以偽貝氏廣播為基礎的媒體存取控制之協定

Abstract

估計出參與在無線通訊系統中，想要重送封包的使用者數目，對於提升某些分散式媒體存取控制協定系統的效能，是特別重要的。為了解決這個困難的問題，Rivest提出偽貝氏演算法，在穩定的時槽式Aloha系統中，可以不需要知道使用者的數目，簡單又有效地解決上述的問題。他的運作方式是藉由保持住，在每一個時槽開始時，要重送的使用者的估計值。在這篇論文裡，我們著重在以時槽式Aloha為基礎的分散式媒體存取控制之協定中，維持要重送的使用者個數的估計值。最基本的協定形式是從穩定的Aloha開始，接著如果發生某些情況，則系統進入第二模式，第二模式是一個縮短的樹狀分裂的演算法。今過一段有限的時間後，系統會重新回到穩定的Aloha模式下。第二模式是一個提高效能的模式。如同Rivest的偽貝氏演算法，我們提出一個方法，可以不需要知道使用者的個數，並且提高有效傳輸量。

Estimating the number of backlogged users participating in wireless communications is particularly important for enhancing the performance of some distributed MAC protocols. To resolve such a difficult issue, Rivest’s Pseudo- Bayesian algorithm is a simple and effective way to doing so in stabilized slotted aloha without knowledge of the number of nodes. It operates by maintaining an estimate of the backlogged user at the beginning of each slot. In this thesis, we focus on estimating the number of backlogged users for the slotted-aloha-based distributed MAC protocols. A basic form of these protocols is starting from the stabilized aloha and then if certain condition occurs, the system goes to second mode which is indeed a truncated tree splitting algorithm. After a finite period, the system moves back to the mode of stabilized aloha. The second mode could be a performance boost mode. As in Rivest’s Pseudo-Bayesian algorithm, our proposal does not require the acknowledgement of the number of nodes, and increase the throughput.