Abstract
With the development of computer technology, computational-intensive and delay-sensitive applications are emerging endlessly, and they are limited by the computing power and battery life of Smart Mobile Devices (SMDs). Mobile edge computing (MEC) is a computation model with great potential to meet application requirements and alleviate burdens on SMDs through computation offloading. However, device mobility and server status variability in the multi-server and multi-task scenario bring challenges to the computation offloading. To cope with these challenges, we first propose a parallel task offloading model and a small area-based edge offloading scheme in MEC. Then, we formulate the optimization problem to minimize the completion time of all tasks, and transform the problem into a deep reinforcement learning-based offloading scheme by Markov decision approach. Furthermore, we present a deep deterministic policy gradient (DDPG) approach for obtaining the offloading strategy. Experimental results demonstrate that the DDPG- based offloading approach improves long-term performance by at least 19% in ultra-low latency, efficient usage of servers, and frequent mobility of SMDs over traditional strategies.
Highlights
In recent years, the widespread use of Smart Mobile Devices (SMDs) is accelerating the massive growth of computation-intensive and delay-sensitive mobile applications, such as online videos, virtual reality (VR), and augmented reality (AR)
MOBILE EDGE COMPUTING SYSTEM WITH MULTI-SERVER MULTI-SMD AND MULTI-TASK We propose a mobile edge computing system consisting of one Macro Base Station (MBS) and m Small Base Stations (SBSs)
Initialization: Initialize mobile edge computing environment, including the location of base station equipped with edge server, SMDs movement trajectory based on small area and task information generated by SMDs; Randomly initial Critic network Q Sl, al; w and Actor μθ Sl with weights w and θ ; Initialize target network Q Sl, al ; w and μθ Sl with weights w Q ← wQ and θ μ ← θ μ; Initialize replay memory pool R; Iteration: 1: for each round t = 1 to T do 2: Accept initial Mobile Edge Computing (MEC) system state S0; 3: Initialize a random process N for action exploration; 4: for time slots l = 0, . . . , lmax do
Summary
The widespread use of Smart Mobile Devices (SMDs) is accelerating the massive growth of computation-intensive and delay-sensitive mobile applications, such as online videos, virtual reality (VR), and augmented reality (AR). The authors of [10] proposed a high-performance offloading management scheme in the small-cell-networks MEC system to minimize the energy consumption of all user devices through jointly optimizing computation offloading and computation resource allocation (i.e., spectrum, power, and computation resource). 2) Frequent mobility of SMDs: Smart mobile devices are on-the-move, which causes a dynamic communication environment and affects the upload delay in tasks offloading To address these challenges, we design a Deep Deterministic Policy Gradient-based task offloading approach that can achieve the efficient use of edge servers and satisfy the frequent mobility of SMDs in a multi-server multi-SMD and multi-task MEC environment. We incorporate the dynamic communication environment and the variable server status information over continuous time slots into the problem transformation process and solve the offloading scheme through a Deep Deterministic Policy Gradient-based offloading approach.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
