Abstract : Device-to-device (D2D) communication has emerged as a vital paradigm within fifth-generation (5G) wireless networks, improving system capacity, spectrum performance and energy efficiency. However, the introduction of D2D links poses challenges such as interference with cellular links, complicating spectrum allocation and network quality assurance. This paper presents a novel approach that employs the Temporal Fusion Transformer (TFT) model, an attention-based architecture, to forecast a number of locations of user equipment within D2D networks. The TFT model provides interpretable insights into temporal dynamics, enabling precise mobility forecasting. These forecast results are then used to optimize resource allocation, revealing the significant impact of mobility forecasting on resource management. Furthermore, our study demonstrates the optimal performance of the our model as against traditional ones, offering improved accuracy and efficiency in mobility forecasting and resource allocation.
Index terms : Attention mechanism, D2D communication, Resource allocation, Mobility management, Temporal fusion transformer