Design and Analysis of Full-duplex Systems for Next-generation Wireless Communication Systems
To cater to the ever-growing demand for higher data rates, cellular networks and wireless local area networks (WLANs) need to adopt innovative technologies. Full-duplex (FD) is one such technology that promises to double the data rates by enabling simultaneous transmission and reception over the same frequency band. However, FD raises its own challenges. The first challenge is self-interference (SI) at the FD receivers due to power leakage from the collocated transmitter. Recent advances that employ techniques such as analog and digital domain signal cancellation and electromagnetic isolation of the receiver and the transmitter suppress SI below the noise floor, thereby making FD viable. The second challenge is inter-user interference due to the simultaneous uplink and downlinks in cellular networks and WLANs with FD-capable base stations (BSs) or access points (APs) and half-duplex (HD) users. It can degrade the downlink signal-to-interference- plus-noise ratio (SINR). In cellular networks, effective user-pair scheduling and mode selection algorithms are needed to mitigate it. In user-pair scheduling, the BS determines which users are scheduled on the uplink and the downlink. In mode selection, the BS selects either the FD mode or the HD mode for operation, depending on the feasibility. Since these decisions are taken based on the inter-user interferences between the users and these channels are not connected to the BS, the users need to feed them back. In a cell with N users, there are (N, 2) inter-user links, which can overwhelm the uplink. To limit the feedback overhead, we first propose a novel reduced feedback scheme in which a user feeds back the channel gains of a limited number of inter-user channels that are below a threshold and the corresponding user indices. Then, we propose a user-pair scheduling and mode selection algorithm (UPSMA). In UPSMA, the BS schedules a user based on the uplink and downlink SINRs. Then, the scheduled user feeds back the channel gains and corresponding indices using the above reduced feedback scheme. The BS selects a user for the reverse link from the fed back indices so as to maximize the sum rate. We analyze the spectral efficiency of UPSMA and the threshold for two different channel models. In the first channel model, the channels undergo multi-path fading and path-loss. In the second channel model, the channels undergo shadowing along with multi-path fading and path-loss. Next, we consider FD in WLANs. Since these operate in low mobility environments, the inter-user channels vary slowly, and the overhead of feeding channel gains back occasionally is small. The focus shifts to redesigning the conventional carrier sense multi-access with collision avoidance (CSMA/CA) based medium access control (MAC) protocols that do not allow simultaneous transmissions. We propose a MAC protocol called asymmetric FD-MAC (AFD-MAC) for WLANs with the FD-capable AP. AFD-MAC intelligently exploits hidden users to enable multiple transmissions. It uses a random back-off counter-based channel contention mechanism of the 802.11 HD MAC protocol to give channel access to a user or AP for transmission. Then, the AP schedules a user on the reverse link. It introduces a new signal to reduce the probability of packet collisions and a new management packet to inform users about uplink transmission opportunities when the AP wins the channel contention. We analyze AFD-MAC using a novel renewal-theoretic framework. We obtain accurate user-specific and network topology-specific expressions for the saturation throughput for the general scenario in which different users see different sets of hidden users. The analysis captures the impact of collisions due to interfering transmissions from hidden users on the saturation throughput. It leads to different statistical parameters for the AP and each of the users. In this case, we find that depending on the system parameters, FD can increase the throughput by a factor as large as two.