Multiobjective Optimization and Sensor Correlation Framework for IoT Data Validation

Abstract

Internet-of-Things (IoT) systems generate massive data, but not all of them provide valuable information about an oddity in the monitored environment. Also, not all sensors in an IoT application require a constant operation, especially, in gently fluctuating environments. Hence, extensive bandwidth and power get wasted. As a result of the power wastage, the sensor network life gets reduced. Also, the data reliability is diminished without a data validation mechanism. Sensor correlation and selective activation of sensors to validate a reported change in the environment parameter can address these issues. Thus, this article proposes a framework for sensor correlation and the selection of sensors to activate and verify a reported abnormality in the monitored environment. We formulate and solve the sensor selection as a multiobjective optimization (MOO) problem, considering the power demand of sensors, network quality, remaining battery level, correlation value of sensors, and free CPU and RAM of the edge gateway device. Finally, we provide detailed results of our extensive evaluation of the proposed framework on test hardware. Implementation results show that the proposed framework consumes minimal edge gateway resources, that is, 3.47%RAM, 29.1%CPU, and 0.29 A current. The maximum execution time was 1006.28 ms.