Task Scheduling for Smart City Applications Based on Multi-Server Mobile Edge Computing

• Main Authors: Yiqin Deng, Zhigang Chen, Xin Yao, Shahzad Hassan, Jia Wu
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: Task scheduling; smart city; mobile edge computing; Internet of Vehicle; alternating direction method of multipliers (ADMM) algorithm
• Online Access: https://ieeexplore.ieee.org/document/8618300/

The smart city is increasingly gaining worldwide attention. It has the potential to improve the quality of life in convenience, at work, and in safety, among many others' utilizations. Nevertheless, some of the emerging applications in the smart city are computation-intensive and time-sensitive, such as real-time vision processing applications used for public safety and the virtual reality classroom application. Both of them are hard to handle due to the quick turnaround requirements of ultra-short time and large amounts of computation that are necessary. Fortunately, the abundant resource of the Internet of Vehicles (IoV) can help to address this issue and improve the development of the smart city. In this paper, we focus on the problem that how to schedule tasks for these computation-intensive and time-sensitive smart city applications with the assistance of IoV based on multi-server mobile edge computing. Task scheduling is a critical issue due to the limited computational power, storage, and energy of mobile devices. To handle tasks from the aforementioned applications in the shortest time, this paper introduces a cooperative strategy for IoV and formulates an optimization problem to minimize the completion time with a specified cost. Furthermore, we develop four evolving variants based on the alternating direction method of multipliers (ADMM) algorithm to solve the proposed problem: variable splitting ADMM, Gauss-Seidel ADMM, distributed Jacobi ADMM, and distributed improved Jacobi (DIJ)-ADMM algorithms. These algorithms incorporate an augmented Lagrangian function into the original objective function and divide the large problem into two sub-problems to iteratively solve each sub-problem. The theoretical analysis and simulation results show that the proposed algorithms have a better performance than the existing algorithms. In addition, the DIJ-ADMM algorithm demonstrates optimal performance, and it converges after approximately ten iterations and improves the task completion time and offloaded tasks by 89% and 40%, respectively.