Implementation Of Cleaning Techniques For RFID Data Streams

RFID data are streaming, massive, time-dependent, semantic-rich, and inherently unreliable. With the extensive applications of RFID technology, how to effectively clean unreliable data generated by RFID systems is really an urgent problem to be solved,Traditional data cleaning techniques are not fully qualified for cleaning RFID data. Available RFID data cleaning systems usually provide various smoothing filters that interpolate for lost readings and aggregate data via sliding windows. They work well under various conditions, but they mainly focus an individual reader and have disregarded the very high cost of cleaning in a real application that has thousands of readers and millions of tags. Given the huge volume of data, diverse error sources, and rapid response requirements, setting the window size is still a challenging task.Based on the analysis of existing technology, we address the problem of adaptive cleaning with minimum costs, propose the concept of proximity group, and exploit proximity readers to extend the multi-tag cleaning mechanism of SMURF. Then, we design and implement a rule-based data cleaning system which is composed of reader configuration, cache management, rule management and data cleaner. In support of RFID data cleaning, the system incorporates declarative cleaning operations such as smoothing, arbitration, correction and combination. Finally various experiments are conducted on simulated and real RFID data sets to verify the effectiveness of the proposed approach. In comparison, the promising empirical results reveal that the proposed adaptive grouped cleaning is effective for lost readings and redundant RFID data.