Skip to Main Content
The recent proliferation of system-of-systems (SoS) trade spaces has allowed the investigation of highly complex interactions that have historically not been analyzed. Because of the size and complexity of these trade spaces, traditional methods of learning, exploring, and eventually understanding the complex interactions are inadequate. The first third of this paper is used to discuss the design approach of learning SoS through simulation, while the second third exposes many of the inadequacies of the current techniques. The final third of this paper calls for the development of new methods by presenting specific areas in which new techniques must be improved. Some examples are adaptive sampling methods which adjust local repetitions depending on local variance and batch experiment sizes that allow a designer to dynamically change the number of experiments depending on the number of available computational nodes.