具高效率節能之層級式小細胞網路資源分配

Abstract

在本篇論文中，透過恰當的分配資源區塊(resource block)，我們研究了小細胞網路(small cell network)最小化傳送功率的問題，並且符合服務質量(quality-of-service) 與功率預算的限制。我們提出一種集中式(centralized)的運算方式，在此方式之中，中心處理器擁有基地台與用戶裝置之間的通道資訊。為了方便處理問題，我們將零與一的分配問題，放寬至零到一的範圍，以利執行最佳化程式。我們也透過對分法(bisection method)限制每次最佳化的使用的資源區塊數量。 此外，根據文獻，我們將原來資源分配問題，化成層級式演算法(hierarchical algorithm)，總共會有三個階段。每一個階段會先處理過一次再傳到下一階段，因此所需要的overhead就會比較小。在第一階段每一個小細胞基地台會收集使用者的平均通道增益，並且估計需要多少個通道才足以滿足使用者們的服務質量。通道需求量會提交給資料中心，統一由資料中心透過干擾圖(interference graph)發配通道，即為第二階段。最後一階段，小細胞基地台會根據資料中心授權，將可用通道轉換成資源區塊服務用戶，透過類似於集中式演算法的做法，達成最小化傳送功率與維持服務質量的需求。 我們和文獻中的兩種方法比較，啟發型(heuristic)資源分配演算法[3]與層級式演算法[6]。為了符合我們的情境，啟發型演算法當中的資源我們改成資源區塊而不是子通道(subchannel)。此外我們還將啟發型演算法套入集中式與層級式演算法中。模擬結果顯示，我們提出的集中式演算法可以最省能量，最消耗能量的則是層級式演算法[6]。我們也使用每能量速率(rate per energy)當作比較依據，越高的每能量速率表示越高的系統效能。結果顯示我們提出的集中式演算法在低服務質量需求時，可以擁有最好的系統效能。

In this work, we study the problem of maximizing the sum rate of the network, with quality-of-service (QoS) constraints, by properly allocating resource blocks to the users in small cell networks and in a sense to restrain the energy consumption by using as few resource blocks as possible in that situation. Centralized resource allocation algorithm (CA) is proposed where the centralized coordinator such as Data Center (DC) knows the information channels between all the small cell base stations (SBSs) and user equipments (UEs). To make the problem tractable, we relax the integer constraint of association indicator to any real value between 0 and 1. We first fix the number of used resources as a constant in each iteration and then employ bisection method to obtain the optimal solutions. Simulation results show that the optimum solutions of the relaxed problem are close to the boundary, i.e., either the result is close to 0 or 1, due to the fact that the objective function is linear and monotonic. On the other hand, in order to further reduce the overhead of channel state information (CSI) feedback to the DC, we propose another hierarchical algorithm (HA) that decomposes the centralized algorithm into three phases. First, the load of each SBS is estimated. Second, based on each SBS’s load demand, the channels are allocated to SBSs by the technique of graph coloring. Finally, the resource allocation is performed at each SBS with QoS constraints and limited power budget. We compare the proposed centralized algorithm (CA) and hierarchical algorithm (HA) with the algorithms proposed in [6] and [3]. Simulations show that the CA consumes less energy than the HA, while HA requires less overhead than CA for CSI feedback in the network. We also compare the proposed CA and HA with the heuristic algorithm studied in [3] in which RBs are allocated to users sequentially until QoS requirement is satisfied. Simulation results demonstrate that both the proposed CA and HA schemes requires less energy than the heuristic scheme in [3]. We also take rate-per-energy as a benchmark. The proposed CA is the best energy-efficient allocation among other algorithms in lower QoS requirements.