Max Weighted Queue (MWQ) control policy based on the theory of Lyapunov optimization is a cross-layer control policy that achieves queue stability and optimal delay performance. For the real-time and delay-sensitive demand in Device-to-Device (D2D) communication services, the MWQ algorithm, in which the Channel State Information (CSI) of PHY layer and the Queue State Information (QSI) of MAC layer are collectively considered, makes the maximum system throughput as the objective function and controls the power of D2D nodes dynamic. In this paper, a novel MWQ algorithm in the D2D communication was proposed. Compared to the algorithm with fixed power, the CSI-based algorithm and the QSI-based algorithm, the MWQ algorithm can decrease the average delay about 0.5 s when the average packets arrival rate exceeds 10 Mb/s and require less 26 dB power while having the same average delay. So the MWQ algorithm can achieve a good performance and have a reference to obtain low latency in D2D communication.