Structured Simulation Optimization and Analysis

Cornell University, Ithaca NY

Investigators

Abstract

A huge number of applications involve simulation optimization, i.e., the optimization of a function that is computed via Monte Carlo simulation. The primary goal of the research is to develop solution methods for structured simulation optimization problems. The word "structured" suggests that the problem possesses certain structural properties that can be leveraged by optimization algorithms to, for example, reduce solution time or improve the likelihood of convergence. In more detail, the primary goals of this research are as follows. First, to further develop simulation optimization methods for structured problems, where the structure may be detected and verified with minimal user input. Second, to apply those methods on a range of practical problems, with emphasis on emergency services in rural areas. Third, to create a library of simulation optimization problems that can be used to help compare proposed optimization methods, and help guide the development of new simulation-optimization algorithms. The development of structured simulation-optimization methods will allow far larger models and associated optimization problems to be solved than is possible today. The potential list of applications is enormous, since Monte Carlo simulation is now used in a tremendous range of fields including, but certainly not limited to, emergency service planning, manufacturing, communications systems, service system design and health care. The research explicitly includes applications to emergency service planning in rural areas, with the goal of obtaining more effective response times and coverage. Other potential benefits include improved methods for operating policy design for networks such as arise in power systems, communication systems and manufacturing. Applications also exist in health care, such as in external beam radiation treatment planning for cancer patients.