Adaptive Finite Blocklength for Ultra-Low Latency in Wireless Communications

With the very stringent demand for real-time transmission of wireless communication services, the requirement of sub-millisecond ultra-low end-to-end delay has been initially proposed in the sixth generation (6G) communication networks. Finite blocklength transmission is one of the potential technologies to meet such a low end-to-end delay demand for the next generation networks. However, as the finite blocklength decreases, the transmission delay decreases while the queuing delay increases, which results in the tradeoff between the transmission delay and the queuing delay. To achieve the optimal balance, in this paper we propose an adaptive blocklength transmission framework to minimize the important part of the end-to-end delay of wireless networks, where we focus on the transmission delay and queuing delay. A dynamic buffering model for variable transmission time interval (V-TTI) is introduced for the time-varying arrival of packets adaptation. Then, we propose the Flexible proximal Alternating direction method of multipliers based Blocklength Optimization (FaBo) scheme to minimize the important part of the end-to-end delay for the single user case. We also propose the Multiple deep Q-learning network based Resource Allocation (MuRa) scheme, which can efficiently balance the transmission delay and queuing delay, to minimize the important part of the end-to-end delay for the multi-user case. Numerical results show that the proposed adaptive blocklength framework can reduce the important part of the end-to-end delay compared with that of long-term evolution and the fifth generation (5G) new radio. We also show that our proposed schemes can quickly converge to the minimum end-to-end delay.