This paper summarizes some of our unclassified work on concepts and techniques for performing automated runtime time-critical resource management (especially scheduling) in large scale, dynamic, control systems such as for battle management. The approach described here is based on application-level quality of service metrics, such as track quality and weapon spherical error probable. These metrics are used to derive parameters for thread time constraints in the form of time/utility functions. Threads are scheduled according to application-specific optimality criteria that seek to maximize accrued utility to the system. Two worked examples illustrate the cost-effectiveness of this approach for this class of application.
E. Douglas Jensen