Modeling ATR applications for intelligent execution upon a heterogeneous computing platform

The information processing tasks associated with real-time applications (e.g., automatic target recognition, intelligent robotics, information fusion) have very diverse computational requirements that result in different needs for different computing system capabilities. Heterogeneous parallel computing provides a variety of architectural capabilities, orchestrated to perform an application whose tasks have such diverse execution requirements. A key issue that must be addressed in embedding real-time applications on heterogeneous parallel computing architectures is the design of a high-level operating system for selecting algorithms, matching subtasks to processors, and scheduling subtask execution. Essentially, high-level descriptions of an application would be “intelligently executed” by such an operating system. This paper presents on-going research in the development of an Intelligent Operating System, specifically describing the mechanism for the selection of algorithms that functionally satisfy the processing needs of ATR subtasks. The methodology presented here can also be used for other application domains and classes of hardware platforms whose characteristics are similar to those of the applications and platform considered here