AI-Driven Traffic-Aware Dynamic TDD Configuration in B5G Networks

The advent and anticipated evolution of Beyond Fifth Generation (B5G) networks raise critical issues for the static Time Division Duplex (TDD) radio resource allocation technique. In Static TDD, the fixed allocation of uplink and downlink resources leads to poor resource utilization, with uplink channels often congested and downlink channels underutilized. This study addresses static TDD limitations by proposing a novel TDD configuration called Traffic-Aware Dynamic TDD (TA-TDD), aiming to satisfy the high-speed and low-latency communication requirements of various applications. Specifically, the proposed TA-TDD utilizes Convolutional Long Short-Term Memory (ConvLSTM) model to predict traffic before allocation of uplink and downlink resource. This method effectively manages uplink-centric traffic in wireless networks, to improve both network quality and user experience. Compared to static TDD, the proposed TA-TDD notably improves network throughput by as much as 20% in scenarios with high uplink demand. The findings demonstrate that dynamic TDD configurations significantly enhance network throughput compared to static setups, which offers an effective solution for network management.