Power Modelling of Sensors for IoT Systems using Reinforcement Learning

Speaker:  Pradeep Kumar TS – Chennai, India
Topic(s):  Networks and Communications

Abstract

Wireless Sensor nodes consist of communication devices, physical (environmental Sensors), a processing unit, a small memory and radio. The power consumed during communication is high and optimising the power and energy during communication is really necessary. The main aim of these networks is to increase the lifetime of the nodes in the network. So modelling these nodes for better lifetime is necessary. 


This presentation addresses this issue by modelling the sensor nodes for Mission Critical Systems (MCS). Mission Critical systems are systems use tasks to accomplish that have real time deadlines. If a deadline is not met, some catastrophe may occur and the sensor nodes which sleep during critical times will lead to an unstable system. So, instead of going to a sleep state, the state changes to an idle state for handling critical tasks. In this thesis, the motes are characterised using semi Markov decision process (SMDP). Various policies were framed for non-critical and mission critical systems. Mission critical systems uses nodes that meet the deadlines thereby optimising the power and energy used. 

This lecture also proposes the efficiency of the MCS by 25%. The model is validated using Crossbow Micaz motes. Also the model selects the actions in the node state and suggests the best policy for better energy optimisation. Results show the ability of the motes to go from active to sleep state for non-critical applications and active to semi-sleep state for mission critical application. Additional power saving achieved is seen as 25%. 


This lecture also models the sensors for IoT application in the multi layered IoT network. The sensors are modelled using a reinforcement learning approach that works with all the layers like routing, physical and network layers. A framework called as EEIT (Energy Efficient Internet of Things) is being designed that models the sensors in the network layers. Numerous experiments have been conducted using EEIT framework and evaluated the performance of the sensors that were helpful for the design of complex IoT systems for the future. 

About this Lecture

Number of Slides:  67
Duration:  60 minutes
Languages Available:  English
Last Updated: 

Request this Lecture

To request this particular lecture, please complete this online form.

Request a Tour

To request a tour with this speaker, please complete this online form.

All requests will be sent to ACM headquarters for review.