부산대학교 도서관

Since unlicensed band already hosts well-established technologies such as Wi-Fi, the transmissions from another network systems in unlicensed spectrum could make Wi-Fi suffer from severe performance degradation without a fair coexistence mechanism. 3GPP has proposed their definition of fairness criterion, referred to as “3GPP fairness”, which restricts the Wi-Fi performance not to be affected by other unlicensed nodes. Thus it is of paramount importance to establish an access mechanism that guarantees the 3GPP fairness. This paper investigates how to maintain the 3GPP fairness between Wi-Fi and other unlicensed nodes, in the meanwhile, obtain the maximum of total throughput. We first obtain a benchmark by solving the optimization problem of total throughput under the 3GPP fairness constraint. Then we propose a deep reinforcement learning (DRL) mechanism based on Double Deep Q-network (DDQN) to help unlicensed nodes make access decisions while coexisting with Wi-Fi, so that they can learn to maximize total throughput without against the 3GPP fairness. Extensive simulations reveal that the DRL mechanism can approach the benchmark and therefore provides an approach for the coexistence of unlicensed nodes and Wi-Fi nodes.