Building Zero-touch Service Management Framework for Automotive Services Using the Smart Highway Testbed

Abstract

In Beyond 5G (B5G) and towards the Sixth Generation (6G), the performance of vehicular communications can be evaluated and improved through services located in the cloud and closer to the users in the Multi-Access Edge Computing (MEC) units. The increasing demand for connected cars requires an optimal distribution of the available network resources to ensure required levels of End-to-End (E2E) latency and the reliability of the vehicular services which contributes to the safety of the participants in traffic. The allocation of these resources can be improved with the support of Zero-touch Network and Service Management (ZSM). In this paper, we present a ZSM framework for automotive services using the Smart Highway testbed to improve the performance of vehicular services in a realistic environment. The proposed framework consists of decision-making processes that follow the principles of ZSM and intent-driven management. The decisions are based on the MEC units' workload and availability to match the demand from vertical services. The collected datasets from the decisionmaking process algorithm are used to train a Deep Reinforcement Learning (DRL) model and compare its results with a simple rule-based algorithm. The results show that DRL can quickly adapt to the dynamic environment of a testbed and outperform the conventional rule-based approaches. This indicates that the DRL algorithm can improve the decision-making process and ultimately decrease the E2E latency of vehicular services.