions with no more than 1.06% deviation from their optimal counterparts; an outstanding result that existing algorithms have failed to achieve. The variations in AoAI and latency trends were compared and found to be in-line for fixed network configurations.

Timely gathering of sensing data is critical in wireless sensor networks (WSNs). However, in delay-sensitive applications, maintaining the freshness of collected data poses a significant challenge. To tackle this issue, an age of information (AoI)-aware data collection method leveraging uncrewed aerial vehicle (UAV) and intelligent reflective surface (IRS) is proposed in this work. Particularly, a UAV is employed to traverse over ground sensor nodes (SNs) and reliably collect their sensing data where the received signal strength is enhanced through IRS. The UAV's flight trajectory and its association with SNs, as well as the IRS phase control strategy are jointly optimized to minimize the weighted sum of the average AoI of the SNs and energy consumption of the UAV. However, this optimization is complicated by potential inaccuracies in IRS channel state estimation. To tackle this challenge, we propose an enhanced deep reinforcement learning (DRL) framework that incorporates a dual-network agent with two nested neural networks (NNs): UAV-NN, which jointly optimizes the UAV trajectory and SN association, and IRS-NN, which dynamically adjusts IRS phase shifts based on sampled channel states, UAV position, and associated SN. By integrating this architecture into proximal policy optimization (PPO) and deep Q-network (DQN), we develop two novel algorithms: PPO-RAC and DQN-RAC, tailored for IRS-assisted UAV data collection. Extensive simulations validate their effectiveness across diverse scenarios, demonstrating significant AoI reduction compared to baseline methods.

The rapid proliferation of Internet of Things (IoT) devices with sensing, monitoring, and control capabilities has fueled the emergence of diverse real-time IoT applications. These applications often require the efficient transmission of collected sensitive data to centralized or distributed processing units, such as cloud servers, edge nodes, or fog computing platforms—to enable timely responses. Ensuring the freshness of information while maintaining sustainable data transmission under energy constraints remains a critical challenge. To address this issue, we quantify information freshness using Age of Aggregated Information (AoAI) and investigate its minimization within energy-harvesting multi-hop IoT networks. Specifically, we develop data transmission schedules that jointly consider energy harvesting and packet aggregation. This problem is formulated as a Mixed-Integer Linear Programming (MILP) model, integrating data scheduling, packet aggregation, link interference, and energy resource management across multiple nodes. To tackle the computational complexity of this NP-hard problem, we propose an Energy-Aware Column Generation Algorithm (CGA). The original problem is decomposed into a Restricted Master Problem (RMP) and a Pricing Problem (PP). The RMP models data aggregation at Base Station (BS), while the PP incorporates an Energy-Aware Scheduling Algorithm (EASA) to generate feasible scheduling solutions. Simulation results demonstrate that the proposed CGA consistently outperforms EASA, the Genetic Algorithm (GA), and the Juventas Algorithm, achieving an average deviation of only 1.84% from the global optimal solution.

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 %).

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.