A2C Deep Reinforcement Learning-based MEC Network for Offloading and Resource Allocation
Haoran Cui, Dongyu Wang, Qi Lecky Li, Xiaoyu Liu, Hua Lü · 2021 7th International Conference on Computer and Communications (ICCC) · 2021
Nowadays, various neural network models based on Deep Reinforcement Learning (DRL) have been proposed to find the optimal strategy of computation offloading and resource allocation especially in Mobile Edge Computing (MEC). However, due to the complexity and the different types of computing tasks in 5G, the current DRL algorithms are difficult to be applied to complex MEC scenarios directly. This paper applies the normalized characteristic matrix to split the complex network into small network units and proposes a normalized network model for complex network scenarios based on Advantage Actor Critical (A2C) algorithm to find the computation offloading and resource allocation strategy.