Multi-Armed Bandits for Spectrum Allocation in Multi-Agent Channel Bonding WLANs

While dynamic channel bonding (DCB) is proven to boost the capacity of wireless local area networks (WLANs) by adapting bandwidth on a per-frame basis, its performance tied primary and secondary selection. Unfortunately, in uncoordinated high-density deployments where multiple basic service sets (BSSs) may potentially overlap, hand-crafted spectrum management techniques perform poorly given complex hidden/exposed nodes interactions. To cope with such challenging Wi-Fi environments, this paper, we first identify machine learning (ML) approaches applicable problem at hand justify why model-free RL suits it most.We then design complete framework call into question whether use algorithms helps quest for rapid realistic scenarios. Through extensive simulations, derive that stateless form lightweight multi-armed-bandits (MABs) an efficient solution adaptation avoiding definition broad and/or meaningless states. In contrast most current trends, envision MABs as appropriate alternative cumbersome slowly convergent methods Q-learning, especially, deep reinforcement learning.