ORDINAL OPTIMIZATION APPROACH TO RARE EVENT PROBABILITY PROBLEMS

In this paper we introduce a new approach to rare event simulation. Be cause of the extensive simulation required for precise estimation of p erformance criterion dependent on rare event occurrences, obstacles su ch as computing budget/time constraints and pseudo-random number gener ator limitations can become prohibitive, particularly if comparative s tudy of different system designs is involved. Existing methods for rar e events simulation have focused on simulation budget reduction while attempting to generate accurate performance estimates. In this paper w e propose a new approach for rare events system analysis in which we r elax the simulation god to the isolation of a set of ''good enough'' d esigns with high probability Given this relaxation, referred to as ord inal optimization and advanced by Ho et al. (1992), this paper's appro ach calls instead for the consideration of an appropriate surrogate de sign problem. This surrogate problem is characterized by its approxima te ordinal equivalence to the original problem and its performance cri terion's dependence not on rare event occurrences, but on more frequen t events. Evaluation of such a surrogate problem under the relaxed goa ls of ordinal optimization has experimentally resulted in orders of ma gnitude reduction in simulation burden.