Q-Learning-Based Spatial Reuse Method Considering Throughput Fairness by Negative Reward for High Throughput

In this paper, we propose a Q-learning-based spatial reuse method considering throughput fairness in Wireless LANs (WLANs). In Spatial Reuse (SR) methods, wireless nodes try to use wireless resources efficiently by controlling both the Transmission Power (TP) and Carrier Sense Threshold (CST). When wireless nodes are densely deployed, the SR methods have difficulty to achieve both the high aggregate throughput and throughput fairness because the mutual interference among the wireless nodes becomes severe. The proposed method removes the difficulty by utilizing Q-learning where wireless nodes can learn the adequate CST and TP by themselves. The proposed method motivates nodes to use wireless resources actively by rewards, while it suppresses nodes with high throughput using the resources by negative rewards. As a result, the wireless resources are distributed among nodes with low throughput, and the proposed method achieves both the high aggregate throughput and throughput fairness. Simulation results show that the proposed method improves the aggregate throughput with keeping throughput fairness.