Dynamic Resource Allocation Scheme and Deep Deterministic Policy Gradient-Based Mobile Edge Computing Slices System

• Main Authors: Yin Ren, Aihuang Guo, Chunlin Song, Yidan Xing
• Format: Article
• Language: English
• Published: IEEE 2021-01-01
• Series: IEEE Access
• Subjects: Deep deterministic policy gradient; mobile communication; mobile edge computing; network slicing; resource allocation
• Online Access: https://ieeexplore.ieee.org/document/9452161/

The development of multi-industry compatibility and the coexistence of multiple services and multiple functional communication networks will cause rapid growth in mobile communication system traffic. Users will have increasingly strict requirements for quality of service (QoS), e.g., a high rate, low latency, and low energy consumption. To address these problems, it is helpful to combine network slicing and mobile edge computing (MEC) to provide customized networks while reducing the service processing time. Due to the uncertainty of user requests and the environment, reasonable resource allocation is always particularly challenging. A novel dynamic resource allocation scheme for MEC slice systems, which formulates resource allocation and computation offloading issues as an optimization problem subject to the latency and rate, is proposed. Based on the dynamics of the slice requirements, quantity, and service time, the proposed problem is converted to a Markov decision process (MDP), and a state, action, and reward function are proposed. By exploiting the deep deterministic policy gradient (DDPG) algorithm, the wireless resources and computing resources are configured dynamically according to the requirements of different types of slices to maximize the revenue of the network operator. The simulation results demonstrate the influence of the slice arrival rate and total resources on the allocation policy. Compared with other schemes, the proposed scheme can provide a more effective performance when resources are scarce.