利用佇列狀態的裝置對裝置通訊網路資源分配演算法

Abstract

在行動網路資料流量快速成長的情況下，裝置對裝置的通訊技術也顯得日益重要。我們在論文中提出的網路資源分配演算法，將傳輸功率的控制、使用裝置對裝置通訊前後的網路吞吐量增益、以及最重要的佇列狀態評估都納入考量。傳輸功率的控制和吞吐量的增益目的都是為了提升系統吞吐量而做的管理。然而，如果只考量到吞吐量，某些裝置可能會產生資源飢渴的問題。因此，佇列的大小評估就被加入了我們的分析，用於減少囤積過多封包量的問題。面對佇列的問題，我們提出了同時將封包抵達速率、服務速率、及當下佇列大小都納入考量的「佇列權重函數」。在使用了我們提出的網路資源分配演算法後，平均的佇列大小以及封包延遲時間都能有效的改善，並同時維持甚至提升吞吐量的表現。

Facing the rapid growth of cellular data traffic, device-to-device (D2D) communicaiton becomes more and more important nowadays. In the thesis, we propose a resource allocation algorithm that takes into consideration several network factors, including power, throughput gain, and queue stability. We first use a power control algorithm to decide the transmission power. In addition, we use the weighted matching algorithm to find the optimal device that could share a channel with a specific user equipment (UE). However, a throughput-oriented solution may lead to a starvation problem. We proposed a queue-aware reosurce allocation scheme which takes both arrival rate, service rate, and current queue size into account. The proposed queue-aware resource allocation scheme could reduce the average queueing delay without decreasing the network throughput.