Multi-Armed Bandits in IEEE 802.11ac: Efficient Algorithms and Testbed Experiments

In response to growing wireless demands, new WiFi standards have significantly increased the number of transmission rates. This makes it challenging to identify the optimal rate quickly. We propose WiFi-UTS, a unimodal MAB algorithm, to address this problem. As a baseline algorithm, we use the state-of-the-art rate adaptation algorithm, Minstrel-HT, and implemented Thompson sampling. The algorithms were evaluated within an RF-shielded WiFi testbed, rigorously assessing their throughput performance to showcase their efficiency under controlled conditions. Our experiments show that WiFi-UTS outperforms the baseline algorithm with an average improvement of 47% and at least 12% in all scenarios considered.