A penalty function approach for simulation optimization with stochastic constraints

This paper is concerned with continuous simulation optimization problems with stochastic constraints. Thus both the objective function and constraints need to be estimated via simulation. We propose an Adaptive Search with Discarding and Penalization (ASDP) method for solving this problem. ASDP utilizes the penalty function approach from deterministic optimization to convert the original problem into a series of simulation optimization problems without stochastic constraints. We present conditions under which the ASDP algorithm converges almost surely, and conduct numerical studies aimed at assessing its efficiency.