藍芽 (Bluetooth) 是一種以提供短距離、低功率、低成本且無所不在之無線環境為目標的嶄新無線技術，其 MAC 層資料傳輸頻寬的分配，係由Master 以輪詢 (Polling) 排程方式，動態分配可供封包傳送之時槽 (time slot) 給所有與其連線的Slave 裝置。輪詢期間，若Master沒有資料傳給該Slave，則需送出POLL封包；同理，若Slave沒有資料可回應予Master，則需送出NULL封包。此不帶實際資料的POLL-NULL 封包總共佔用兩個可用時槽。雖然已經有很多學者提出一些演算法來避免拜訪沒有封包待傳的Slave以減少POLL/NULL封包之傳送，但仍有改進的空間。
本論文提出一個全新、有效率且可實際應用於微網 (Piconet) 之輪詢演算法—CRR (Credit-based Round Robin)，根據過去及目前Master與Slave間之封包流量，動態地調整詢問Slave的次數以避免拜訪沒有資料傳送的Slave，降低傳送或接收POLL/NULL的封包數。本論文針對四項主要的課題進行研究並改善，以提升整體網路系統之效能：一、提高系統在單位時間內資料封包的傳送及接收率。二、減少不必要的POLL/NULL封包以避免佔用頻寬。三、降低傳送、接收資料封包之延遲時間。四、避免Starvation及Interruption效應。
模擬結果證明了CRR在不同的slave數及不同的idle slave數均能有效地提升throughput，進而提升資料封包的傳送及接收率以及減少不必要的POLL/NULL封包所佔用的頻寬；並降低封包傳送的End-to-end delay time。此外，CRR預定的Threshold有助於預防Starvation現象，而實驗更證實CRR的獎懲方式有助於減緩Interruption效應。

Bluetooth is an emerging technology for providing short range, low power, low cost, and ubiquitous wireless environments. In a Bluetooth piconet, the Bluetooth device, known as master, dynamically allocates time slots by way of polling other individual Bluetooth devices called slaves. During the polling interval, a POLL packet will be sent if master has no data to send to the slave; while a NULL packet will be sent if slave has no data to send to the master. Both the POLL and NULL packets are of zero payload and each captures one time slot. In order to avoid transmitting POLL/NULL packets, many polling schemes have been proposed by skipping visit to the slaves that might have no data packet to be sent to the master, but they still have much space to be further improved.
In this thesis, we devise a new, efficient and practicable polling scheme—CRR (Credit-based Round Robin). In this scheme, the visit to each slave is dynamically adjusted based on the current and historical traffic between the master and slave to reduce the visit to idle slaves.
CRR improves the prior polling schemes on four key issues：The first is to enhance the packet transmitting rate and packet receiving rate in a unit time. The second is to avoid unnecessary POLL/NULL packets to save bandwidth and power consumption. The third is to reduce the transmission delay. The fourth is to avoid the starvation and interruption effect.
Our simulation results demonstrate that CRR can enhance the throughput among various numbers of active slaves and idle slaves, can improve the transmitting rate and receiving rate of data packets, can reduce the bandwidth by diminishing the visit to idle slaves, and can decrease the packets end-to-end delay. Moreover, to restrict the number of frames exchanged during a single visit to the slave, a threshold is used to prevent the starvation and interruption effect.

中文摘要…………………………………………………………………I
英文摘要……………………………………………………..…………III
目錄...…………………………………………………………………....V
圖目錄………………………………………………………………….VII
表格目錄…………………………………………………………......….X
第一章 簡介……………………………………………………………1
第二章 相關研究…………………………………………………..…15
第三章 CRR輪詢演算法……………………………………..……...20
第3.1節 問題探討………………………………………………..20
第3.2節 CRR之設計動機……...………………………………..28
3.2.1. 針對系統產能之提升……………………………….28
3.2.2. 針對減少POLL/NULL封包對頻寬所造成的浪費31
3.2.3. 針對減少封包之端對端延遲……………………….33
第3.3節 CRR設計概述………………………...………………..35
3.3.1. 參數定義…………………………………………….35
3.3.2. CRR演算法…………………………………………36
第四章 實驗結果與分析……………………………………………..38
第4.1節 隨著Slave連線數及ON/OFF比例變化之Throughput評估…...………………………………………………...43
第4.2節 隨著Piconet中Idle Slave數及ON/OFF比例變化之Throughput評估…………………..……………………48
第4.3節 End – to –end Delay………….………………………....52
第五章 結論…………………………………………………………..54
參考文獻………………………………………………………………..55

圖目錄
圖1-1 Bluetooth的通訊協定階層………………………………..……..1
圖1-2 藍芽使用的79個跳頻頻道……………………………..………2
圖1-3 藍芽微網以及多組微網形成散網的應用實例…...…………….5
圖1-4 一個Scatternet中四種Bluetooth節點的角色………………….6
圖1-5 各狀態監聽Master傳訊的時序狀態圖……..……………..…...8
圖1-6(a) 單一Time Slot的資料傳遞示意圖..…………………………8
圖1-6(b) Multi-slot的資料傳遞示意圖………..………………………..9
圖1-7 Bluetooth之Time Slot配置範例….……………………….........10
圖1-8 Baseband層資料封包格式……………………………………...12
圖1-9 藍芽Mac排程之示意圖.……………………………………….13
圖3-1(a) 一個Piconet由一個Master及七個Slaves所組成………….21
圖3-1(b) Piconet中各佇列流量分佈情形之示意圖………………...21
圖3-2(a) 一個由7個Piconets所組成的Scatternet…………………..24
圖3-2(b) S1的資料欲送達S22所經過的路徑……….………………..24
圖3-3 Slave於Sniff Interval運作的情形……….....………………....31
圖4-1 簡單的NS-2架構…………………...………….……………….39
圖4-2 NS-2 中C++與Otcl 物件架構關係……….…………………..40
圖4-1-1 ON/OFF 比例為1:2的Throughput………………………...45
圖4-1-2 ON /OFF 比例為1:3的Throughput....................................45
圖4-1-3 ON /OFF 比例為1:4的Throughput....................................45
圖4-1-4 ON /OFF 比例為1:5的Throughput…………………........46
圖4-1-5 ON /OFF 比例為1:6的Throughput....................................46
圖4-1-6 ON /OFF 比例為1:7的Throughput……………………....46
圖4-1-7 ON /OFF 比例為1:8的Throughput…………………..…..47
圖4-1-8 ON /OFF 比例為1:9的Throughput…………………….....47
圖4-1-9 ON /OFF 比例為1:10的Throughput………………………47
圖4-2-1 隨著Idle Slave數目改變，ON /OFF 比例為1:2 的Throughput
…………………………………………………………………………..49
圖4-2-2 隨著Idle Slave數目改變，ON /OFF 比例為1:3的Throughput
…………………………………………………………………………..49
圖4-2-3 隨著Idle Slave數目改變，ON /OFF 比例為1:4的Throughput
…………………………………………………………………………..49
圖4-2-4 隨著Idle Slave數目改變，ON /OFF 比例為1:5的Throughput
…………………………………………………………………………..50
圖4-2-5 隨著Idle Slave數目改變，ON /OFF 比例為1:6的Throughput
…………………………………………………………………………..50
圖4-2-6 隨著Idle Slave數目改變，ON /OFF 比例為1:7的Throughput
…………………………………………………………………………..50
圖4-2-7 隨著Idle Slave數目改變，ON /OFF 比例為1:8的Throughput
…………………………………………………………………………..51
圖4-2-8 隨著Idle Slave數目改變，ON /OFF 比例為1:9的Throughput
…………………………………………………………………………..51
圖4-2-9 隨著Idle Slave數目改變，ON/OFF 比例為1:10的Throughput………………………………………...……………………51
圖4-3-1 平均 End-to-end Delay………………………………………53

表格目錄
表 1 各個Master輪詢Slave的順序…………………...……………..25
表2 以t=5之LRR輪詢方法運作所得到的結果.……………………..27
表3 以t=5之LRR輪詢方法，但有效避開Idle Slave之運作所得到的結果……………………………………………………………………..27

[1] Bluetooth Special Interest Group, Specification of the Bluetooth System. Version 1.2, Volume 0, November 2003.
[2] A. Capone, R. Kapoor, and M. Gerla, “Efficient Polling schemes for Bluetooth picocells,” in Proceedings of IEEE International Conference on Communications ICC 2001, vol. 7, (Helsinki, Finland), pp. 1990–1994, June 2001.
[3] Ka Lok Chan, Vojislav B. Mišić, and Jelena Mišić, “Efficient Polling Schemes for Bluetooth Picocells Revisited,” 37th Annual Hawaii International Conference on System Sciences (HICSS'04), Big Island, Hawaii, January 05 - 08, 2004.
[4] Indraneel Chakraborty, Abhishek Kashyap, Anupam Rastogi Huzur Saran, Rajeev Shorey and Apurva Kumar, “Policies for Increasing Throughput and Decreasing Power Consumption in Bluetooth MAC,” IEEE International Conference on Personal Wireless Communications, pp.90-94, 2000.
[5] N. Johansson, U. K