Throughput Enhancement of Full-Duplex CSMA Networks via Adversarial Multi-Player Multi-Armed Bandit

This paper investigates the network-level throughput of a full-duplex (FD) enabled CSMA network, considering transmit power (TP) control and carrier-sensing threshold (CST) adjustment. With the FD technique, a transmitter-receiver pair can transmit and receive simultaneously in the same frequency band. The motivation is to find an optimal combination of TP and CST for each link so as to maximize the network throughput. The challenge is that adjusting each link's TP and CST will change the network's carrier-sensing relation and interference relation, which consequently leads to a computationally intractable network optimization problem. To overcome the complexity challenge, we model this network throughput maximum problem within a multi-player multi-armed bandit (MP-MAB) framework, in which the players are the FD-enabled links and the arms are the combinations of TP and CST. The proposed framework can also be viewed as an adversarial MP-MAB due to the hostile contention among links. Furthermore, we propose a refined Exponential-weight algorithm for Exploration and Exploitation (Exp3) to solve this adversarial MP-MAB problem. The refined Exp3 algorithm proceeds in epochs and starts with some prior knowledge. The numerical results show that the proposed method can improve the network throughput by more than 42%, compared with the random selection method. Meanwhile, the proposed algorithm exhibits a fast convergence rate in random network scenario.