Performance Evaluation of User Scheduling in Full-Duplex Cellular Networks

摘要

Full-duplex (FD) communication has attracted research attention because of its high spectral efficiency. In FD cellular (FDC) networks, a base station (BS) simultaneously transmits and receives signals on the same frequency band, which has the potential for doubling the spectral efficiency of conventional cellular networks. However, self-interference and inter-user interference are important factors limiting the performance of FDC networks. Self-interference is caused by signals transmitted from the BS, which interferes with uplink (UL) signals. Inter-user interference is caused by signals transmitted from a UL user, which interfere with downlink (DL) signals. Many papers discussed self-interference suppression techniques. Propagation, analog circuit, and digital domain cancellation methods reduce self-interference. Nevertheless, residual self-interference may remain after interference cancellation. Therefore, the performance of FDC networks depends on residual self-interference and inter-user interference, and it is necessary to schedule users taking these interferences into account. In this paper, in order to analytically study the impact of residual self-interference and inter-user interference on the performance of FDC networks, we formulate the capacity and fairness. Because in FDC networks, the scheduler decides not only a DL user but also a UL user, the performance analysis of FDC networks is not so simple as that of conventional cellular networks. Thus, we assume that the user pair is an independent set, and this assumption enables us to derive the performance of user scheduling in FDC networks in the same way as the derivation of the performance of user scheduling in conventional cellular networks. The performance under the assumption can be used for a lower bound of the capacity because a multi-user diversity gain is not fully exploited. We analyze the performance of the scheduler that decides DL and UL users according to the maximum SINR normalized by average SINR. This scheduler is based on the same idea as proportional fair scheduling (PFS) in FDC networks. We confirm the validity of the assumption through numerical examples.