A Chernoff relaxation on the problem of application-oriented finite sample experiment design

In this paper, application-oriented experiment design formulated as a chance constrained problem is investigated. The chance constraint is based on the presumption that the estimated model can be used in an application to achieve a given performance level with a prescribed probability. The aforementioned performance level is dictated by the particular application of interest. The resulting optimization problem is known to be nonconvex in most cases. To this end, convexification is attempted by employing a Chernoff relaxation. As an application, we focus on the identification of multiple input multiple output (MIMO) wireless channel models based on a general L-optimality type of performance measure.