Recent progress in global optimization has raised new interest in the application of global methods to lens design. This paper has several goals. We describe the attributes of a new global algorithm, Global SynthesisTM (GS), and we distinguish it from previously reported methods on the basis of efficiency and the ability to handle many (i.e., > 50) variables and comparable numbers of active, nonlinear, equality, and inequality constraints. Many experienced designers doubt, often with good reason, that multiple minima exist for practical problems, so we present meaningful examples showing additional minima often do exist. These test cases, some with known optimal solutions, can be used to study and benchmark the performance of different methods, and we describe recent results with GS on these test problems. We discuss what is meant by `success' in global optimization, and point out the practical limits of current methods. Finally, we discuss what impact global optimization may have on optical design as it becomes a mainstream tool of designers.
|