Efficient subset selection for the expected opportunity cost

A lot of problems in automatic control aim at seeking top designs for discrete-event systems. In many cases, these problems are most suitable to be modeled as simulation optimization problems, and a key question for solving these problems is how to efficiently and accurately select the top designs given a limited simulation budget. This paper considers the generalized problem of selecting the top m designs from a finite set of design alternatives based on simulated outputs, subject to a constraint on the total number of samples available. The quality of the selection is measured by the expected opportunity cost, which penalizes particularly bad choices more than the slightly incorrect selections and is preferred by risk-neutral practitioners and decision makers. An efficient simulation budget allocation procedure, called EOC-m, is developed for this problem. The efficiency of the proposed method is illustrated through numerical testing.