Resource and Computationally Efficient Subchannel Allocation for D2D in Multi-Cell Scenarios With Partial and Asymmetric CSI

In underlay device-to-device (D2D) communication, assigning more D2D pairs to a subchannel can increase the spectral efficiency but it also increases the inter-D2D interference and causes interference to the cellular users (CUs). We consider the assignment of at most K D2D pairs per subchannel in a multi-cell scenario with multiple uplink subchannels. We propose a q-bit quantized feedback and resource allocation model that provides a quality-of-service guarantee to the CUs and ensures that the rates assigned by the base station (BS) to the D2D pairs can be decoded with a pre-specified outage probability even with unknown intercell and inter-D2D interferences. We propose a novel, polynomial-time, cardinality-constrained subchannel assignment algorithm (CCSAA) that applies for any K and achieves at least 1/2 and 1/3 of the optimal D2D sum throughput for q = 1 and q >= 2 bits, respectively. We also propose an alternate cardinality-constrained locally greedy algorithm (CCLGA) that has an even lower complexity and is just as effective in practice. We present a rate upgradation step that exploits the inherent asymmetry in the channel state information at the BS and D2D users to improve spectral efficiency. Our approach also addresses a novel extension to dynamic two-way D2D communications.