PAM: Predictive analytics and modules-based computation offloading framework using greedy heuristics and 5G NR-V2X

Abstract

Recent advancements in distributed computing systems have shown promising prospects in enabling the effective usage of many next-generation applications. These applications include a wide range of fields, such as healthcare, interactive gaming, video streaming, and other related technologies. Among such solutions are the evolving vehicular fog computing (VFC) frameworks that make use of IEEE and 3GPP protocols and use advanced optimization algorithms. However, these approaches often rely on outdated protocols or computationally intensive mathematical techniques for solving or representing their optimization models. Additionally, some of these frameworks have not thoroughly considered the type of application during their evaluation and validation phases. In response to these challenges, we have developed the “predictive analytics and modules” (PAM) framework, which operates on a time and event-driven basis. It utilizes up-to-date 3GPP protocols to address the inherent unpredictability of VFC-enabled distributed computing systems required in smart healthcare systems. Through a combination of a greedy heuristic approach and a distributed offloading architecture, PAM efficiently optimizes decisions related to task offloading and computation allocation. This is achieved through specialized algorithms that provide support to computationally weaker devices, all within a time frame of under 100 ms. To assess the performance of PAM in comparison to three benchmark methodologies, the evaluation pathways that we employed are average response time, probability density function, pareto-analysis, algorithmic run time, and algorithmic complexity.