Data Validation and Sensor Life Prediction Layer on Cloud for IoT
Viju Chacko, Vikram Bharati · 2017
The Internet of Things (IoT) opens up unlimited possibilities. The core of IoT lies in its ability to collect data and actuate any "thing" ranging from monitoring and controlling complicated machines, to setting up mood lighting at home or monitoring the vital signs of a bedridden patient. At the heart of each of these possibilities-of all IoT possibility-is a wide range of sensors connected to a cloud infrastructure that can collate sensor data. The shelf-life and reliability of these sensors will vary widely, primarily because of the diversity of the environments and conditions in which they sensors operate. While it's possible to build data validation logic close to the sensor, it's not always practical given restrictions around power requirements, compute availability, and cost. The paper discusses pragmatic approaches to validating and correcting sensor data, and leveraging faulty sensor data to train cloud-based models that would predict sensor life. In addition, the paper discusses the possibilities of validating the correctness of the prediction models.