PROJECT TITLE :

Quantile-Based Simulation Optimization With Inequality Constraints: Methodology and Applications

ABSTRACT:

Several automation or manufacturing systems are too advanced to be modeled by analytical approaches and will solely resort to quick-running simulation. Stochastic Nelder–Mead simplex methodology (SNM) may be a newly developed methodology for simulation optimization with expected-worth-primarily based objective functions. Quantile, as an important various to the usual expected value, provides further info about the distribution of system performance. In particular, it's helpful in describing the tail behavior of the distribution. During this paper, we tend to exploit the structure of SNM and extend it to unravel simulation optimization issues with quantile-based objective functions and inequality constraints. The proposed technique, called SNM-QC, utilizes the identical search strategy as SNM however more incorporates effective quantile estimation techniques and penalty operate approaches to unravel the problem. We have a tendency to prove that SNM-QC has the fascinating global convergence guarantee, i.e., the algorithm is absolute to converge to the true optima with probability one. One advantage of SNM-QC is that it's an instantaneous search method that determines the moving direction simply by comparing a group of solutions instead of estimating gradient, so it can handle several sensible issues where gradient will not exist or is difficult to estimate. An extensive numerical study shows that the performance of SNM-QC is promising compared to the prevailing heuristics. 2 illustrative applications are provided in the tip to demonstrate the viability of SNM-QC in sensible settings.