Deep Reinforcement Learning-based Edge Discovery within the 3GPP Framework for C- ITS

With the evolution of edge computing, addressing challenges within the framework of the Third Generation Part-nership Project (3GPP) standard has garnered attention. In particular, challenges such as edge discovery and relocation, session management function (SMF) selection, and edge lifecycle man-agement pose significant hurdles in providing seamless services, especially in advanced Cooperative Intelligent Transportation Systems (C- ITS). This paper proposes an intelligent solution for edge discovery tailored to continuously provision C- ITS services to users within the 3G PP framework. Leveraging deep reinforcement learning (DRL), our proposed algorithm facilitates optimal edge discovery based on specific user requirements. We demonstrate the compatibility of our approach with 3GPP standard operations and address the critical challenge of edge discovery by cmplovina an intelliaent DRL-based methodology,

I. Introduction

The immense number of growing mobile devices in the 5G era has tremendously increased the mobile data traffic. Traditional cloud-based computing services served well for a number of applications in conventional networks owing to its huge computation and storage capabilities. However, the advanced 5G services such as massive Internet of Things (IoT), autonomous and remote driving, smart-city, healthcare, and augmented and virtual reality have stringent quality-of-service (QoS) demands [1]. Edge computing has emerged as a viable solution to satisfy the QoS demands of these advanced services thanks to its decentralized processing model. The ultra-low latency, limited energy consumption, and self-provisioning capabilities of edge computing enable the support of these advanced services [2]. These advantages of edge computing over cloud computing have made it a popular subject of research in both industry and academia. Although a well-researched topic, edge computing is still an evolving field with the emergence of machine learning technology [3].