Foresighted delayed CSIT feedback for finite rate of innovation channel models and attainable NetDoFs of the MIMO interference channel

Abstract

Channel State Information at the Transmitter (CSIT) is of utmost importance in multi-user wireless networks, in which transmission rates at high SNR are characterized by Degrees of Freedom (DoF, the rate prelog). In recent years, a number of ingenious techniques have been proposed to deal with delayed and imperfect CSIT. In this paper we consider Finite Rate of Information (FRoI) channel models (CM) introduced earlier, which captures the DoF of the channel coefficient time series. Both the block fading model and the stationary bandlimited channel model are special cases of the FRoI CM. However, the fact that FRoI CMs model stationary channel evolutions allows to exploit one more dimension: arbitrary time shifts. In this way, the FroI CM allows Foresighted Channel Feedback (FCFB) which provides CSIT at all times, even in the presence of CSIT FB delay, by increasing the FB rate. DoF-optimal schemes in the presence of perfect CSIT then maintain optimal DoF when combined with FCFB. This is applicable to any multiuser network. In this paper we analyze NetDoFs, which account for training overhead and FB DoF consumption on the reverse link. We provide attainable NetDoF expressions in the MIMO Interference Channel (IC) with linear transceivers for the DoFoptimal schemes of Interference Alignment (IA) with FCFB and several simpler schemes. Numerical evaluations show that the attainable NetDoF may be far away from the DoF and that simpler schemes may be close to optimal in terms of NetDoF.

Net degrees of freedom of decomposition schemes for the MIMO IC with delayed CSIT

“…As mentioned in [22], the numbers of active cells and active antennas N t and N r need to be optimized to find the right channel learning/using compromise because serving more users (or having more active antennas) means a larger DoF but also larger overheads. This why for small T c , TDMA, i.e., single user MIMO, is optimal.…”

“…In [22] net DoF of proper schemes were evaluated. Here we are interested in evaluating the net DoF of the two decomposition schemes, especially because of the robustness of ergodic IA to FB delay that was recently discovered and could bring net DoF gains.…”

From MU massive MISO to pathwise MU massive MIMO

“…To carry out this Tx/Rx design in a distributed fashion, global CSIT is required at all BS [2]. The overhead required for this global distributed CSIT is substantial, even if done optimally, leading to substantially reduced Net DoF [3]. In [4] the simplified SIMO uplink (UL) problem with only CSIR acquisition is considered and it is shown that it is impossible to maintain any positive Net DoF if one wants to design a cellular network that extends infinitely far (the problem arises already at a finite network size that depends on the Doppler bandwidth).…”