Power modelling of sensors for IoT using reinforcement learning

Abstract

Internet of things (IoT) is a technology where all things like household equipments, industrial elements, etc. are monitored by sensors and controlled by actuators. For a large scale IoT application, sensors are needed in huge numbers and all these sensors are powered by small battery. Hence, these miniature devices' lifetime can be improved by means of optimising the power and hence modelling of these sensors is a must for such application. This paper models the sensors for IoT application in the multi-layered IoT network. Reinforcement learning is used for modelling the sensors that model in the physical, routing and network layer. EEIT framework is used to model the nodes that optimise energy consumption in physical, routing and network layer. Physical layer modelling deals with the hardware aspects like transmission power, radio, etc. of the sensors. Routing and networking layer deals with the communication (transmitting and receiving data, dissemination, routing, etc.) capabilities of the sensors. We conduct numerical simulations and emulations using EEIT framework for IoT systems that are helpful for the design for complex IoT systems. Our results are quantified empirically based on the facts lifetime of the sensors, energy usage and communication costs.