5G毫微型基地台網路中人工智慧物聯網技術之研究

Femtocell networks and Internet of Things (IoTs) are important features in the future 5th generation (5G) communication system. When D2D communication is employed for IoT applications underlying the 5G Femtocell cellular network, it can enable a direct data transmission between User Equipment (UE) without the relay by Base Station (BS) so as to enhance the spectrum utilization, power-saving efficiency and network capacity. However, it might also cause inter-interferences between Femtocell cellular networks, degrading the overall system performance. In this project, we investigated the inter-interference problem in the Heterogeneous Networks (HeNets) which consist of Femtocells and D2D communications, and apply an artificial intelligence (AI) technology to improve the system utility of HeNets and Quality of Experience (QoE) for users. We provided an analytical model of inter-interferences in HeNets consisting of Femtocells and D2D communications. Based on the analytical model, we developed the power control and resource allocation approaches using the AI technologies, Long Short-Term Memory (LSTM) and Reinforce Learning (RL). The proposed approaches will adjust the transmission power of D2D communications and allocate radio resources dynamically according to the interference scenario, user density and traffic types to improve the total system utility of HeNets and QoE for users. By both the theoretical analyses and solutions expected to be investigated in this project, we hope to thoroughly provide an in depth view about the design and implementation of the future communication system.