This paper investigates the problem of ranking and selection under input uncertainty with simultaneous resource allocation. In this problem, two types of resources are sequentially allocated at the same time to collect input data to reduce input uncertainty and run simulations to reduce stochastic uncertainty. We formulate the simultaneous resource allocation problem as a concave optimization problem that aims to maximize the asymptotic probability of correct selection (PCS) through the allocation policy for both input data collection and simulation, based on a moving-average estimator for aggregation of simulation outputs and its asymptotic normality. The two optimal policies are interdependent since they jointly affect the PCS. We derive the optimality equations to characterize the optimal policies and develop a fully sequential algorithm that demonstrates high efficiency through numerical experiments.
Paper
Full text
Input Data Collection Versus Simulation: Simultaneous Resource Allocation
Semantic Scholar · Mathematics · 2023
Abstract
This paper investigates the problem of ranking and selection under input uncertainty with simultaneous resource allocation. In this problem, two types of resources are sequentially allocated at the same time to collect input data to reduce input uncertainty and run simulations to reduce stochastic uncertainty. We formulate the simultaneous resource allocation problem as a concave optimization problem that aims to maximize the asymptotic probability of correct selection (PCS) through the allocation policy for both input data collection and simulation, based on a moving-average estimator for aggregation of simulation outputs and its asymptotic normality. The two optimal policies are interdependent since they jointly affect the PCS. We derive the optimality equations to characterize the optimal policies and develop a fully sequential algorithm that demonstrates high efficiency through numerical experiments.