MSE estimation of multichannel signals with model uncertainties

We consider the problem of multichannel estimation, in which we seek to estimate multiple input vectors that are observed through a set of linear transformations and corrupted by additive noise. The input vectors x_k are known to satisfy a weighted norm constraint. We discuss both the case where the linear transformations are fixed (certain) and the case where they are only known to reside in some deterministic uncertainty set. We seek the linear estimator that minimizes the worst-case mean-squared error (MSE) across all possible values of the linear transformations and possible values of x_k. We show that for an arbitrary choice of weighting matrix, the minimax MSE estimator can be formulated as a solution to a semidefinite programming problem (SDP). In the case in which the linear transformations are fixed and the norms are unweighed, the minimax MSE multichannel estimator has an explicit closed from solution. Finally, we demonstrate through examples, that the minimax MSE estimator can significantly increase the performance over conventional least-squares based methods.