Data Compression for Active Diagnosis

点击翻译按钮获取中文摘要

The goal of online diagnosis for open embedded systems is the localization and recovery from failures in open systems, where components can dynamically join and leave the system in order to provide safety-relevant services. In previous work, an approach for online diagnosis based on semantic web technology was introduced. New diagnostic information is derived using SPARQL queries from sensory data, status information and already inferred diagnostic information. Large amounts of diagnostic information need to be stored in local real-time databases and communicated between components of the system, although communication bandwidths in many applications are very limited. The primary objective of the project is increased efficiency and the reduction of overhead for online diagnosis in open embedded systems using data compression. New compression techniques and the adaptation of existing techniques is required to support the specific characteristics of online diagnosis such as large numbers of partly correlated data streams, which need to be stored and processed as dynamic and continuous time windows for the evaluation of diagnostic queries. The integration of compression components into an online diagnosis system requires also extensions in the inference process on diagnostic information and the scheduling of the inference. The developed algorithms for solving these problems will be prototypically implemented and evaluated in different application scenarios.