Multi-agent DRL-based Task Offloading in Hierarchical HAP-LAP Networks

Future wireless networks promise to offer ubiq-uity connection to numerous Internet of Things devices with various demands. Aerial access networks that combine satellite and unmanned aerial vehicle communications with mobile edge computing can offer a unique opportunity to address such demands promptly. In this paper, we investigate a hierarchical aerial network in which task offloading for ground devices is supported collaboratively by high and low altitude platforms. In addition, non-orthogonal multiple access is employed to enhance the transmission rate. We transform the problem into a partially observable Markov decision process and use a multi-agent deep reinforcement learning method to find a solution to minimize energy consumption and task execution delay of all devices.