Robust transceiver design for geometric mean decomposition systems with limited precoder feedback

With the assumption of a slow-time-variant multiple-input multiple-output (MIMO) channel, the use of channel state information at the transmitter coupled with the use of a jointly optimized linear transceiver can achieve excellent performance. The geometric mean decomposition (GMD) when combined with BLAST detection can provide the same diversity order as a more complex ML detector without sacrificing the rate of the MIMO system. The main obstacle to the practical implementation of this scheme is whether or not it performs well when the transmitter has limited feedback. In this paper we propose a decoder designed to be robust against quantization errors, as well as a quantizer that reduces the number of required feedback bits to approximate the performance of GMD when it has infinite feedback for an N times M MIMO system. Our results show that we can reduce the amount of feedback bits from 64 to 10 bits for a 2 times 2 and require only 30 bits for a 3times3 MIMO system, while achieving performance nearly identical to infinite feedback case.