A Bayesian Optimization Algorithm to Improve the Spatial Reuse in the Next-Generation WLANs

The dense deployment of wireless nodes in the next generation of wireless local area networks (WLANs) poses a potential threat to network performance. Enhancing spatial reuse (SR) in WLANs can effectively address this issue. Dynamic clear channel assessment (CCA) threshold and transmit power control are crucial techniques to improve SR. This paper formulates the SR problem as a multi-armed bandit problem. A Bayesian optimization online learning algorithm with Gaussian process is proposed to optimize CCA thresholds and transmit power jointly. Finally, the proposed algorithm is compared with the default configuration and Thompson sampling algorithm across four performance metrics with the NS-3 simulator. The results demonstrate that our algorithm can significantly diminish cumulative regret, amplify total throughput, reduce the number of nodes in starvation, and improve overall network fairness.