A COMPUTER ENGINEERING BENCHMARK APPLICATION FOR MULTIOBJECTIVE OPTIMIZERS
Simon Künzli, Stefan Bleuler, Lothar Thiele, Eckart Zitzler · Advances in natural computation · 2004
Among the vaious benchmark problems designed to compare and evaluate the performance of multiobjective optimizers, there is a lack of real-world applications that are commonly accepted and, even more important, are easy to use by different research groups. The main reason is, in our opinion, the high effort required to re-implement or adapt the corresponding programs. This chapter addresses this problem by presenting a demanding packet processor application with a platform- and programminglanguage-independent interface. The text-based interface has two advantages: it allows (i) to distribute the application as a binary executable pre-compiled for different platforms, and (ii) to easily couple the application with arbitrary optimization methods without any modifications on the application side. Furthermore, the design space exploration application presented here is a complex optimization problem that is representative for many other computer engineering applications. For these reasons, it can serve as a computer engineering benchmark application for multiobjective optimizers. The program can be downloaded together with different multiobjective evolutionary algorithms and further benchmark problems from