Complex Event Processing over Unreliable RFID Data Streams

Existing RFID complex event processing (CEP) techniques always assume that raw RIFD data has been first cleansed to filter out all unreliable readings upfront. But this may cause delayed triggering of matched complex events. Furthermore, since the cleansed event streams need to be temporarily buffered for CEP evaluation, it may generate a large number of intermediate results. To address these issues, we propose an approach to perform CEP directly over unreliable RFID event streams by incorporating cleansing requirements into complex event specifications, and then employ a non-deterministic finite automata (NFA) framework to evaluate the transformed complex events. Experimental results show that our approach is effective and efficient.