The unpredictable run-time configurations of dynamic, data-driven application systems require flexible simulation components that can adapt to changes in the number of interacting components, the syntactic definition of their interfaces, and their role in the semantic definition of the entire system. Simulation coercion provides one solution to this problem through a human-controlled mix of semi-automated analysis and optimization that transforms a simulation to meet a new set of requirements posed by dynamic data streams. This paper presents an example of one such coercion tool that uses off-line experimentation and similarity-based lookup functions to transform a simulation to a reusable abstract form that extends a static feedback control algorithm to a dynamic, data-driven version that capitalizes on extended run-time data to improve performance.
展开▼
