Decision Optimization Module for The GASP IV Simulation Language.

This Thesis presents the development of an optimization module that can be added to the GASP IV simulation language to provide automated optimization of a user specified objective function over a set of feasible values of user defined decision variables. The optimization module employs a Hooke-Jeeves pattern search which is modified to accept both discrete and continuous decision variables. In addition, the decision variables may be bounded or unbounded. A unique feature incorporated in the optimization module is a procedure for automatically controlling each simulation run length by statistically testing the objective function value at user scheduled time intervals to determine if it is significantly better or worse than the best value thus far obtained.