A Machine Learning Approach to the Estimation of Near-Optimal Electrostatic Force in Micro Energy-Harvesters

Abstract

Wearable medical sensors are one of the key components of remote health monitoring systems which allow patients to stay under continuous medical supervision away from the hospital environment. These sensors are typically powered by small batteries which allow the device to operate for a limited time. Any disruption in the battery power could lead to temporary loss of vital data. Kinetic-based micro-energy-harvesting is a technology that could prolong the battery lifetime or equivalently reduce the frequency of recharge or battery replacement. Focusing on a Coulomb-Force Parametric Generator (CFPG) micro harvesting architecture, several machine learning approaches are presented in this paper to optimally tune the electrostatic force parameter; and therefore, maximize the harvested power.