蜂巢式網路用戶與V2X通訊共存異質性網路之功率控制與資源分配演算法

在現今資訊暴漲的時代，無線網路是由許多的物聯網和通訊裝置所組合起來，而對於基地台原本所服務的蜂巢式網路用戶來說，因為基地台所需要服務的用戶不斷的增加，導致了基地台之間嚴重的互相干擾，為此我們通過提出一個下行鏈路干擾緩解方案，在確保了蜂巢式網路用戶的前提下，也保障了系統內的其他次級用戶不受到干擾，本文中以V2X通訊代表次級用戶。本論文建立了一個有多個多輸入單輸出(MISO)小區的環境，並在其中設置了數台採用C-V2X通訊的無人車，並使用人工智慧中的強化式學習模型Deep Q-learing 結合波束成形技術，提出了一種功率調整與波束成形演算法，每個基地台都代表一個代理(Agent)，並擁有獨立的神經網路，能夠根據基地台目前的環境做出適當的決策，我們的研究結果表明此演算法能夠有效保障蜂巢式網路用戶的權益(Utility)，並透過波束成形技術避開無人車，從而達到降低干擾並提升系統效能的目的。

關鍵字

強化式學習；深度學習；深度強化式學習；系統容量；波束成形

並列摘要

In today's era of soaring information, wireless networks are combined by many Internet of Things and communication devices. For cellular network users originally served by base stations, because the number of users that base stations need to serve continues to increase, This leads to serious mutual interference between base stations. For this reason, we propose a downlink interference mitigation plan, while ensuring the rights of cellular network users, and also ensuring that other secondary users in the system are not interfered. An environment with several multiple input single output (MISO) regional combinations is established, and a number of unmanned vehicles using C-V2X communication are set up in it, using the reinforcement learning model in artificial intelligence is used in combination Beamforming technology proposes a power adjustment and beamforming algorithm. Each base station represents an agent and has an independent neural network that can make appropriate decisions based on the current environment of the base station. Research results show that this algorithm can effectively protect the rights of cellular network users, and avoid unmanned vehicles through beamforming technology, thereby reducing interference and improving system performance.