Predictive Machine Learning Analysis for Reliable D2D Discovery in 6G Critical Communications

With the advent of sixth-generation (6G) networks, significant advancements are anticipated in wireless communication, particularly in critical communication scenarios where reliable and efficient communication is of utmost importance. Device-to-Device (D2D) communication is a pivotal technology in 6G which allows direct communication between nearby devices with or without relying on centralized infrastructure and enables several services in 6G such as emergency search and rescue operations, target monitoring, etc. This paper presents a comprehensive measurement campaign conducted to assess the performance of ProSe direct discovery in terms of connectivity. Additionally, we have implemented and analyzed 5 machine learning (ML) models: multiple linear regression (MLR), polynomial regression (PR), support vector regression (SVR), decision tree regression (DTR), and random forest regression (RFR) that were trained on real-time experimental data collected in diverse heterogeneous environments to predict the probability of direct discovery in out-of-coverage emergency scenarios. Comparative analysis reveals that the DTR, RFR, PR, and SVR, are 30%, 29%, 18% and 12% more efficient compared to the MLR model, respectively, for predicting D2D discovery. This work suggests that ML models can help to deploy efficient D2D networks and make the node Intelligent to establish a reliable D2D link in 6G.