关键词： Mobile edge computing   Trajectory optimization   Offloading strategy   Unmanned aerial vehicle   Deep reinforcement learning  

摘要： This paper addresses latency challenges in cloud-based task offloading caused by geographical server-user disparities by integrating Unmanned Aerial Vehicles (UAVs) with Mobile Edge Computing (MEC). We propose a dynamic UAV-assisted framework that optimizes real-time parameter adjustments to minimize system latency. Key challenges include joint task offloading, UAV trajectory optimization, and User Equipment (UE) occlusion detection under mobility constraints. To resolve these, we transform the problem into a Markov Decision Process (MDP) and develop an enhanced Adaptive Delayed Deep Deterministic Policy Gradient (AD3PG) algorithm, which improves upon DDPG by incorporating delayed updates and neural network tuning. The algorithm dynamically optimizes three critical aspects: UAV-UE connectivity establishment, occlusion-aware dual noise power configuration, and adaptive task offloading ratios. Extensive simulations demonstrate AD3PG's superiority over baselines such as DDPG and Twin-Delayed DDPG (TD3) in reducing total system latency by 13.2-35.3 % under dynamic scenarios (e.g., 8 UEs with 100MB total task volume). Specifically, AD3PG achieves a task completion delay of 83.5-88 s across varying UE quantities, outperforming DDPG (89-115 s) and TD3 (87-95 s). These results validate the proposed framework's efficacy for latency-sensitive applications in UAV-MEC systems.

关键词： Mobile edge computing   Task offloading   Cloud computing   Mobility prediction   Fault tolerance   Deep reinforcement learning   Energy efficiency  

摘要： Mobile edge computing (MEC) improves resource-limited mobile devices by transferring demanding computational activities to edge servers and cloud infrastructure, thereby mitigating issues related to user mobility and network fluctuations. This paper introduces Flexible Mobility-Based Task Offloading (FlexiMRO), a mobility-oriented, fault-tolerant framework for adaptive task offloading in MEC and cloud settings. The suggested methodology incorporates a long short-term memory model to forecast user trajectories and network states with high transmission precision, alongside a deep reinforcement learning strategy utilizing Q-Deep networks for optimal job distribution at the edge and cloud layers. A random forest classifier guarantees fault tolerance by accurately predicting server failures and facilitating the reallocation of ongoing jobs. Comprehensive simulations illustrate the superiority of FlexiMRO compared to current methodologies, achieving a latency reduction of up to 70%, an enhancement in energy efficiency of 60%, and a failure rate of 80% in contrast to 70%. FlexiMRO offers a scalable and adaptable solution for 5G and IoT applications by minimizing latency, energy consumption, and quality of experience, efficiently utilizing edge and cloud computing to guarantee uninterrupted service delivery in dynamic MEC environments.

关键词： Video caching   Retention time   Mobile edge computing   Deep reinforcement learning   Storage cost  

摘要： The current massive video requests have caused severe network congestion. To reduce transmission latency and improve user Quality of Experience (QoE), caching infrastructures are deployed closer to the edge. Nowadays, most caching systems tend to cache content with a high programming voltage to ensure a long retention time, which leads to significant cache damage. However, as new videos emerge every second, the rapidly changing popularity makes long retention time wasteful in terms of caching resource. Moreover, with the rise of emerging video formats (such as virtual reality content), the diverse requirements for transmission latency across various video categories make balancing user QoE more challenging. To tackle these challenges, we propose a joint optimization framework that balances user QoE and operational costs through video category recognition and adaptive retention time selection. First, we model user QoE as transmission latency cost and further formulate the optimization problem as a Markov Decision Process (MDP) to minimize the system cost. To solve the proposed problem, we design a two-step Double Deep Q-Network (DDQN)-based scheme. The scheme first determines the optimal retention time through unifying the process of action selection and state-value evaluation. Secondly, it makes replacement decisions according to the computed caching value of each content. By validating on three datasets, the experiments show that the proposed scheme outperforms the baseline algorithms in both cache hit rate and system cost.

关键词： Service function chaining   Mobile edge computing   Federated learning   Model compression   Anomaly detection  

摘要： The escalating demand for network services has prompted the evolution of Service Function Chaining (SFC) within 6G networks to deliver sophisticated, customized services while ensuring robust cybersecurity. This paper introduces an efficient and secure framework for SFC in Mobile Edge Computing (MEC) environments, termed the Federated Learning-based SFC (FL-SFC), which integrates SFC, MEC, and Federated Learning (FL) to enhance service policy decision-making and safeguard user privacy. The FL-SFC framework enables dynamic updating of service policies and optimizes communication efficiency. We propose an anomaly detection model, CNN-GRU, which combines Convolutional Neural Networks (CNNs) and Gated Recurrent Units (GRUs) to significantly improve anomaly detection performance at the network edge. Additionally, to address the high communication costs associated with service policy models, we have designed a model compression mechanism leveraging sparsification and quantization techniques, which substantially reduces communication overhead during model training. Simulation experiments demonstrated the superiority of the FL-SFC framework and the CNN-GRU model in detection performance over existing methods. Results indicate that our model excels in accuracy, precision, recall, and F1-score while significantly reducing the number of communication bits, thereby validating the effectiveness of our approach.

关键词： Mobile edge computing   Task offloading   Content caching   Deep reinforcement learning  

摘要： In the context of 5G architecture, mobile edge computing (MEC) has emerged as a promising computing paradigm within this framework. It involves deploying servers at the edge of the network with computing and storage resources to meet the low-latency and high-energy requirements of emerging applications. While many studies have focused on task offloading decisions in MEC, the performance improvement brought by content caching has often been neglected. In this paper, the joint optimization problem of task offloading and content caching in a multi-user, multi-server MEC system is investigated. An offloading method based on multi-agent deep reinforcement learning is proposed, which integrates self-attention with multi-agent deep deterministic policy gradient (MADDPG) to determine task offloading policies. To further enhance the performance of offloading tasks, a dynamic cache decision optimization algorithm (DCDOA) is developed for more effective content caching decisions. The simulation results demonstrate that the proposed method outperforms other baseline algorithms in different scenarios. Specifically, the proposed method reduces the long-term average user cost by 45.4%, 22.2%, 10.6%, and 8%, respectively, compared to the all-cloud execution method, the random offloading method, the DDPG method, and the MADDPG method.

关键词： Servers   Resource management   Computational modeling   Optimization   Measurement   Quality of experience   Computer architecture   Wireless communication   Videos   Multitasking   Mobile edge computing   task offloading   deep reinforcement learning   AIGC services  

摘要： Artificial Intelligence-Generated Content (AIGC) has emerged as a transformative paradigm, enabling the autonomous creation of diverse content. By offloading model inference tasks to the network edge that is closer to mobile users (MUs), Mobile Edge Computing (MEC) has the potential to significantly enhance the performance of AIGC services. In practice, however, it is challenging to optimally manage MEC-empowered AIGC services, due to the lack of well-defined AIGC-specific metrics, as well as the dynamic workload and computation-intensive nature of AIGC services. In this paper, we first define a novel AIGC metric based on extensive real data experiments, and then study the joint task offloading and resource allocation problem in a generic MEC-empowered AIGC network, where MUs can offload model inference tasks to local or remote Base Stations (BSs), aiming at maximizing their Quality of Experience (QoE). The problem is challenging due to the fast and randomly changing of environments, as well as the necessity for real-time, asynchronous decision-making. To tackle these challenges, we propose two deep reinforcement learning algorithms based on the Proximal Policy Optimization (PPO) framework: Single-Layer PPO (SL-PPO) and Multi-Layer PPO (ML-PPO), designed for slow-changing and fast-changing environments, respectively. In the SL-PPO algorithm, both task offloading and resource allocation decisions are made simultaneously when tasks arrive. In the ML-PPO algorithm, the task offloading decision is made immediately when tasks arrive, while the resource allocation decision is deferred until tasks are scheduled for processing or transmission in the corresponding queues. Simulation results show that (i) both algorithms outperform existing methods in the literature, and can increase the average utility by up to 47% and 48.8%;(ii) both algorithms can effectively manage the trade-off between latency and energy consumption.

关键词： Internet of Things   Mobile edge computing   Nonorthogonal multiple access   Unmanned aerial vehicles   Physical layer security  

摘要： This paper studies security task offloading for dual unmanned aerial vehicle (UAV)-assisted mobile-edge computing (MEC) in Internet of Things (IoT) systems, in which one UAV is used to aid ground edge devices (EDs) in computing the offloaded tasks, another UAV as a jammer to discourage an eavesdropper. We employ uplink nonorthogonal multiple access (NOMA) to EDs to carry out the offloading procedure and so offer a four-phase protocol to ensure secure task offloading. To evaluate the system's security offload performance, we derive a closed-form formula for the secrecy successful computation probability (SSCP) using the fading Nakagami-m channel model. In addition, to increase the system's security offload performance, we create an SSCP maximization problem by optimizing parameters such as the location and altitude of dual UAVs, and the offloading ratio of ED. To solve this problem, a firefly algorithm (FFA)-based technique was used. Finally, Monte Carlo simulation results are presented to validate the accuracy of our analysis by simulating system performance under various system parameter values, such as dual UAV's location and altitude;average transmit signal-to-noise ratio (SNR), offloading ratio and comparison between NOMA and orthogonal multiple access (OMA) schemes for the considered system.

关键词： Autonomous aerial vehicles   Trajectory   Array signal processing   Optimization   Integrated sensing and communication   Resource management   Clutter   Vectors   Accuracy   Signal to noise ratio   AAV-relayed communication   integrated sensing and communication   mobile edge computing   trajectory design   joint optimization  

摘要： In this paper, we investigate joint trajectory and beamforming design for autonomous aerial vehicle (AAV)-relayed integrated sensing and communication (ISAC) systems with mobile edge eomputing (MEC) under the clutter environment. Due to the limited on-board computing capability, the AAV has to offload sensing echoes to the base station (BS) for efficient processing. A novel relay-based ISAC-then-offload frame structure is considered. We aim to maximize the throughput of the BS-AAV-user relaying link while ensuring sensing accuracy and efficient sensing data offloading. The non-convex problem is solved using an alternating optimization algorithm based on successive convex approximation (SCA). Simulation results illustrate that our proposed algorithm achieves near-optimal communication performance while guaranteeing sensing accuracy, addressing the balance between the communication and sensing performance. Furthermore, we evaluate the impact of critical system parameters including sensing constraints, power control factor, and AAV flight duration on communication performance, and explore the trade-offs between energy efficiency and spectral efficiency under varying sensing data intensity and offloading duration.

关键词： Unmanned aerial vehicle   STAR-RIS   Mobile edge computing   Lyapunov optimization   Successive convex approximation  

摘要： Unmanned aerial vehicle (UAV)-assisted wireless power transfer (WPT) is a promising technology for delivering sustainable energy to energy-constrained ground users (GUs) in mobile edge computing (MEC) networks. Nevertheless, the network performance is severely limited by channel fading. Compared with the conventional reconfigurable intelligent surface (RIS) limited to half-space coverage, the simultaneously transmitting and reflecting (STAR)-RIS achieves full-space coverage, thereby enhancing both WPT efficiency and computational task offloading performance. To fully investigate the potential of STAR-RIS, this paper proposes a STAR-RISassisted UAV-enabled MEC system aiming to minimize long-term latency while ensuring system stability by jointly optimizing time slot allocation, STAR-RIS coefficient matrices, and UAV trajectory. By leveraging the Lyapunov optimization method, the original long-term stochastic optimization problem is transformed into tractable deterministic sub-problems, which are then solved by using successive convex approximation, penalty functions, and convex optimization techniques. Simulation results demonstrate that the proposed scheme effectively balances long-term latency reduction and system stability, achieving significant performance gains compared with baseline schemes.

关键词： Active reconfigurable intelligent surface   Mobile edge computing   Non-orthogonal multiple access   Semi-grant-free transmission  

摘要： In this paper, an active reconfigurable intelligent surface (RIS) and non-orthogonal multiple access (NOMA)-assisted mobile edge computing (MEC) wireless system with semi-grant-free (SGF) transmission is investigated. In the considered system, a grant-based user (GBU) is required to offload its data to a base station (BS) for processing with a certain computing ability within a time deadline and an energy budget in front of an eavesdropper, and meanwhile a grant-free user (GFU) in a group of GFUs is scheduled to offload its data to the BS based on SGF-NOMA, which can interfere with the eavesdropper to enhance the GBU secrecy performance. Two scheduling strategies for GFU are employed: one prioritizes scheduling the GFU with the maximum communication channel gain (MCCG), while the other prioritizes scheduling the GFU with the maximum jamming channel gain (MJCG). The approximate expressions of the system success probability (SSP) under adopted strategies are derived. Based on a power budget consisting of the GBU and the RIS, we further derive the optimal amplification factors of the active RIS, which have an effect on the performance of GBU's maximum achievable secrecy rate. Then, we take the passive RIS system with the same power budget into consideration, based on which various system parameters on the SSP and GBU's maximum achievable secrecy rate are simulated. Simulation results verify the theoretical results and show that the active RIS aided system with adopted strategies achieves better secure performance and effectively improves the SSP. Moreover, for a fixed power budget, allocating more power budget to the active RIS instead of the GBU is more effective for our focused performance indexes. Interestingly, with an ample power budget, the MCCG strategy achieves the upper limit of the SSP while the MJCG strategy shows its superiority for an insufficient power budget.