Filtered gradient algorithm for closed loop MIMO systems

This paper proposes a new adaptive beamforming algorithm for FDD MIMO systems with closed loop feedback. We base our work on recently proposed sign gradient methods [Banister, BC et al., 2003][Raghothaman, B 2003] that aim to maximize the signal to noise ratio (SNR) at the user equipment (UE) by adapting the complex baseband weights of the base station (BS) transmit antenna array, based on the quantized feedback information. The methods use simplified gradient estimates, suitable for slowly fading channels. In order to improve tracking in fast fading channels, we propose an algorithm that utilizes filtering of the gradient estimates and orthogonal stochastic search directions. Simulations show an improvement of 1 dB in uncorrelated channels at 30 km/h.