An Adaptive Multi-Agent System for Integrative Multidisciplinary Design Optimization

Abstract

Multidisciplinary optimization (MDO) problems are a specific class of optimization problem where the number of variables and disciplines involved is to important to directly apply classical optimization methods. Most of the existing approaches concentrate on separating the problem in distinct sub problems and using standard optimization methods on these sub problems while trying to maintain consistency among the variables shared by the sub problems. Basically these methods try to help the user to find an optimization process which reduces the complexity of the problem. However, a shortcoming of these MDO methods is that they require a strong expert knowledge of both the problem to be solved and the method which is applied, in order to obtain interesting results.