Bandit-Based Multiple Access Approach for Multi-Link Operation in Heterogeneous Dynamic Networks

The Multi-Link Operation (MLO) is a key technique in the upcoming IEEE 802.11be Extremely High Throughput standard, also known as Wi-Fi 7. MLO enables Multi-Link Devices (MLDs) to utilize multiple communication links simultaneously, thereby enhancing throughput. However, research has shown that MLO performance deteriorates in heterogeneous networks coexisting with Single-Link Devices (SLDs), especially in asymmetric scenarios. In this paper, we address the multiple access problem for MLDs in both homogeneous and heterogeneous dynamic uplink networks. We propose a Multi-Player Combinatorial Upper Confidence Bound (MP-CUCB) algorithm designed for Simultaneous Transmit and Receive (STR) MLDs to better exploit the potential of MLO. The MP-CUCB algorithm enables each MLD to independently make access decisions to improve throughput while guaranteeing the fairness requirements of SLDs. Simulation results demonstrate that the MP-CUCB algorithm can: 1) better utilize the potential of MLO, achieving a 20% improvement compared to the standard CSMA scheme while ensuring fairness requirements for both MLDs and SLDs; 2) address the severe performance degradation issue in asymmetric heterogeneous scenarios, achieving a 33% improvement compared to the standard CSMA; 3) effectively adapt to dynamic networks with new accessing nodes, achieving short convergence times.