Yeongjun Seok, Hyun-kyo Lim, Ihsan Ullah, Youn-Hee Han,

Journal of Advanced Technology Research, Vol. 7, No. 2, pp. 13-16, Dec. 2022
10.11111/JATR.2022.7.2.013, Full Text:
Keywords: IoT, 5G network, Partial Task Off-Loading, Resource Consumption, Delay time, Edge Cloud Computing Systems, Hierarchical Deep Reinforcement Learning, Task Allocation Optimization
Abstract

As IoT and 5G network services continue to evolve, partial task offloading is becoming an important solution to meet energy and delay requirements on resource-constrained local devices. However, determining the offloading ratio in partial task offloading is a combinatorial optimization problem, which is difficult to solve. In this paper, we propose using hierarchical deep reinforcement learning in edge cloud computing systems to solve this combinatorial optimization problem for task offloading. We also define a Markov Decision Process (MDP) and objective function for the hierarchical deep reinforcement learning-based task allocation optimization problem. Each layer in the hierarchical deep reinforcement learning encourages the use of different algorithms, and simulation results show that the proposed method is superior in terms of resource consumption and delay time compared to other methods.

Statistics
Cite this article