合作式蜂巢網路中之多接力點分配演算法研究

Abstract

合作式通訊為增進蜂巢式網路之空間多樣性（spatial diversity）以及覆蓋率（coverage area）之著名技術。於合作式通訊中，接力點（relay node）將幫助使用者終端裝置（user equipment）向目標傳輸資訊。接力點分配（relay assignment）即是將合適的接力點分配給使用者以達到系統效能的最大化。而以往的接力點分配研究成果皆於匹配（matching）架構下完成，其意義為每一位使用者只能與至多一個接力點連結。因在任一瞬間閒置的使用者皆可被雇用為接力點，匹配架構之系統效能顯然不為最佳解。若在多個接力點可同時服務一位使用者的架構下，雖消耗更多傳輸功率，但系統之總容量（aggregate capacity）可望更有效率地提升。對多接力點分配問題而言，為了最大化系統總容量，在給定的迴圈數裡面，粒子群最佳化（particle swarm optimization）可獲得近似以暴力演算法（exhausted search algorithm）所求得之全局最佳解（global optimal）。隨著使用者與接力點的數量增加，暴力法求解所需要的演算複雜度（computational complexity）呈現指數成長。相對地，當迴圈數目增加，粒子群最佳化所得之解將逼近全局最佳解。而當使用者與接力點數目增加時，獲得此解亦得付出指數成長之粒子與迴圈數量。因此，本論文旨在探討如何開發出在演算複雜度上有效率之多接力點分配演算法。剩餘接力點分配演算法（residual relay node assignment）因此提出以求得線性成長演算複雜度之多接力點分配解。根據在LTE-Advanced通訊系統環境下的模擬結果顯示，隨著接力點數量增加，使用此論文提出之演算法之多接力點支援系統效能可被更有效率地提升。

Cooperative communication is a well-known technique to exploit the spatial diversity and extend the coverage area in cellular networks. In cooperative communication, the relay node (RN) can help user equipment (UE) to transmit its information to the destination. Relay assignment is to assign the proper RNs to the UEs in order to optimize the system performance. The traditional research works on relay assignment are with matching architecture, which yields that each UE can only be matched with single RN. This is obviously a suboptimal solution since the free UEs can be deployed as RNs. With multi-relay assignment architecture described in this work, the aggregate capacity of overall system can be power-efficiently raised at the price of power consumption on originally leisure RNs. For multi-relay assignment problem, to maximize the aggregate capacity or instantaneous information rate, the particle swarm optimization (PSO) is introduced to obtain a close-global optimal solution with given number of iteration instead of exhausted search algorithm, which is with computational complexity of exponential growth as the number of RNs increases when tackling the problem of maximizing the aggregate capacity. As the number of iteration grows, the local-optimally solution of PSO converges to the globally optimum. However, the number of particle and iteration required by PSO to converge to the global optimal solution grows exponentially when the number of RN increases. Therefore, this work aim at finding a computationally efficient algorithm to coordinate a better tradeoff between computational complexity and aggregate instantaneous information rate of each UE. The residual RN assignment algorithm is therefore proposed. The simulation results have shown that as the number of RN increases, the multiple relay-assisted system with the multi-relay assignment algorithm proposed in this work can be power-efficiently raised. The topology is in LTE-Advanced cellular networks. Simulation results have shown that PSO and RRNA provides great performance enhancement comparing with the optimal solution of single-relay assignment with efficient power consumption. Meanwhile, the fairness index is improved in such a resource allocation problem.