Research On The Key Techniques For High Available Sensing Data Acquisition

In recent years, with the rapid development of the IOT (the Internet of Things), the type of sensing devices has become increasingly diversified, leading to abundant sensory data of the IOT. Moreover along with the development of sensing network, it usually consists of a host of mutually independent data sources, which can be used to monitor measured objects from multiple dimensions. Based on information mentioned above, this thesis divides the multi-source sensory data into four types including isomorphism single-mode, isomorphism multi-modal, heterogeneous single-mode and heterogeneous multi-modal. Even if the precision of sensing devices is significantly enhanced, there still exist false negative readings, false positive readings and wrong readings. The availability of data is reduced due to the inherent hardware limitations, system aspects and environment interference, etc. However, the data is the carrier of information, whether it can accurately reflect the real physical world information is the key of its utility in the upper applications. So how to improve the availability of sensory data is an urgent problem to be solved. Therefore, this article focuses on the research of high available sensory data acquisition key techniques.Initially, two methods are proposed towards Fixed Reader Mobile Tag (FR-MT) model and Mobile Reader Fixed Tag (MR-FT) model respectively, in terms of isomorphism multi-source single-mode state of sensory data. Under the FR-MT model, a new data cleaning strategy---LC-INFER is proposed, which is more suitable to practical applications based on data redundancy from multiple readers considering of the RFID data features, prior knowledge about the readers and the environment, and given constraints in target applications. We deploy the real supply chain environment to derive the real-world data. Our experimental results, using both real-world data and large simulated data, demonstrate the accuracy and efficiency of our proposed algorithm. ATOrientation, another new RFID data acquisition method raised under MR-FT model, uses passive tags to location the moving readers. This method determines the probabilistic locations of reader through the readings of this reader to massive tags, and then divides the probabilistic area into grids for precise orientation. In particular, the ATOrientation algorithm is suitable for a wide range of monitoring area where the tag deployment is intensive, and the deviation of location obtained by experimental verification is generally very small.Secondly, in view of sensor data, which is the presentation of isomorphism multi-source multimodal state sensory data, a high available sensor data acquisition method based on information quality is proposed. It is easy to identify two aspects of information quality-data reliability and data share. The former one is mainly measured by the deviation between sensor data and reference data, and the latter one emphasize the correlation among sensor data transmitted. Furthermore, based on the information quality, we transfer the data acquisition problem into an optimization one, which is aiming at maximizing the reliability of sensory data and the minimizing the information sharing under the energy constraints of sensor node thereby solving this problem through the GA algorithm. And at1ast, the experimental results derived by a lot of experiments show that our algorithm is efficient and scalable.Finally, in view of heterogeneous multi-source multimodal sensory data, we propose the high available sensory data acquisition method based on data quality (DQ) in terms of three aspects-accuracy, integrity, and consistency. By modeling these three aspects respectively, we propose metrics to estimate the comprehensive data quality method of heterogeneous multi-source multimodal sensory data. According to the given precision, a data source selection algorithm is presented based on DQ, which selects a part of data sources in order to achieve a balance between DQ and consumption of network resource. At last, the effectiveness of the algorithm is verified by experiments.In conclusion, based on the perception of the IOT, this thesis divides the sensory data into four types according to the type and function of sensory devices, and carries out high available sensory data acquisition techniques for each type of sensory data.