An improvement of Rosenbrock's algorithm for model based design optimization of microsystems

Due to the scaling of microelectronic as well as for example mechanical, optical or fluidic microsystems and their devices the complexity of these systems and the number of design variables is increasing constantly. In consequence of this tendency the exploration of the design space cannot be handled manually anymore and therefore automated design optimization is becoming an extremely important task in the development process. Suitable optimization strategies have to be chosen for this task and since the evaluation time of the objective function is very large, these algorithms have to be adopted in such a way that they avoid unnecessary calls to the objective function. A method derived from the well known barrier function approach is presented which omits calculations of the objective function at points that do not tend to deliver better results than the ones already obtained.