Sleeping Multi-Armed Bandits for Fast Uplink Grant Allocation in Machine Type Communications

Abstract

Scheduling fast uplink grant transmissions for machine type communications (MTCs) is one of the main challenges of future wireless systems. In this paper, a novel fast uplink grant scheduling method based on the theory of multi-armed bandits (MABs) is proposed. First, a single quality-of-service metric is defined as a combination of the value of data packets, maximum tolerable access delay, and data rate. Since full knowledge of these metrics for all machine type devices (MTDs) cannot be known in advance at the base station (BS) and the set of active MTDs changes over time, the problem is modeled as a sleeping MAB with stochastic availability and a stochastic reward function. In particular, given that at each time step, the knowledge on the set of active MTDs is probabilistic, a novel probabilistic sleeping MAB algorithm is proposed to maximize the defined metric. Numerical results show that the proposed algorithm has logarithmic regret, and hence is optimal. The results also show that the proposed framework yields a three-fold reduction in latency compared to a random scheduling policy since it prioritizes the scheduling of MTDs that have stricter latency requirements. Moreover, by properly balancing the exploration versus exploitation tradeoff, the proposed algorithm is able to provide system fairness by allowing the most important MTDs to be scheduled more often while also allowing the less important MTDs to be selected enough times to ensure the accuracy of estimation of their importance.