A Practical Algorithm for Realizing GDFE Precoder for Multiuser MIMO Systems

It is well known that multiuser multiple-input multiple-output (MU-MIMO) systems can achieve superior data rates compared to single user MIMO links. The improvement in data rates offered by the MU-MIMO systems is dependent on the design of precoding scheme for the broadcast channel (BC). A precoding scheme based on generalized decision feedback equalizer (GDFE) is known to achieve MIMO BC capacity. However, GDFE precoder suffers from huge computational complexity and is not suitable for practical systems. Much of computational cost of implementing a GDFE precoder can be attributed to its reliance on the covariance matrix corresponding to the ``least favorable noise'', which has prohibitive computational complexity. In this paper, we provide an alternative framework for realizing a GDFE precoder, which avoids the need to compute ``least favorable noise''. While maintaining capacity optimality of the GDFE precoder, the proposed algorithm has significantly lower complexity that is comparable to other MU-MIMO precoding schemes. Additionally, the proposed algorithm provides a useful tradeoff between desired complexity and performance.