On the performance of MIMO-based ad hoc networks under imperfect CSI

In this paper, we investigate the performance of MIMO-based ad hoc networks under imperfect channel state information (CSI). Most existing results on MIMO ad hoc networks assume perfect CSI. This assumption is usually unrealistic in practice because the feedback overhead of perfect CSI is prohibitively high. Thus, it is of great interest to study how to optimize the performance of MIMO-based ad hoc networks under imperfect CSI. In this paper, we propose a solution based on stochastic approximation technique to address imperfect CSI. To analyze the performance gap due to the existence of CSI error, we first show that our proposed solution is stable and converges with probability one to a neighborhood of an optimal solution obtained under perfect CSI. Based on this stability result, we characterize the performance gap due to imperfect CSI. We show that the boundary of the stability neighborhood is constituted by points for which the excess capacity of each link is equal to the ergodic link capacity loss due to imperfect CSI. Thus, as the amount of CSI feedback increases, the performance gap of our proposed solution vanishes.