Model-Free Dynamic Traffic Steering for Multi-Link Operation in IEEE 802.11be

The IEEE 802.11be extremely high throughput (EHT) amendment (commercially known as Wi-Fi 7) aspires to further improve the network performance on data rate, latency, and reliability. As a core capability in the IEEE 802.11be EHT amendment, multi-link operation (MLO) defines the multi-link device (MLD) architecture with the simultaneous transmit and receive (STR) functionality for both station (STA) and access point (AP) to achieve concurrent operations across multiple available links. To unleash the full potential of MLO, it is essential that the AP MLD create proper dynamic traffic steering for STA MLDs. Therefore, in this paper, we consider MLO with the STR functionality and propose two application-agnostic model-free dynamic traffic steering methods. Specifically, we develop an adaptive scoring heuristic algorithm based on a normalized weighted score featuring lightweight computational complexity and a deep reinforcement learning (DRL) approach based on standard soft actor-critic (SAC) featuring robust adaptability. Simulation results demonstrate the effectiveness of the proposed model-free dynamic traffic steering methods in terms of reward convergence and average network throughput. Besides, we investigate the load balancing ability of the proposed model-free dynamic traffic steering methods.