A Tchebycheff-Guided MARL-Based Channel Access in IEEE 802.11ah Networks

Le résumé fourni par la source

This paper addresses the challenge of balancing per-station throughput and overall fairness for IEEE 802.11ah uplink scenarios with heterogeneous link conditions and shared-medium contention. We propose a learning-based MAC-layer enhancement that replaces the backoff mechanism with a policy-driven transmission decision, enabling explicit Pareto-guided control of the throughput-fairness trade-off. We formulate the problem as a multi-objective optimization task and propose a Tchebycheff-based Mixed Multi-Agent Deep Deterministic Policy Gradient (TM-MADDPG) framework. By modeling the channel access problem as a mixed cooperative-competitive task, the framework enables decentralized agents to learn coordinated policies that account for local and global objectives. To capture the structure of competing objectives, we design a Tchebycheff-guided dual-critic architecture, which supports adaptive and efficient policy learning. Extensive evaluations show that TM-MADDPG achieves superior trade-offs between throughput and fairness compared to existing baselines.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.