Skip to Main Content
For multiple-input/multiple-output (MIMO) wireless communications systems employing spatial multiplexing transmission we evaluate the convergence performance of genetic algorithm (GA)-based detection against the maximum-likelihood (ML) approach. We consider transmit-correlated Rayleigh and Rician fading with Laplacian power azimuth spectrum. The values of the azimuth spread (AS) and Rician K-factor are selected according to the measurement-based WINNER II channel models, for several relevant scenario types. We consider the effect on GA convergence speed and population size requirements of the following: number of antennas, modulation constellation size, scenario (i.e., AS and K values), and rank of the deterministic component of the channel matrix. We find that the GA population size needs to be carefully adjusted to the antenna geometry and modulation constellation in order to maintain fast convergence. On the other hand, changes in the channel fading type and geometry do not appear to affect GA convergence. GA is shown to achieve ML-like performance, possibly for lower complexity, i.e., more efficient hardware and power usage.