Optimization of limited feedback in two-way communication

In this paper, we study the role of combined CSI and ARQ feedback in the context of two-way wireless communications. We consider two nodes that wish to exchange data over, a frequency division duplex, wireless additive white Gaussian noise with Rayleigh fading channel. In two-way scenarios the data and resources allocated to feedback and channel estimation share the same link, leading to interesting tradeoffs. To analyze these, we present a generalized framework for feedback, which combines limited CSI feedback and ARQ. All resources in the two-way communication, that is, 1) training (estimating channel state), 2) feedback (ARQ and CSI), and 3) data are taken into account, and share the same noisy fading channel. We obtain an expression that captures the tradeoffs between allocating resources for these different tasks on the overall achievable throughput in each direction, which we numerically evaluate. In particular, we obtain the optimal resource allocations corresponding to different channel conditions, SNR regimes, transmitter protocols, and receiver feedback protocols.