聯合封包排程與天線功率控制的無線資料下載

Abstract

在無線環境中，通訊頻道的狀況隨時不斷地改變，如果能利用「使用者多樣性」的特性，便能增加系統的負載量、提高服務品質。QUALCOMM 提出了一種公平演算法(Q-PFA)，然而這種演算法在實際使用是不足的，因此我們提出了一些方法以改進Q-PFA，這些方法包含了設定初值，及觀察使用者的行為以更正確地估計使用者的資料速率等，以達到更高的使用者多樣性增益。我們也提出了一個創新的方法，它是藉由觀察使用者的負載以動態調整指數平滑常數。 在細胞網路中，相鄰的基地台間，其通訊的區堿有部份重疊以使得無線服務的涵蓋範圍更完整。這些重疊使得virtual soft handoff成為可能，但也造成了基地台嚴重地互相干擾而減低資料速率，近年來有人提出可以利用控制天線功率，以減低干擾的程度。在我們的研究中，假設基地台的天線功率是可以控制的，並且提出了一個「機率式功率控制演算法」(PPC)，這個方法首先計算相鄰基地台間的相對負載，並利用這個相對負載來決定天線功率高低的機率。 我們使用電腦來模擬演算法的執行結果，結果顯示我們的封包排程演算法與天線功率控制演算法皆可有效地增進某些效能的指標，如：資料處理量、延遲、暫時性的流量高負載區的行為等。In wireless environment, channel condition varies from time to time. If we can exploit multiuser diversity, the system capacity can be increased and the service quality is improved. To achieve this goal, QUALCOMM’s proportional fair algorithm (Q-PFA) is effective. But it might be insufficient at practical applications for some reasons. We propose viable techniques to improve Q-PFA, which essentially resolve issues related to the loss of multiuser diversity gain from initial values and dynamic ranges of monitored throughputs due to the arrival, absence, presence, and departure of an active user. Additionally, we propose a novel approach to further exploiting the multiuser diversity gain through dynamically adjusting the exponential smooth constant in throughput monitoring. In cellular networks, wireless service coverage is improved through partly radio coverage overlays between neighboring base stations (BSs). For wireless forward link data services, the overlay provides virtual soft handoff opportunities, but also introduces significant inter-call interference which reduces feasible service data rate. Using a power control algorithm to alleviate inter-cell interference is an emerging concept. We assume the BS’s transmitting power is adjustable between full power and low power if necessary and proposed the probabilistic power control algorithm (PPC). PPC first calculates the relative loading of current cell and neighboring cells. The result is used to determine if a BS should transmit at full power according to some probabilities depending on the relative loading. Our simulation shows that both the packet scheduling policies and PPC can remarkably improve some performance features like throughput, latency, and the behavior of temporal traffic hot spot.