Cooperative Dynamic Voltage Scaling and Radio Resource Allocation for Energy-Efficient Multiuser Mobile Edge Computing.

Yanting Wang,Min Sheng,Xijun Wang,Jiandong Li

DOI: https://doi.org/10.1109/icc.2018.8422629

2018-01-01

Abstract:Mobile-Edge Computing (MEC) could relieve computing pressure of resource-constrained Smart Mobile Devices (SMDs) by offloading computation-intensive tasks to nearby/MEC server. However, how to achieve energy efficient computation offloading for SMDs under application-dependent latency constraints remains challenging in multiuser MEC systems. Specifically, the optimal system operations are not only inner- coupled for each SMD due to parallel local and cloud execution, but also inter-coupled among SMDs due to competition for limited radio resource. Additionally, the inner- and inter-coupling influence each other, which further complicates multiuser offloading strategy design. In this paper, we address such a challenge by jointly optimizing computational speed of SMDs via Dynamic Voltage Scaling (DVS) technology, subcarrier allocation, transmit power per subcarrier, data size sent per subcarrier, and offloading ratio, to minimize weighted sum of mobile energy consumption, resulting in a mixed-integer optimization problem. To tackle this NP-hard problem, we propose a fast-convergent suboptimal algorithm with the Lagrangian dual decomposition. Additionally, simulation results verify that the algorithm converges fast and significantly outperforms existing schemes in energy consumption reduction. Meanwhile, we discover that given latency mean, total mobile energy consumption remains stable or increases with the variance of latency requirements, which could direct admission control in practice.