TINYML ENABLED EDGE AI BASED ENERGY EFFICIENT WSN USING DIGITAL TWIN SIMULATION

Abstract

Abstract -In present time WSNs are very much used for continuous health care but traditional WSN systems do have high energy issue which they face because of continuous data transfer. In this paper we present an energy efficient patient monitoring system that we have put together using Tiny Machine Learning (TinyML) and the Digital Twin concept. We use physiological parameters like temperature, heart rate, and blood oxygen saturation (SpO2) to monitor patient health. Also we have put in a TinyML based edge decision which classifies the data into normal and abnormal right at the sensor node. As opposed to what we see in present WSNs which send out all the collected data, our put forth approach only sends out abnormal data which in turn greatly reduces unneeded communication. Also we have developed a Digital Twin model in MATLAB which we use to do real time patient monitoring and study system behavior using data driven input. We evaluate the performance of the put forth system by comparing it with a traditional WSN model in terms of data transfer and energy use. We present that the TinyML based WSN does in fact reduce the transmission rate and achieve great energy savings while at the same time does not sacrifice monitoring quality. This approach we put forth improves the efficiency and scalability of health care monitoring systems which in turn makes it very suitable for real time and resource constrained environment.