MAC and Routing Protocols Design in Multi-Hop Terahertz Networks

Abstract

Communication at Terahertz (THz) frequencies is strongly affected by high path loss and blockage, making multi-hop strategies essential, especially in obstacle-dense scenarios such as Industrial Internet of Things (IIoT) environments. We propose a joint Medium Access Control (MAC) and Network (NET) layer design, centered on a distributed Table-Based (TB) routing protocol that removes control-plane signaling by leveraging only user-plane data transmissions for route discovery and maintenance. The MAC layer employs a simple contention-based unslotted Aloha protocol, whose fully distributed nature makes it well-suited to THz IIoT multi-hop networks while avoiding the overhead of more complex contention-free schemes. The performance of TB is evaluated against two well-established NET-layer benchmarks: a Table-Less (TL) broadcast-forwarding scheme and the reactive Ad hoc On-Demand Distance Vector (AODV) protocol. Results show that AODV suffers from fairness degradation due to high control overhead and asynchronous route discovery, while TL can achieve higher throughput under congestion, but incurs higher latency due to suboptimal path selection. Overall, TB provides the best trade-off between fairness, throughput and latency. To further minimize the necessity of hyperparameter tuning, we also extend TB with a decentralized Multi-Agent Deep Reinforcement Learning (MADRL) mechanism that enables autonomous MAC-layer parameter adaptation. Simulations in a realistic IIoT scenario demonstrate that MADRL-TB delivers competitive performance compared to optimal static configurations while ensuring high robustness against parameter variations.