Magneto Inductive (MI) Channel Variables Prediction Through Machine Learning Linear Regression Method, for Underwater and Underground WSNs

In today’s digital era, where information flows seamlessly and is readily available and accessible. However, these information and communication systems are highly dependent on the robustness and adaptability of communication links, which must endure the dynamic and often unpredictable changes occurring in the communication environment, especially in underwater and underground communication. As technology advances and demands for reliable data transfer grows, the resilience of these links becomes increasingly critical to ensure an uninterrupted access to information. For this we will be analyzing a new paradigm in Wireless Sensor Networks (WSNs), i.e. Magneto Inductive (MI) communication. Which has the capability to provide robustness to the communication link as well as the security of the information can also be controlled, by utilizing the directionality factor of the created magnetic flux around the MI tri-directional (MI-TD) coils. In this study we will examine MI communication variable factors, which have effects on the output of the MI channel’s response, such as frequency of applied excitation current, number of turns, radius of the communicating coils, and changing communication distance between nodes. Than we will use a linear regression machine learning (ML) model to predict the values of the aforementioned variables, which will help to predict the MI channel’s response in the form of Path loss/Bit Error Rate (BER). The model that we have created for the prediction purpose, archives an estimation accuracy of 92%.