Distributed Federated Learning for Ultra-Reliable Low-Latency Vehicular Communications

Abstract

In this paper, the problem of joint power and resource allocation (JPRA) for\nultra-reliable low-latency communication (URLLC) in vehicular networks is\nstudied. Therein, the network-wide power consumption of vehicular users (VUEs)\nis minimized subject to high reliability in terms of probabilistic queuing\ndelays. Using extreme value theory, a new reliability measure is defined to\ncharacterize extreme events pertaining to vehicles' queue lengths exceeding a\npredefined threshold. To learn these extreme events, assuming they are\nindependently and identically distributed over VUEs, a novel distributed\napproach based on federated learning (FL) is proposed to estimate the tail\ndistribution of the queue lengths. Considering the communication delays\nincurred by FL over wireless links, Lyapunov optimization is used to derive the\nJPRA policies enabling URLLC for each VUE in a distributed manner. The proposed\nsolution is then validated via extensive simulations using a Manhattan mobility\nmodel. Simulation results show that FL enables the proposed method to estimate\nthe tail distribution of queues with an accuracy that is close to a centralized\nsolution with up to 79% reductions in the amount of exchanged data.\nFurthermore, the proposed method yields up to 60% reductions of VUEs with large\nqueue lengths, while reducing the average power consumption by two folds,\ncompared to an average queue-based baseline.\n