In the forthcoming era of 6G, the deployment of dense and diverse wireless networks equipped with edge servers makes parallel offloading a crucial technology for multiple access edge computing. However, the effectiven...
详细信息
In the forthcoming era of 6G, the deployment of dense and diverse wireless networks equipped with edge servers makes parallel offloading a crucial technology for multiple access edge computing. However, the effectiveness of parallel offloading is constrained by heterogeneous edge servers (HES) and heterogeneous wireless networks (HWN). Existing solutions often assume that the link state of the wireless network and the computing resources of the edge server are either known or can be estimated as prior knowledge, which is insufficient, especially for HWN. To address this challenge, we propose a novel online learning parallel offloading (OLPO) framework aimed at jointly selecting multiple radio networks and multiple edge servers (MRMS) by learning the unknown and stochastic metrics of wireless links and edge servers. Specifically, we design and compare three OLPO algorithms using multi -armed bandits (MAB) and construct a comparison matrix with the analytic hierarchy process (AHP). Our experimental results demonstrate the effectiveness of the proposed OLPO algorithms and establish their applicability for various traffic types, considering metrics such as delay, delay jitter, packet loss, rate, and energy delay product.
Recent years have witnessed the unprecedented performance of convolutional networks in image super-resolution (SR). SR involves upscaling a single low-resolution image to meet application-specific image quality demand...
详细信息
ISBN:
(纸本)9798350383515;9798350383508
Recent years have witnessed the unprecedented performance of convolutional networks in image super-resolution (SR). SR involves upscaling a single low-resolution image to meet application-specific image quality demands, making it vital for mobile devices. However, the excessive computational and memory requirements of SR tasks pose a challenge in mapping SR networks on a single resource-constrained mobile device, especially for an ultra-high target resolution. This work presents TileSR, a novel framework for efficient image SR through tile-granular parallel offloading upon multiple collaborative mobile devices. In particular, for an incoming image, TileSR first uniformly divides it into multiple tiles and selects the top-K tiles with the highest upscaling difficulty (quantified by mPV). Then, we propose a tile scheduling algorithm based on multi-agent multi-armed bandit, which attains the accurate offload reward through the exploration phase, derives the tile packing decision based on the reward estimates, and exploits this decision to schedule the selected tiles. We have implemented TileSR fully based on COTS hardware, and the experimental results demonstrate that TileSR reduces the response latency by 17.77-82.2% while improving the image quality by 2.38-10.57% compared to other alternatives.
Mobile edge computing (MEC) enabled artificial intelligence-generated content (AIGC) has garnered significant attention. To support AIGC metaverse applications within MEC, it is effective to offload computation tasks,...
详细信息
ISBN:
(纸本)9798350377675;9798350377682
Mobile edge computing (MEC) enabled artificial intelligence-generated content (AIGC) has garnered significant attention. To support AIGC metaverse applications within MEC, it is effective to offload computation tasks, particularly those involving neural networks AIGC, from mobile devices to edge clouds. Existing solutions typically assume the availability of a dedicated and powerful edge server for each user, which can handle the entire AIGC service offloading. However, in practical scenarios, the availability of such dedicated and powerful servers may be limited, necessitating the utilization of less capable alternatives. Thus, we propose the delay-aware parallel offloading AIGC framework which partitions multi-modal content and offloads partial diffusion tasks to multiple servers. The proposed scheme accelerates mobile deep vision multi-modal metaverse applications by leveraging parallel offloading for any server provisioning. Our framework integrates recurrent region proposal prediction algorithm with delay-aware multi-modal parallel diffusion offloading scheme, optimizing communication and computing resources while minimizing delay in wireless environments. Simulation results show that our approach can reduce delay by 20% compared to conventional algorithms.
Vehicle edge computing (VEC) enhances the distributed task processing capability within intelligent vehicle- infrastructure cooperative systems (i-VICS) by deploying servers at the network edge. However, the prolifera...
详细信息
Vehicle edge computing (VEC) enhances the distributed task processing capability within intelligent vehicle- infrastructure cooperative systems (i-VICS) by deploying servers at the network edge. However, the proliferation of onboard sensors and the continual emergence of new applications have exacerbated the inadequacy of wireless spectrum resources and edge server resources, while the high mobility of vehicles reduces reliability in task processing, resulting in increased communication and task processing delays. To address these challenges, we propose a mobile-aware Many-to-Many parallel (MTMP) offloading scheme that integrates: a) millimeter- wave (mmWave) and cellular vehicle-to-everything (C-V2X) to mitigate excessive communication delays;and b) leveraging the underutilized resources of surrounding vehicles and parallel offloading to mitigate excessive task processing delays. To minimize the average completion delay of all tasks, this paper formulates the objective as a min-max optimization problem and solves it using the maximum entropy method (MEM), the Lagrange multiplier method, and an iterative algorithm. Extensive experimental results demonstrate the superior performance of the proposed scheme in comparison with other baseline algorithms. Specifically, our proposal achieves a 47 % reduction in task completion delay under optimal conditions, a 31.3 % increase in task completion rate, and a 30 % decrease in program runtime compared to the worst-performing algorithm.
As mobile devices continuously generate streams of images and videos, intelligent mobile vision applications are rapidly emerging. An ideal object detection system for mobile vision applications should be accurate and...
详细信息
As mobile devices continuously generate streams of images and videos, intelligent mobile vision applications are rapidly emerging. An ideal object detection system for mobile vision applications should be accurate and real-time. Nevertheless, it is non-trivial to achieve these goals utilizing resource-constrained mobile devices. In this work, we propose an efficient and robust intelligent mobile vision system AREdge via small object aware parallel offloading. We find that the detection performance of small objects is a core factor that affects detection accuracy. To tackle this, we design a local lightweight DNN model that runs on mobile devices to detect big objects fast and identify the regions of interest (RoIs) that may have small objects. These areas are then cropped and offloaded to multiple edge servers for more accurate detection based on complex and large-scale DNN models. To further improve the performance, we propose a dynamic area-aware parallel offloading scheme for fine-grained parallel execution on multiple edge servers. Experimental results show that the accuracy of AREdge is 214.27% higher than that of the local detection in small objects. It also reduces the detection latency by 20.68% on average over the offloading method based on full images and well-used object detection models.
Although mobile health monitoring where mobile sensors continuously gather, process, and update sensor readings (e. g. vital signals) from patient's sensors is emerging, little effort has been investigated in an e...
详细信息
Although mobile health monitoring where mobile sensors continuously gather, process, and update sensor readings (e. g. vital signals) from patient's sensors is emerging, little effort has been investigated in an energy-efficient management of sensor information gathering and processing. Mobile health monitoring with the focus of energy consumption may instead be holistically analyzed and systematically designed as a global solution to optimization subproblems. This paper presents an attempt to decompose the very complex mobile health monitoring system whose layer in the system corresponds to decomposed subproblems, and interfaces between them are quantified as functions of the optimization variables in order to orchestrate the subproblems. We propose a distributed and energy-saving mobile health platform, called mHealthMon where mobile users publish/access sensor data via a cloud computing-based distributed P2P overlay network. The key objective is to satisfy the mobile health monitoring application's quality of service requirements by modeling each subsystem: mobile clients with medical sensors, wireless network medium, and distributed cloud services. By simulations based on experimental data, we present the proposed system can achieve up to 10.1 times more energy-efficient and 20.2 times faster compared to a standalone mobile health monitoring application, in various mobile health monitoring scenarios applying a realistic mobility model.
In this paper, we consider parallel and sequential task offloading to multiple mobile edge computing servers. The task consists of a set of inter-dependent sub-tasks, which are scheduled to servers to minimize both of...
详细信息
In this paper, we consider parallel and sequential task offloading to multiple mobile edge computing servers. The task consists of a set of inter-dependent sub-tasks, which are scheduled to servers to minimize both offloading latency and failure probability. Two algorithms are proposed to solve the scheduling problem, which are based on genetic algorithm and conflict graph models, respectively. Simulation results show that these algorithms provide performance close to the optimal solution, which is obtained through exhaustive search. Furthermore, although parallel offloading uses orthogonal channels, results demonstrate that the sequential offloading yields a reduced offloading failure probability when compared to the parallel offloading. On the other hand, parallel offloading provides less latency. However, as the dependency among sub-tasks increases, the latency gap between parallel and sequential schemes decreases.
暂无评论