Evidence Hub

Physics-Informed Multi AI-Agent Localization-Assisted THz--OWC L3-Aware MPTCP-Inspired Dual-Subflow Cross-Layer Framework for Extreme-Band 6G Networks

Ogbodo EU, Mendes LL

2026

Abstract

Abstract Terahertz (THz) and optical wireless communication (OWC) are complementary technologies for sixth-generation (6G) networks, but their practical integration is challenged by severe THz propagation loss and blockage, OWC geometric and field-of-view constraints, mobility, and localization uncertainty. These effects influence not only physical-layer link quality but also Layer~3 (L3) path selection and transport-layer traffic distribution. Existing studies commonly address these functions separately, leaving a gap in reproducible cross-layer frameworks that jointly connect heterogeneous physical states and localization information with L3 path intelligence, multipath transport adaptation, and packet-level network performance. This work proposes a physics-informed (PI) multi-agent artificial intelligence (AI) framework with a Multipath TCP (MPTCP)-inspired THz--OWC dual-subflow architecture. The intelligent PI-DDQN--GAT--SAC strategy combines a Double Deep Q-Network (DDQN) for physics-constrained access and stability decisions, a Graph Attention Network (GAT) for heterogeneous THz--OWC L3 path ranking and selection, and Soft Actor-Critic (SAC) for continuous transport-layer subflow allocation. A complementary PI-Static 50/50 strategy provides a deterministic non-learning dual-path reference for persistent parallel utilization. The framework combines Sionna ray-tracing states for the THz--OWC environment, with the OWC branch additionally anchored to calibrated…

Read the paper

Related studies

Go deeper in Evidence Hub

Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.

Open this study Sign up free