摘要：In this talk, we provide an overview on the frontier of research in the area of ultra-reliable low latency communications (URLLC). In particular, we first introduce a novel framework that combines deep reinforcement learning with generative adversarial networks (GANs) to enable model-free URLLC under limited data availability and without requiring any knowledge or assumptions on the delay models of the wireless users. We show how the proposed framework can intelligently optimize wireless resources while balancing the tradeoff between reliability, latency, and rate. This approach present a major departure from prior URLLC approaches that often ignore the rate constraints of the users and rely on historic data or on unrealistic delay modeling assumptions. Then, we turn our attention to the problem of joint communications and control for autonomous connected vehicles. In this area, we introduce a new model for characterizing the wireless reliability of an autonomous vehicle system while being explicitly cognizant of its control system requirements. After characterizing reliability, we show how one can optimize the operation of the autonomous vehicle system while jointly taking into account the delay of the vehicular network and the stability of the control system The synergies between URLLC and control system designs are then discussed. Then, if time permits, we will briefly look at how deep learning can be used to enhance the communication latency of drone systems. We conclude the talk with an overview on future opportunities in these exciting areas.