LTE-Advanced中離散式接力傳輸資源分配演算探討

Abstract

在新一代通訊系統中，接力資源分配方式將會對效能影響甚巨。接力資源分配是指在網路環境中，將接力點分配給傳輸效能不好的使用者，用以提升這些使用者的傳輸速度。目前為止，已經有許多研究文獻對此議題提出數個利用集中式演算法的解決方法，但這些方式讓基地台必須收集所有網路系統中的通道資訊，造成基地台嚴重負擔，並且未完全契合設計接力點的真正功能---提升原本傳輸效能較差的使用者。 這份研究提出一個基於離散式演算法的解決方案。透過隨機自動學習機制，使用者根據網路環境回傳的效能值，自行選擇適合的傳輸方式。實驗結果證明我們提出的分配演算法，有良好的收斂特性與負載平衡特性，效能方面也十分卓越。

Relay assignment is a crucial issue and it affects the performance of cooperative communication networks, which means assigning the proper relay nodes (RN) to cell-edge users (UE) in order to exploit the spatial diversity through relay nodes and improve cell-edge performance. Several assignment strategies have been proposed in literatures. Nevertheless, the previous works solved this problem by centralized way, where base station (eNB) will serve as a control node to collect the channel conditions and location information and make the final decision. Maximize aggregate performance is a typical objective of centralized assignment strategy to improve system capacity in the network using Hungarian algorithm. Another optimally centralized algorithm, max-min feature, is to maximize the minimum data rate among all users. It is shown to find the optimal objective regardless the initial relay node. Although centralized methods could have better performance for system, it takes high operation and maintenance tasks in eNB, where it infringes the operational requirements of smallcell enhancements. This work is motivated to propose a distributed algorithm, which means that each UE would choose its own appropriate RN individually. To achieve this goal, we proposed a strategy called ”Distributed Relay Assignment by Learning” (RAL) based on stochastic learning automata. Under the new assignment algorithm, not only the local performance is preserved, but also the benefited probability of UEs is raised obviously.