Robust multiuser CDMA receivers based on the worst-case performance optimization

The performance of blind multiuser detection methods may become severely degraded in the presence of even small errors between the actual and the presumed desired user signatures. Such errors can frequently occur in practice due to an imperfect knowledge of the channel impulse response. We propose a new robust approach to blind multiuser detection in the presence of unknown arbitrary-type (but norm-bounded) errors of the desired user signature. Our approach is based on the explicit modeling of uncertainties in the covariance matrix of the desired user signature/data covariance matrix and optimization of the worst-case performance. Computationally efficient online implementations of the proposed multiuser receivers are developed.