Heterogeneous-system Based Real-time Data Processing

This dissertation addresses three key challenges in heterogeneous-system based real-time data processing: real-time scheduling, simulation-based verification for real-time scheduling and visualization of real-time data. The effectiveness of the solutions is verified by their application to air-borne SAR high-resolution real-timeimaging system.A hierarchical scheduling framework and a hierarchical scheduling model for real-time heterogeneous systems are first presented. Real-time heterogeneous systems are modeled with them in four aspects: system structure, tasks, function modules and scheduling plans. And a new concept 'function class number' is used in them. A novel dynamic task assignment and scheduling algorithm for real-time heterogeneous systems is proposed based on the framework and the model. It adopts a new task assignment policy, and takes into account compatibility of static task assignment and scheduling algorithm. Simulation shows that the algorithm has obvious advantages over other related algorithms.Next, a simulation-based verification approach for real-time scheduling in heterogeneous systems is discussed. It consists of several stages: modeling, model-language-conversion, SMML translation, compiling, linking, simulation verification, online diagnosis and visualization. In the approach, a system model markup language is designed to describe real-time heterogeneous systems and their scheduling.Finally, this dissertation describes the real-time data visualization approach suitable for air-borne SAR high-resolution real-time imaging system. Different algorithms for real-time data visualization and generation of various views such as thumbnail are described.