Quantized-feedback optimal adaptive multi-beamforming

A novel approach to adaptive transmit multi-beamforming is developed for frequency division duplex (FDD) multiple-input multiple-output (MIMO) systems. The method adapts each beam by computing at the receiver the steepest-ascent adaptation direction and feeding the quantized direction vector back to the transmitter; we call it the quantized-feedback optimal adaptive multi-beamforming algorithm. The method is widely flexible in the number of quantization bits and offers an adaptive diversity technique that allows for transmission of multiple data streams on orthogonal beams, thereby substantially increasing the system capacity or providing a new radio link for a new service. The convergence speed is maximized by computing the optimal increment vector which points in the direction of largest directional derivative of the objective function. We show that the algorithm converges globally irrespective of quantization error: the quantized increment vector always provides positive directional derivative at any non-stationary point in the solution space containing no local maxima. Simulation results are presented to illustrate the performance of the method.