关键词： Computational modeling   Adaptation models   Training   Data models   Transformers   Deep learning   Routing   Costs   Biological system modeling   Performance evaluation   Mixture-of-experts   continual knowledge adaption   mobile edge computing  

摘要： Continual machine learning in the context of limited computational resources and data availability is critical in the connected digital world. Current intelligent applications predominantly rely on deep learning models requiring labor/computation-intensive training. These models often struggle to adapt effectively to new data while preserving performance on previously learned knowledge. In this paper, we introduce a lightweight method for continual knowledge adaptation that can address these challenges. To prevent disruption of the existing services, we propose a Mixture-of-Experts (MoE) adapter that integrates seamlessly with the existing vision model to encode new data. The weights of the original model are kept fixed during the adaptation process, ensuring the preservation of previously learned knowledge. The MoE technique enables scaling up the parameters of the adapter while maintaining a relatively low computation, making it fit for constrained devices in mobile computation scenarios. Furthermore, to enhance learning efficiency and accelerate convergence with new data, we implement a knowledge fusion mechanism that facilitates interaction between the existing knowledge and the information extracted from new data. The timing of employing the fusion module is further investigated. We find that it is conducive in scenarios where the task's performance requirements are enhanced. The MoE adapter and knowledge fusion module are integrated at each stage with minimal trainable parameters, efficiently optimizing resource usage. Extensive experiments and ablation studies validate the effectiveness of the proposed method. Specifically, the proposed method prevents an accuracy drop of 43.02% on the previous data compared to the continual train method, while achieving an accuracy of 44.81% on the new data, which is even 0.34% higher than fully training a new model.

关键词： Communication   Computing   Cost reduction   Caching   Control   Latency minimization   Mobile edge computing   Optimization   UAVs  

摘要： Unmanned aerial vehicles (UAVs) enable a mobile edge computing (MEC) paradigm with reduced latency by bringing computational resources closer to the network edge. However, UAV-MEC servers have less computation and caching resources than ground base stations (BSs). The management of communication and control resources is crucial to coordinate communication, computing, and caching due to the involvement of aerial networks. Thus, managing joint caching, communication, computing, and control (4C) resources is vital in UAV-assisted MEC networks. To address these challenges, we developed a computational model for efficient resource management to reduce the linear combination of network cost and latency under constrained caching, computing, and offloading. We used binary decision variables for the allocation of computational and offloading resources. The formulated problem is a binary linear programming problem incorporating binary decision variables and linear constraints. We propose an interior point method-based heuristic to obtain a sub-optimal solution with low complexity. Simulation results demonstrate the effectiveness of our proposed approach compared to the branch and bound algorithm.

关键词： Conformalized Graph Neural Network   Embedded system   Mobile edge computing   Narwhal Optimizer   Power consumption  

摘要： This work presents a novel Multiple Object Visual Tracking (MOT) approach integrating Conformalized Graph Neural Network (CF-GNN) with Narwhal Optimizer (NO) for enhanced tracking accuracy and efficiency in IoT and Mobile Edge Computing (MEC) applications. Addressing challenges like noise, occlusion, and high computational complexity, the method employs Robust Double-Weighted Guided Image Filter (RDWGIF) for noise reduction, Improved YOLO v5 with Residual Group Attention Network (RGA-Net) for detection, and DualBranch Geometric Attention Network (DBGANet) for optimized feature extraction. Achieving 99.9% accuracy, 99.83% precision, 99.85% recall, and 99.88% F1-score, along with reduced power consumption (4.44 PC) and a 99% AUC, this method outperforms existing tracking solutions, making it highly effective for resourceconstrained edge environments. While the proposed method demonstrates exceptional performance on the MOT16 dataset, its generalizability to more diverse or complex real-world scenarios and datasets requires further investigation, and scalability may be challenged in ultra-constrained hardware environments.

关键词： Anomaly detection   Deep learning   Differential privacy   Generative adversarial networks   Healthcare data security   Mobile edge computing   Privacy preservation   Reinforcement learning   Smart healthcare systems   Stochastic vector autoencoder  

摘要： Data have always been a top goal in achieving intelligent healthcare in a smart city, especially with the rapidly growing deployment of the Internet of Things and connectivity. The significant challenges that remain in the evolving landscape of mobile edge computing and smart healthcare are ensuring data privacy and detecting anomalies in sensitive medical information. The research proposes a novel technique, namely a lightweight differential extreme gradient reinforcement neural network with stochastic vector autoencoder-based adversarial Bayes convolutional regression (LDExGRNN_SVA-AdBCR), which combines multiple machine learning strategies to preserve privacy and detect anomalies in healthcare data. In this approach, a lightweight differential extreme gradient reinforcement neural network that integrates differential privacy techniques with XGBoost and deep Q-learning reinforcement strategies is utilized to achieve privacy preservation. Anomaly prediction is carried out using a stochastic vector autoencoder-based adversarial Bayes convolutional regression model that combines a stochastic vector autoencoder with multiple classifiers and regressors, including Na & iuml;ve Bayes, support vector machines, generative adversarial networks, and vector autoregression. To enhance detection robustness, outputs from multiple classifiers and the GAN discriminator score are integrated using a majority voting mechanism. This ensemble fusion boosts reliability in detecting diverse anomalies across healthcare datasets. The experimental analysis is evaluated on various healthcare datasets in terms of privacy rate (98.1%), prediction accuracy (98.75%), scalability (98.3%), and F1-score (98.34%). Overall, the LDExGRNN_SVA-AdBCR method provides a scalable solution for detecting anomalies in MEC environments and securing healthcare data.

关键词： Mobile edge computing   Task offloading   Temporal dependency   Deep reinforcement learning  

摘要： In Mobile Edge Computing (MEC) environments, offloading computational tasks to edge devices can reduce delay and energy consumption, thereby enhancing the user experience. In practical application scenarios, some complex computational tasks often exhibit temporal dependencies, meaning that the execution of one task relies on the results of previously executed tasks. When offloading such tasks, it is necessary to consider both the execution order and the data dependencies between tasks, which increases the complexity of modeling and solving the task offloading problem. Therefore, this paper constructs a task offloading model with temporal dependencies by comprehensively considering the communication cost between tasks, offloading delay, and energy consumption. Additionally, a Deep Q-Network (DQN)-based algorithm is proposed to address the temporal dependency task offloading model. The state space, action space, and optimized reward function tailored to the temporal dependency task offloading scenario are designed to accelerate the convergence of the algorithm. Finally, experimental results in real datasets demonstrate that the method proposed in this paper can achieve more consumption-effective offloading decisions while maintaining strong performance in various temporal dependency task offloading scenarios.

关键词： echo tracking optimization   healthcare system   mobile edge computing   quality of service   wireless body area network  

摘要： In the context of the Internet of Medical Things (IoMT), the rapid expansion of wearable medical devices and healthcare data presents tremendous challenges related to the improved Quality of Service (QoS) and computing task offloading for Smart healthcare systems. Further, the Mobile Edge Computing (MEC)-enabled healthcare systems, which allow computation offloading to edge servers nearby, are attracting great attention as a result of the extraordinary development in Wireless Body Area Network (WBAN) users and applications based on 5G. However, the existing systems in MEC-enabled WBAN-based healthcare systems produce too many control frames while transmitting data, resulting in increased latency, energy wastage, and a lack of flexibility. Therefore, this research aims to design a routing algorithm in WBAN that efficiently allocates resources and consumes less energy utilizing the Echo Tracking Optimization-based MEC-enabled WBAN systems. Specifically, the proposed model provides ultra-reliable data transfer and processing with extremely low latency and energy consumption to meet the demands of healthcare services and applications. More effectively, the proposed approach exploits the Echo Tracking Optimization (ETO) that handles the resource allocation and enhances the QoS by addressing the problem of selection of the target tasks on analyzing the medical criticality, highest relative computing capacity, and energy constraints for effective task offloading. Compared to the other existing techniques, the proposed ETO-QoS aware scheduling effectively lowers latency and energy consumption while increasing throughput and overall WBAN utilization by reporting a delay of 0.102 ms, energy loss of 7.523 J, packet loss of 95, and throughput of 0.723 Kbps outperforming the other existing techniques.

关键词： Vehicular edge computing   latency   communication   computation   storage   mobile edge computing  

摘要： Low-latency Vehicular Edge Computing (VEC) applications require an efficient VEC resource allocation considering all components contributing to application latency, i.e., computation, communication, and storage. While the optimization of communication and computation resources is broadly addressed in literature, storage, a significant source of latency in the computing stack, is often ignored in existing works. Thus, in this paper, we optimize the communication and computation resources together with the storage resources to minimize the latency of VEC applications. The problem of jointly minimizing communication, computation, and storage latency under practical constraints is NP-hard. Hence, we employ dual decomposition and Lagrangian relaxation to achieve a computationally viable solution for the joint communication, computing, and storage resource allocation to VEC applications. To this end, we define a dual problem of the assignment of VEC applications to base stations. This problem corresponds to the perfect matching problem in a weighted bipartite graph and optimally solvable by algorithms with polynomial computation complexity. Then, as the solution to the dual problem may violate some constraints of the main resource allocation problem, we find a feasible solution to the main resource allocation problem using Lagrangian relaxation. We show that the joint optimization of all three aspects, i.e., communication, computation, and storage, reduces the overall offloading latency up to 60% compared to state-of-the-art works.

关键词： integrated energy system   data monitoring   network slicing   mobile edge computing   resource allocation  

摘要： With the continuous growth in both volume and variety of monitoring services in integrated energy systems, the disparities between time scales and tasks of heterogeneous energy flow data monitoring pose significant challenges to rational resource allocation and efficient data transmission. To address these challenges, a monitoring resource slicing allocation method with low-cost optimization is presented for energy flow monitoring services in integrated energy systems. Firstly, a dynamic network slicing method for heterogeneous energy flow is proposed, which realizes rational resource allocation for diverse monitoring tasks and time scales through an integrate-then-slice approach. Secondly, a data transmission strategy with a cost representation method for network slicing is proposed. By establishing precise modeling of the complete data monitoring process within each slice, it solves the quantitative problem of data monitoring costs. Thirdly, an adaptive monitoring resource slice allocation algorithm is proposed, which addresses the cost optimization problem in data monitoring under slicing modes through optimized allocation of both intra-slice and inter-slice resources. Finally, tests are conducted on an integrated energy system in China. The results demonstrate that the proposed method successfully achieves data monitoring of heterogeneous energy flows across multiple time scales, while significantly reducing the data monitoring costs.

关键词： Composite service provisioning   Service coverage problem   Mobile edge computing   DQN   MLP   Quality of service  

摘要： Mobile Edge Computing (MEC) facilitates rapid data handling by processing real-time data in proximity to users. However, due to constrained resources of MEC servers and strict requirements of end users, the need for optimal server placement, and associated provisioning of composite services are required to enable better resource management of the MEC servers while enabling higher Quality of Services (QoS) to end users. In this paper, we propose a framework to enable proper provisioning of the end users' composite services considering their mobility, while facilitating placement of MEC servers and enabling load balancing to enhance QoS. We formulated the composite service placement in MEC as a Service Coverage Problem (SCP) and solved it using our Adaptive Composite Service Deployment (ASDMD) using Multilayer Perceptron (MLP)-Based Deep Q-Network (DQN). Following that, we formulated the MEC server placement as a constraint-based P-Median problem and solved it using the Constraint-Based Knapsack Heuristic (CB-KH) algorithm. To ensure higher QoS and reduction of congested servers, we used a load balancing technique to iteratively monitor the loads of the servers and transfer the congested servers' users to the most appropriate alternatives. Experiments and simulations based on real-world datasets showcase the significance of our model in better management of resources of MEC servers while ensuring proper processing of data and enhanced QoS for end users. The results indicate a minimum of 4.8% improvement in the reduction of composite services deployed, which improves resource usage, enhancing the user coverage by a minimum of 19.8%, to serve a greater number of users, and enabling a minimum of 12.7% reduction in the average latency, enhancing the user satisfaction, compared to the state-of-the-art methods.

关键词： Cuckoo search optimization   Edge servers   Hashing technology   Internet of Things   Mobile edge computing   Quantum computing   Unmanned aerial vehicles  

摘要： Efficient allocation of Internet of Things (IoT) devices and timely real-time data processing in disaster-affected areas pose significant challenges due to energy constraints, limited infrastructure, and the urgent need for rapid response. Unmanned Aerial Vehicles (UAVs), when integrated with Edge Computing (EC), offer a promising solution for collecting and processing data. In this paper, we propose a novel hashing-based quantum-inspired cuckoo search optimization (QICSO) algorithm to address key challenges in disaster management, including user allocation, processor allocation, energy consumption, and time delay factors. The QICSO algorithm is designed to optimize resource utilization by addressing key objectives such as minimizing energy consumption, reducing data processing delays, and improving the allocation rates of users and processors. A novel fitness function is formulated by considering these factors to ensure efficient disaster response. A novel representation of the quantum cuckoo and qubit has been developed, where quantum cuckoos are encoded and decoded using a novel hashing technique. The fitness function incorporates four objectives: user allocation rate, processor allocation rate, energy consumption, and delay. The proposed approach is simulated across multiple scenarios against the latest existing algorithms. Results demonstrate that the proposed algorithm significantly outperforms the compared methods. Additionally, statistical analyses, including analysis of variance (ANOVA) and the Friedman test, have been conducted. The Taguchi parametric statistical technique is employed to further evaluate performance. The simulation results indicate that QICSO surpassed existing approaches in terms of energy consumption by 18.3% and delay by 3.5%.