基於A2C結合LSTM之D2D通訊於蜂巢式網路功率調整演算法

Abstract

我們考慮到裝置對裝置之通訊(Device-to-Device communication, D2D)有可能在非理想情況下被建立，我們提出一個下行鏈路干擾緩解機制，在保障蜂巢式用戶的鏈路品質前提下，提升整體系統容量，我們模擬多個D2D用戶時採用抑制干擾範圍與正交資源分配，接著使用本實驗室學長先前所撰寫之位置推薦的結果，並稱之為傳統方法，本篇研究接續著傳統方法，搭配人工智慧的方法針對D2D提出功率調整演算法(Location Recommendation and Power Adjustment using A2C with LSTM, LR&PA_A2C+LSTM)，管理D2D複用蜂巢鏈路的資源時，基地台對D2D的干擾及D2D對MUE的干擾。我們比較LR的兩種策略，並且證明以最短移動距離作為唯一考量的策略較為實際，因此在我們的演算法中將採取此策略。模擬結果證明， 我們所提出的方法在提升D2D容量方面優於傳統方法，而蜂巢式用戶容量的減少相較於D2D容量的提升是非常小的，因此整體系統容量可以明顯地有所改善。

We consider the Device-to-Device (D2D) communication is likely to be established in non-ideal condition. The new downlink interference mitigation mechanism is proposed to enhance the overall system capacity on the premise of guaranteeing the links quality of macrocell user equipments (MUEs). We simulate the result of interference-suppression-area (ISA) and orthogonal resources allocation scheme with multiple D2D user equipment (DUE), using the Location Recommendation method the senior in our laboratory proposed before, and that is called conventional scheme. This paper proposes a power adjustment using A2C with LSTM (PA_A2C+LSTM) algorithm for DUEs to manage the interference from base station (BS) to DUEs and DUEs to MUEs when sharing same resources. We compared the two strategy about LR scheme and showed the strategy of LR by considering distance is more practical for users and hence this strategy is adopted by our proposed system. Numerical results show that our proposed algorithm is better than the conventional scheme in terms of DUE capacity. The loss of MUEs capacity is small compared to the gain of DUE capacity. The system capacity can thus be obviously improved.