Detection and Classification of a Wireless Communication Disruption

Motivation

The digital revolution is lately being driven by Internet of Things (IoT) devices, especially in areas where real time data analytic is essential. One of the wide spread usage is in the area of industrial automation where modern day industrial equipment have evolved to exploit the advancements in the smart factory context. With its wireless interfaces, the IoT devices can be portable to carry around and also provide simple means of data accessibility.

Although abundant research has been conducted on the wireless communication, it is evident that guaranteeing highly reliable communication over a long period of time could still be challenging. This holds true specifically in industries where multiple communication technologies interfere with one another. The primary reason behind such a behavior is the sharing of the unlicensed band of frequency spectrum. The implications of an unreliable communication could be catastrophic with heavy machinery and human beings working in close quarters.

Topic

In order to improve reliability in communication, it is essential to design a network which is resilient to communication disruptions. The primary component behind a resilient network is its ability to identify disruptions, by constant monitoring. This helps in taking counter measures to mitigate the adversities.

The usage of Artificial Intelligence has provided promising improvements in the field of wireless communication. It could be used to classify communication disruptions by modelling with suitable inputs. The following activities are necessary:

• Creating a simple ad-hoc wireless network using COTS devices and generating background traffic using IEEE 802.15.4 to induce disruption
• Developing test application to measure and gather network characteristics such as latency, throughput, energy metrics for multiple scenarios
• Designing machine learning models to classify the network state based on the network characteristics
• Evaluating the accuracy of the model for different combinations of network properties like latency, throughput, etc

Note: This thesis has to be written in English. In case you are interested in this topic, feel free to contact me for an appointment.

Skills required

• Python, Git
• Linux experience

What you will learn

• Networking in Linux
• Artificial Intelligence modelling

References