Coordination among multiple access points (APs) is integral to IEEE 802.11bn (Wi-Fi 8) for managing contention in dense networks. This letter explores the benefits of Coordinated Spatial Reuse (C-SR) and proposes the use of reinforcement learning to optimize C-SR group selection. We develop a hierarchical multi-armed bandit (MAB) framework that efficiently selects APs for simultaneous transmissions across various network topologies, demonstrating reinforcement learning’s promise in Wi-Fi settings. Among several MAB algorithms studied, we identify the upper confidence bound (UCB) as particularly effective, offering rapid convergence, adaptability to changes, and sustained performance.

The densification of Wi-Fi deployments means that fully distributed random channel access is no longer sufficient for high and predictable performance. Therefore, the upcoming IEEE 802.11bn amendment introduces multi-access point coordination (MAPC) methods. This paper addresses a variant of MAPC called coordinated spatial reuse (C-SR), where devices transmit simultaneously on the same channel, with the power adjusted to minimize interference. The C-SR scheduling problem is selecting which devices transmit concurrently and with what settings. We provide a theoretical upper bound model, optimized for either throughput or fairness, which finds the best possible transmission schedule using mixed-integer linear programming. Then, a practical, probing-based approach is proposed which uses multi-armed bandits (MABs), a type of reinforcement learning, to solve the C-SR scheduling problem. We validate both classical (flat) MAB and hierarchical MAB (H-MAB) schemes with simulations and in a testbed. Using H-MABs for C-SR improves aggregate throughput over legacy IEEE 802.11 (on average by 80% in random scenarios), without reducing the number of transmission opportunities per station. Finally, our framework is lightweight and ready for implementation in Wi-Fi devices.