Delay-Aware Scheduling over mmWave/sub-6 Dual Interfaces: A Reinforcement Learning Approach

We consider a transmitter with mmWave/sub6 dual interfaces. Due to the intermittency of mmWave channel, the transmitter must schedule packets wisely across the interfaces to minimize the average delay by observing the system state. We use the well-known dynamic programming methods and Q-learning to find the optimal scheduling policy and investigate the influence of observing CSI on the optimal policy under different levels of knowledge of the environment. We find that only when the channel state transition model is not available, the instantaneous CSI can help in reducing system delay.