Deep Learning Using Path Length Prediction for Internet of Things

Abstract

Sensors are employed in the Internet of Things (IoT) to collect data and establish connections with the internet. An instance of IoT can be seen in a tree-like topology constructed using wireless links. When a topology graph has a path from its root node to any other leaf or child node, and this path is influenced by the quality of wireless connections, it is known as a Destination Oriented Directed Acyclic Graph. The root node of the tree topology is responsible for implementing source routing for downstream paths to the leaf nodes of the tree. If the longest path for any node in a tested network graph, including the root, is determined by the maximum hop count, then the graph is considered to be connected. The real world and its applications are impacted by issues related to network connectivity in IoT. Models are employed to examine how changes in link probability and hop count affect the connectivity of the graph. In this research, the proposed Deep Learning (DL) model is evaluated using the Keras regression model. The simulated dataset is generated using the Cooja emulator. The link probability serves as a feature to predict the maximum hop count in the IoT. The predicted hop count based on the link probability aligns accurately with the tested data.