Low-complexity optimal CSI feedback in LTE

The commercial deployment of 3GPP LTE systems is gaining increased momentum all over the globe. Achieving the ambitious spectral efficiency targeted by LTE poses significant technical challenges in mobile terminal modem development. In this work, we address the challenges associated with developing one of the key modules in the mobile terminal, the (channel state information) CSI feedback module. Specifically, in LTE, CSI feedback from the mobile terminal to the base station consists of a preferred rank index and a preferred precoder matrix index (RI/PMI, corresponding to spatial processing), and a preferred channel quality index (CQI, corresponding to the modulation and coding rate). For the 4×4 MIMO system configuration targeted in LTE, computation of the optimal RI/PMI/CQI places an enormous burden on the mobile terminal, as much as making the CSI feedback module one of the most resource intensive baseband blocks. To overcome this bottleneck, we present a reduced complexity, optimal, CSI feedback algorithm for LTE. The key idea behind the complexity reduction is identifying and exploiting the structure of the LTE precoder codebook (the set of all possible precoder matrices), which makes computation of the optimal precoder matrix amenable to the matrix inversion lemma. In particular, we exploit the precoder codebook structure, in conjunction with the inversion lemma, to significantly reduce the complexity of the matrix inversion operations, that form an integral component of the CSI feedback algorithm. The proposed approach is shown to provide up to 2 times reduction in the number of matrix inversion operations required in CSI feedback. The computational savings are further amplified in the (practically relevant) special case when the interference across the mobile terminal antennas is modeled to be i.i.d.