Model-ML Integrated Intelligence in URLLC Towards End-to-End Delay Fulfillment Over Vehicular Networks

Ultra-Reliable and Low-Latency Communication (URLLC) had been initially proposed as one of the three main application scenarios in the fifth generation of mobile telecommunications systems (5G). While URLLC is expected to support vehicular networking with low latency, current 5G infrastructures still cannot well assure about one-millisecond-level delay for various time-sensitive applications in vehicular networks. To better serve vehicular networks as well as other vertical application scenarios with stringent latency requirement, URLLC remains the hotspot for beyond-5G and 6G and is expected to dig deeper in optimization paradigm together with random-access control technologies. Model-based design principles integrating machine learning (ML) intelligence have been recognized as a competitive candidate to empower URLLC quickly into reality. In this article, we first anatomize the constitution of the end-to-end delay for URLLC towards vehicular networks and concentrate on the ways of how to apply model-ML integrated intelligence to reduce major delay components, including access delay, queuing delay, and transmission delay. Facing the challenging task, we derive an intelligent multi-tier-driven computing framework for access-delay reduction. We then introduce an efficient resource allocation approach driven by multi-deep-reinforcement-learning networks to jointly lower the queuing delay and transmission delay. Finally, we share the discussions about the open issues on latency control for future URLLC.