Black-box optimization in a configuration system

Abstract

The product configurator Merlin is a CPQ solution (Configure, Price, Quote) that enables fast, error-free configuration and quotation generation for products with many variants. In the context of this paper an optimization module was developed and integrated into Merlin. Previously, Merlin could only minimize the number of changes that must be made when a user changes a configuration. With this work, the optimization capability in Merlin was extended in a way, that a user can define a custom target function. Specific features and variables then can be selected for optimization. The optimization module can optimize the values of these attributes and variables with respect to the defined target function. The optimization process has no limited runtime and does not stop automatically when reaching certain predefined values, since in the field of optimization often no promises can be made on finding global extrema. Instead, the optimization process is monitored live by the user and can be terminated at any time as soon as the user is satisfied with the current solution. In addition to the adaptation of the Merlin frontend, two black-box and derivative-free optimization algorithms are implemented and tested for performance to solve the optimization problem.