关键词： Mobile edge computing   Task scheduling   Reinforcement deep learning   DAG task   Internet of Things   Data analysis  

摘要： As the deployment of Internet of Things (IoT) devices becomes more widespread, the demand for data analysis services is also increasing in both range and volume. Despite the increasing demand for services due to the rise of IoT devices, efficiently processing latency-sensitive data analysis tasks (DATs) remains a significant challenge, especially in areas where communication infrastructure is limited or unavailable. Unmanned aerial vehicles (UAVs) can enhance communication coverage and quality by leveraging their flexible deployment characteristics, particularly in areas where network coverage is limited for IoT devices. UAVs can provide edge computing services while simultaneously offering network coverage, making them a valuable asset for improving communication in such areas. In this paper, we propose an Mobile Edge Computing (MEC) system that utilizes multiple UAVs equipped with edge servers to offer latency-sensitive analytical services to IoT devices on the ground. However, traditional task scheduling algorithms are challenging to adapt to dynamic and complex edge network environments, particularly for scheduling tasks with dependencies. To address this challenge, we employ deep reinforcement learning (DRL) to develop Directed Acyclic Graph (DAG) task scheduling algorithms and UAV deployment optimization algorithms. Our algorithm enables us to obtain optimal scheduling and deployment adjustment strategies in dynamic and changing environments. Simulation experiments demonstrate that in a MEC system comprising of multiple UAVs, our algorithm can swiftly converge to the optimal value, resulting in a significant reduction in DAT response time and cluster energy consumption compared to baseline algorithms.

关键词： Resource management   Task analysis   Servers   Delays   Reliability engineering   Error probability   Reliability theory   Time division multiple access   mobile edge computing   finite blocklength   extreme value theory   6G  

摘要： In this paper, we study a multi-access mobile edge computing (MEC) network in the industrial Internet-of-Things (IoT) scenario, which aims at providing a joint computation service for a group of sub-tasks offloaded from multiple user equipments (UEs). The whole MEC service, including a communication phase and a computation phase, is required to satisfy both a low latency and a high reliability requirement. We derive the end-to-end reliability (error probability) of the whole MEC service and provide corresponding reliability-optimal design frameworks, where both the perfect channel state information (CSI) and outdated CSI scenarios are considered. In particular, we characterize the low-latency communication behavior with the consideration of the finite blocklength (FBL) impact, and exploit the extreme value theory to study the delay violation probability in the computation phase. Following the characterizations, in the perfect CSI scenario, a design framework minimizing the instantaneous end-to-end error probability is provided, i.e., via optimally choosing the time length for each user's offloading and the time length for the computation phase. We rigorously prove the convexity of the problem, investigate the relationships among the variables in the optimal solution, based on which a low-complexity method is proposed achieving the optimal solution. In addition, for the scenario with only the outdated CSI, after deriving the expected end-to-end error probability conditioned on the outdated CSI value, a corresponding optimal time allocation design is provided as well, where the convexity of the formulated problem is characterized and the optimal solution is obtained. Via simulations, we validate our analytical model and evaluate the network performance under the design.

关键词： Mobile edge computing   Service caching   Task partitioning   Computation offloading   Deep reinforcement learning  

摘要： Mobile Edge Computing (MEC) federations aim to establish a joint edge service model between Edge Infrastructure Providers (EIPs) and clouds, facilitating the sharing and utilization of MEC services and resources. However, in such a hierarchical multi-EIP MEC federation environment, optimizing service caching, task partitioning, and pricing strategies to maximize the profits gained by EIPs and minimize the cost of mobile devices (MDs) is challenging. To address this challenge, This paper first formulates a two -stage multi -leader, multi -follower Stackelberg game between EIPs and MDs. Then, a new method combining a Stackelberg-based Multi -Agent Deep Deterministic Policy Gradient (STMADDPG) algorithm and a Transformer -based Popularity Prediction (TPP) model is devised to learn the optimal strategies. Extensive experiments with real -world datasets are conducted to demonstrate the excellent performance of the proposed STMADDPG-TPP method.

关键词： Task analysis   Heuristic algorithms   System dynamics   Measurement   Data processing   Minimization   Servers   Age of information   mobile edge computing   post-decision state   deep reinforcement learning  

摘要： 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.

关键词： Task analysis   Processor scheduling   Scheduling   System performance   Probability density function   Probability   Power system reliability   Mobile edge computing   user scheduling   computational access point selection   system outage probability  

摘要： The advent of mobile edge computing (MEC) has effectively resolved the challenges associated with computation-intensive applications. In this letter, we investigate a multi-user MEC system with multiple computational access points (CAPs) in the presence of an eavesdropper. In contrast to existing works that only considered user scheduling or CAP selection, we propose a joint user scheduling and CAP selection policy, and derive the system outage probability under the proposed policy in closed form. Simulations are provided to validate the correctness of the theoretical analysis and demonstrate the effectiveness of the proposed policy. The results confirm that the proposed joint user scheduling and CAP selection policy outperforms various existing policies that independently consider user scheduling and CAP selection.

关键词： Computational modeling   Training   Data models   Costs   Privacy   Mobile computing   Degradation   Differential privacy   federated learning   lottery ticket hypothesis   mobile edge computing   zero -concentrated DP  

摘要： Federated learning (FL) can train collaboratively with several mobile terminals (MTs), which faces critical challenges in communication, resource, and privacy. Existing privacy-preserving methods usually adopt instance-level differential privacy (DP), which provides a rigorous privacy guarantee but with several bottlenecks: performance degradation, transmission overhead, and resource constraints. Therefore, we propose Fed-LTP, an efficient and privacy-enhanced FL framework with Lottery Ticket Hypothesis (LTH) and zero-concentrated DP (zCDP). It generates a pruned global model on the server side and conducts sparse-to-sparse training from scratch with zCDP on the client side. On the server side, two pruning schemes are proposed: (i) the weight-based pruning (LTH) determines the pruned global model structure;(ii) the iterative pruning further shrinks the size of the pruned model. Meanwhile, the performance of Fed-LTP is boosted via model validation based on the Laplace mechanism. On the client side, we use sparse-to-sparse training to solve the resource-constraints issue and provide tighter privacy analysis to reduce the privacy budget. We evaluate the effectiveness of Fed-LTP on several real-world datasets in both independent and identically distributed (IID) and non-IID settings. The results confirm the superiority of Fed-LTP over state-of-the-art (SOTA) methods in communication, computation, and memory efficiencies while realizing a better utility-privacy trade-off.

关键词： 5G/6G vehicular networks   Graph neural network   Mobile edge computing  

摘要： Mobile Edge Computing (MEC) is a key technology that emerged to address the increasing computational demands and communication requirements of vehicular networks. It is a form of edge computing that brings cloud computing capabilities closer to end-users, specifically within the context of vehicular networks, which are part of the broader Internet of Vehicles (IoV) ecosystem. However, the dynamic nature of traffic flows in MEC in 5G/6G vehicular networks poses challenges for accurate prediction and resource allocation when aiming to provide edge service for mobile vehicles. In this paper, we present a novel approach to predict the traffic flow of MEC in 5G/6G vehicular networks using graph-based learning. In our framework, MEC servers in vehicular networks are construed as nodes to construct a dynamic similarity graph and a dynamic transition graph over a duration of multiple days. We utilize Graph Attention Networks (GAT) to learn and fuse the node embeddings of these dynamic graphs. A transformer model is subsequently employed to predict the vehicle frequency accessing the edge computing services for the next day. Our experimental results have shown that the model achieves high accuracy in predicting edge service access volumes with low error metrics.

关键词： Mobile edge computing   Bi-level optimization   Nash equilibrium   Game theory  

摘要： In the mobile edge computing system subject to wireless interference, the Edge Server Provider (ESP) aims to offer profitable computing resources to Device Managers (DMs), who make optimal strategies based on the provided prices. However, the existence of mixed variables typically constitutes an NP -hard problem, posing a significant challenge for optimization. In response to address this issue, we formulate a bi-level optimization problem, where the upper level is devoted to optimizing the pricing of computing resources. The lower level optimizes DM migration strategies and resource allocation at specified prices. Leveraging game theory, we achieve distributed and efficient computation offloading by formulating the distributed computing offloading strategy among lower -level DMs as a task offloading game. Our analysis identifies Nash equilibrium and finite improvement characteristics within the game. Based on these insights, we present a bi-level distributed computation offloading algorithm capable of reaching Nash equilibrium, thus optimizing the profit for both DMs and ESP. The experiments have demonstrated the algorithm's effectiveness in reducing system costs and maximizing the profits of ESP with DMs across diverse scenarios.

关键词： Delays   Multitasking   Internet of Things   Artificial neural networks   Task analysis   Computational modeling   Inference algorithms   Asynchronous advantage actor-critic   distributed DNN inference   mobile edge computing   model partitioning   multi-task learning  

摘要： Model partitioning is a promising technique for improving the efficiency of distributed inference by executing partial deep neural network (DNN) models on edge servers (ESs) or Internet-of-Things (IoT) devices. However, due to heterogeneous resources of ESs and IoT devices in mobile edge computing (MEC) networks, it is non-trivial to guarantee the DNN inference speed to satisfy specific delay constraints. Meanwhile, many existing DNN models have a deep and complex architecture with numerous DNN blocks, which leads to a huge search space for fine-grained model partitioning. To address these challenges, we investigate distributed DNN inference with fine-grained model partitioning, with collaborations between ESs and IoT devices. We formulate the problem and propose a multi-task learning based asynchronous advantage actor-critic approach to find a competitive model partitioning policy that reduces DNN inference delay. Specifically, we combine the shared layers of actor-network and critic-network via soft parameter sharing, and expand the output layer into multiple branches to determine the model partitioning policy for each DNN block individually. Experiment results demonstrate that the proposed approach outperforms state-of-the-art approaches by reducing total inference delay, edge inference delay and local inference delay by an average of 4.76%, 10.04% and 8.03% in the considered MEC networks.

关键词： Resource management   Optimization   Task analysis   NOMA   Vehicle dynamics   Servers   Energy consumption   Mobile edge computing   non-orthogonal multiple access   resource allocation   Lyapunov optimization   vehicular networks  

摘要： Mobile Edge Computing (MEC) has envisioned to be a promising technology to provide more efficient services for computation-intensive but delay-sensitive onboard mobile services. In this paper, the Non-Orthogonal Multiple Access (NOMA) technology is applied in a vehicular edge computing network, in which vehicular users (VUs) can offload partial computation tasks to MEC servers over wireless channels for remote execution. In this network, an optimization problem for the long-term energy consumption of the system is presented and aims to minimize it by jointly optimizing the Successive Interference Cancellation (SIC) ordering of NOMA, the VUs' transmit power for computation offloading, and computation resource allocation of the MEC server. To deal with the intractable long-term optimization problem, we first transform it into an equivalent instantaneous form based on the Lyapunov optimization theory. Since the transformed problem is still highly non-convex, we further decompose it into the interactive resource allocation and SIC ordering sub-problems. For the resource allocation sub-problem, we exploit its convexity through the transformation and reparameterization, and derive the optimal solution in accordance with the Karush-Kuhn-Tucker (KKT) conditions and the gradient descent algorithm. After that, we propose a low-complexity algorithm by leveraging the Tabu search to obtain the sub-optimal SIC ordering. Simulation results validate the effectiveness of the proposed algorithm and the superiority of NOMA compared to Frequency Division Multiple Access (FDMA).