Genetic Algorithm Based Blind Beamforming for Multiusers

In this paper, we proposed a new genetic algorithm (GA) based blind beamforrming algorithm for multiusers without the Gram-Schmidt orthogonalization (GSO). This algorithm can estimate weight vectors by defining a new cost function based on the kurtosis maximization algorithm (KMA) and the cross-correlation of the beamformer port outputs. Moreover, to overcome the local optimal problem existing in the stochastic gradient algorithms, a global optimal solution of nonlinear weighting vector estimation is obtained by complex coding genetic algorithm (CC-GA) proposed in this paper. The coded parameters in the CC-GA are composed of real parts and imaginary parts of complex weight vectors. The fitness function equivalent to the objective function of the traditional optimization techniques is constructed by the new cost function. The performance of the new algorithm is analyzed with computer simulation.