This award provides funding to advance the design of algorithms for solving simulation-optimization (SO) problems on parallel computing platforms. SO problems are optimization problems where the objective function and constraints can only be observed through a stochastic simulation. Virtually all current algorithms for solving SO problems assume a single processor as the computing platform. However, the trend in computing devices is towards multi-processor computers, not just at the laptop/desktop level, but all the way up to cloud computing environments. This research will explore how to design algorithms for solving SO problems that exploit such environments, to attempt to return high-quality solutions at a reasonable cost and within a reasonable amount of time.
If successful, the results of this research will lead to a new line of research - parallel SO - with ensuing improvements in the design and implementation of SO algorithms on parallel computing platforms, thus making this currently computing-intensive technology much more accessible and effective. SO already holds an important place in application fields, as evidenced by the variety of SO problems in an existing testbed <www.simopt.org>. The proposed research will provide further opportunities for major impact through reliable solution of important SO problems. This research is part of a continuing thrust to enhance the implementability of SO by creating methods, theory, and computational tools to facilitate the automatic solution of larger and more realistic SO problems.
