Intelligent Terahertz Medium Access Control (MAC) for Highly Dynamic Airborne Networks

Abstract

The Terahertz (THz) link provides over 100 Gbps of data transmission rate. However, it is extremely vulnerable to obstacles and requires very narrow (highly directional) beams for highly focused energy radiation. The main task of THz medium access control (MAC) protocol is not channel access collision avoidance; instead, it is to coordinate all the one-hop neighbors for scheduled, antenna-aligned line-of-sight communications. This study aims to build an intelligent MAC scheme for highly mobile Terahertz airborne networks (TANs). Our TAN MAC design has three features as follows: (1) <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Spatio-temporal TAN state learning:</i> We will build a predictive network state estimation model through deep learning and generative adversarial network (GAN). Based on the predicted node/link status, all the one-hop neighbors can prepare well for the antenna alignment and THz channel scheduling. (2) <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Accurate, two-level MAC operation control:</i> we propose to use nested deep reinforcement learning (DRL) with outer/inner policy loops for high-/low-level action determination: The outer loop determines the high-level, coarse actions (such as antenna codebook selection); the inner loop determines the low-level, fine actions (such as individual beam control) under the selected high-level action. (3) <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TAN-specific, comprehensive MAC behavior control:</i> Based on the above deep neighbor adaptation (DNA) model, we design a complete TAN MAC protocol that considers the routing context and dynamic network topology. Our simulations demonstrate the smooth, high-rate THz communications in MAC layer with resilient RF links under high node mobility.