TSM: Temporal segmentation and modules-based computation offloading using predictive analytics and NR-V2X

Abstract

Recent increases in the usage of applications dependent on distributed computing have persuaded designers and entrepreneurs to utilize these solutions for diverse purposes, especially latency-constrained computation-intensive applications. Vehicular fog computing (VFC) is the innovative paradigm of distributed computing techniques, and therefore, a number of VFC offloading frameworks have been developed using AI-based advanced optimization procedures, with support from standards maintenance organizations like IEEE, 3rd Generation Partnership Project (3GPP), and some others. However, many of these strategies have not been adapted to specific application data and types, therefore these frameworks may function poorly despite their comprehensive offloading principles. Moreover, most computation offloading frameworks ignore the use of updated V2X protocols. We designed a temporal segmentation and modules (TSM)-based method specific for computation-intensive V2X applications that uses a four-tier hierarchy of resource-rich nodes and works in discrete time periods. TSM relies on status updates from previous time periods and uses predictive analytics to address the stochastic nature of vehicular networks using the latest 3GPP 5G V2X standards. Using an online modular computation offloading structure that heuristically manages the whole process, we were able to successfully and timely execute the latency-sensitive advanced vehicular applications. TSM supports computation-deficient devices in under a hundred milliseconds, makes use of smart vehicles’ processing units as fog nodes, and solves the optimization problem in short, discrete stages. We utilized Monte Carlo analysis, which confirmed that TSM outperformed the three other baseline methods.