Multi-agent DRL for edge computing: A real-time proportional compute offloading

Kunkun Jia, Hui Xia, Rui Zhang, Yue Sun, Kai Wang

Abstract: Highlights•To reduce mobile device runtime and energy use, we must decide if offloading tasks to Mobile Edge Servers is beneficial and how to allocate resources, considering the device’s computational power and bandwidth. This is a long-term optimization problem.•Considering the real-time and large-scale data processing needs of the Industrial Internet of Things (IIoT), we propose a lightweight task offloading framework. This framework addresses issues like device selfishness, limited bandwidth, and low energy storage, making it ideal for IIoT deployment.•We propose MASTD3 and according to the randomness of input state information, we correspondingly construct a network structure to address how to determine offloading decisions and proportions. MASTD3 considers the long-term optimization of task offloading to balance real-time performance and energy efficiency.