Towards 6G for Connected Autonomous Vehicles: A Trial Facility Analysis

Artificial Intelligence (AI) techniques for traffic management and optimization have the potential to funda-mentally change the way transportation systems operate. In particular, Deep Reinforcement Learning (DRL) has been shown to achieve significant improvements in numerous traffic-optimization approaches, allowing transportation systems to become safer, more efficient, and environmentally friendly. How-ever, most approaches proposed for such traffic management have so far been validated only within simulated environments as real-world test facilities tend to be limited in availability and functionality. In this paper we elaborate on Deep Reinforcement Learning (DRL)-based smart traffic management applications for future Connected Autonomous Vehicles (CAV s) and give insights on their vertical integration into emerging 6G infrastructures in view of enabling realistic experimentation. We propose a layered architecture in which the experimenter develops a software as a service (SaaS) service layer on top of platform as a service (PaaS) layers provided by the trial facility. We also present a qualitative analysis of existing trial facilities for smart traffic management and address the potential challenges within these trial facilities.