Spectrum Efficiency Maximization for UCNC Networks With Mobility Prediction

A user-centric no-cell (UCNC) architecture has been proposed to achieve a consistent user experience and enhance system capacity. However, high user mobility and complex network topologies present significant challenges in realizing these objectives. To tackle these challenges, we investigate the clustering and resource allocation problems in the UCNC network to maximize spectral efficiency (SE). We split this problem into two sub-problems: user-centric clustering and resource allocation. First, we propose a clustering algorithm based on mobility prediction and train a Long Short-Term Memory (LSTM) neural network to forecast user trajectories. Second, based on the clustering results, we develop an improved two-stage graph coloring algorithm to minimize interference between users and maximize resource block (RB) gains. Simulation results indicate that our proposed approach significantly outperforms the benchmark in terms of system SE while reducing signaling overhead produced by the updating of clusters.