Joint optimization of rate allocation and BLAST ordering to minimize outage probability

We consider a wireless communication system over a multiple-input multiple-output Rayleigh-fading, channel with successive-cancellation detection. The outage probability of such a system is strongly dependent on two choices: the order in which the layers are detected, and the rate-allocation strategy at the transmitter. We propose the ride-normalized ordering algorithm, a generalization of the BLAST ordering algorithm that is shown to minimize outage probability. We further optimize the allocation of rate and energy at the transmitter, for a variety of receiver ordering strategies. Finally, we jointly optimize the receiver ordering and transmitter rate and energy allocations. Our main conclusion is that, for a wide range of data rates and SNR, the outage probability is minimized by a combination of rate-normalized ordering and a partially uniform rate and energy (PURE) allocation strategy. The jointly optimum system outperforms the Bell Labs layered space-time (BEAST) architecture by 15 dB at 8 b/s/Hz and an outage probability of 10(-3), when operating over a 4-input 4-output Rayleigh-fading channel. Also, the jointly optimum system shows an improvement of 1.5 dB over a recently proposed combination of optimum allocation and fixed ordering.