To Sim or Not to Sim: Computing Budget Allocation in Simulation Optimization of Large Scale Networked Systems
Abstract: Advances in computing technology now allow us to simulate more and more systems with dynamics that can hardly be described by closed-form equations. This is especially true for many large scale networked systems such as remanufacturing, wireless sensor network, smart grid, and smart buildings. In order to find a good (not even the optimal) policy to govern the dynamics of these systems, one has to explore the exponentially exploding policy space using noisy and time-consuming simulation. Therefore it is of great practical interest to study how to efficiently allocate the computing budget so that a good policy can be found within a reasonable period of time. In this talk, we will show the optimal computing budget allocation for policy improvement and discuss its variants when the simulation time is stochastic and when the sample path can be reused. Applications in remanufacturing, wireless sensor network, and smart buildings will be shown.
