多階層行動隨意網路之設計及實作-子計畫四:無線行動隨意網路上之移動支援與電源管理協定(II)

全文

(1)行政院國家科學委員會專題研究計畫. 成果報告. 子計畫四：無線行動隨意網路上之移動支援與電源管理協定 (2/2). 計畫類別：整合型計畫計畫編號： NSC93-2219-E-009-003執行期間： 93 年 08 月 01 日至 94 年 07 月 31 日執行單位：國立交通大學資訊工程學系(所). 計畫主持人：曾煜棋. 報告類型：完整報告報告附件：出席國際會議研究心得報告及發表論文處理方式：本計畫可公開查詢. 中. 華. 民. 國 94 年 8 月 9 日.

(2) 行政院國家科學委員會補助專題研究計畫結案報告無線行動隨意網路上之移動支援與電源管理協定 (2/2). 計畫類別：整合型計畫計畫編號： NSC 93-2219-E-009-003 執行期間： 93 年 08 月 01 日至 94 年 07 月 31 日計畫主持人：曾煜棋. 執行單位：國立交通大學資訊工程研究所. 中. 華. 民. 國. 94. 年. 8. 月. 9. 日.

(3) 摘要此子計畫包含了兩大工作重點，(1)多階層無線隨意行動網上移動之管理和(2)多階層無線隨意行動網上電源之管理。在移動管理方面，我們已知 Mobile IP 是一個被廣泛使用於支援無現網際網路環境的標準，而隨意網路也已提供於多跳躍(Multi-hop)的無線區域網路連上網際網路的可能，所以在移動管理方面，我們建構一個以 Wi-Fi 為基礎的具多步跳躍 (Multi-hop)通訊子網路，作為在多階(Multi-tier)行動隨意網路架構中主要支援行動隨意網路的部分。而為了達到多系統整合與漫遊的目的，我們並提出解決多階行動隨意網路與 Mobile IP 整合在一起的解決方法。在無線網路中電池可說是行動主機的生命來源。如何做好電源管理藉以延長行動主機之運作時間，是無線行動隨意網路能否長期運作的一大關鍵，針對上述之問題我們分別在多階多步跳躍通訊子網路與藍芽通訊子網路在各層協定上提出有效的解決方法，並且對於多階隨意行動網路在整合不同系統時所需考量的省電機制，提出解決良策。此外我們也針對隨意網路提出省電的網路協定，例如繞路協定。. 關鍵詞：無線隨意行動網路，省電模式，IEEE 802.11，藍芽，移動管理，Mobile IP I.

(4) 目錄報告內容. ………………………………………………………………………………… 1. I.. 前言. II.. 研究目的………………………………………………………………………………3. III.. 研究方法…………………………………………………………………………… ...3. …………………………………………………………………………………… 1. A. 無線行動隨意網路上的移動管理協定 – 結合 Mobile IP 與隨意網路 B. Bluetooth 多個 piconets 共存之干擾模型 C. Bluetooth 系統上電源模式之管理 D. IEEE802.11 為基礎之無線隨意多跳躍(Multi-hop)網路上的電源管理協定 E. 無線行動隨意網路上以電源為考量之通訊協定 IV. 研究成果………………………………………………………………………………9. 附錄……………………………………………………………………………………………10 I. 國外差旅費使用情形………………………………………………………………10 II.. 發表論文全文. ……………………………………………………………………11. II.

(5) 報告內容 I. 前言在多階層無線隨意行動網路上，由於行動主機具有可任意移動的特性，因此如何使行動主機的使用者感受不到移動性的影響，而能如同在有線的環境中一樣，以相同的方式運作及接受服務，是無線行動隨意網路能否成功的一大關鍵。計劃第一部分著重於如何掌握行動通訊裝置目前之位置與狀況，使得通訊模式、資料傳輸率可視情況彈性切換，以節省資源並達最大資料傳輸量。因此本子計劃的其中一個目標便是設計適用於隨意網路的移動管理方法，我們整合了 Mobile IP 與隨意網路，使得在隨意網路下的行動主機感受不到因為移動所產生的影響。 Mobile IP 的作用，簡單地說，就是讓同一台電腦，從原本所在的網路移至另一個網路時，可使用原來的 IP address。我們會在研究方法中去敘述，我們如何去結合 Mobile IP 與隨意網路，以達成移動管理的目的。此外多階層無線隨意行動網路上電源之管理也是此子計劃的另一個目標，我們知道無線隨意行動網路是由行動主機所構成之網路，對於所有的行動主機而言，電池是其唯一的動力來源，沒有電源，行動主機將無法運行。然而，在可預期的近幾年之內，電池科技將不會有很大的突破，因此，如何延長電池的使用時間，對於無線隨意行動網路而言，是一個十分重要的課題。本計劃在電源管理方面之主要目標之一，是在以 IEEE 802.11 為基礎之無線隨意行動網路上，研究省電模式之管理。當行動主機進入省電模式後，行動主機必須要去預測其他行動主機醒來的時機，以便彼此可以傳送或接收訊息，然而，由於沒有中央控管的機制，再加上因為不可預期之移動性與無線電干擾造成的封包延遲，使得無線隨意行動網路上的同步難以逹成，也因此讓這項預測的工作變得十分複雜而且困難。除此之外，在無線隨意行動網路上設計省電模式的另一個挑戰，即是行動主機可能無法察覺其鄰居的存在，會發生這個現象的主要原因，是因為彼此醒來時間的不相同，再加上傳送與接收訊息的次數也減少了所造成的，而不精確的鄰居資訊可能會使得繞徑協定發生錯誤。為了解決上述之問題，我們以發送更多的 beacon、讓行動主機定期醒來的時間有所重疊、以及醒來時間的預測等策略，設計了三種省電協定，分別為： dominating-awake-interval protocol、periodically-fullyawake-interval protocol 與 quorum-based protocol，其中，每種省電協定其醒來的模式都不相同，但都能保證不同步的行動主機的醒來時間可以相互重疊。本計劃的另一個目標，即是去探索在藍芽規格裡一種省電模式 – sniff mode。在這個模式下，一個角色為 slave 的移動設備，只要定期醒來接收 master 端的資料，而不須時時刻刻待命，浪費無謂的電池能量。然而，一個重要且具挑戰性的問題是，master 該如何去排程所有. 1.

(6) 在子網內 slaves 的睡眠週期，使得在省電的同時，還能兼顧網路傳輸效能，這樣的一個問題，我們稱之為”省電排程問題”。針對這個問題，我們提出一個可調式的排程機制，動態調整每個 slave 的睡眠參數，使其符合每個 slave 與 master 之間經常改變的不對稱交通流量需求，並逹到 slave 端的省電目標。此外既有的通訊協定(protocol)大部分都只在某些層上去加進省電的功能，例如實體層 (physical layer)、媒體擷取層(MAC layer)、網路層(network layer)，但並非全部。所以我們也發展出一套整合各層的能量感知(power aware)、位置感知(location aware)的通訊協定，以達成隨意網路的電源管理。. 2.

(7) II. 研究目的本子計劃的主要目的為提供無線行動隨意網路的移動管理與電源管理，以細項來說明的話，本計畫的目的包括：（一）整合 Mobile IP 與隨意網路，設計無線行動隨意網路下的移動管理協定。（二）Bluetooth 多個 piconets 共存之干擾模型（三）IEEE 802.11 為基礎之無線隨意行動隨意網路上的省電協定。（四）藍芽網路中省電模式之可調式排程機制。（五）整合網路各層的能量感知(power aware)、位置感知(location aware)的通訊協定，以達成隨意網路的電源管理。. III. 研究方法 A. 無線行動隨意網路上的移動管理協定 – 結合 Mobile IP 與隨意網路在移動管理的部份，我們實作了一個架在 Layer 3(Network Layer)以及 Layer 2(Datalink Layer) 的無線隨意網路通訊協定。並加上 Mobile IP 的機制，使它可以支援跨大到一整個網路的移動管理。 Mobile IP 的作用，簡單地說，就是讓同一台電腦，從原本所在的網路移至另一個網路時，可使用原來的 IP address。我們使用 Mobile IP 來做子網路之間的資料封包傳遞，對於子網路之內的資料封包傳遞，我們則用 Ad Hoc DSDV protocol 來做轉送。如下圖所示，Network1 和 Network2 分屬於不同的子網路，兩者之間的資料傳送靠 Mobile IP 來服務，而子網路之內的資料傳送則使用 Ad Hoc DSDV protocol 的服務。 Internet IBM Compatible. Ethernet. Gateway/Agent network1. Gateway/Agent IBM Compatible. IBM Compatible. network2. Laptop computer. Laptop computer Laptop computer. text. Laptop computer. Laptop computer Laptop computer. Laptop computer. 3. Laptop computer.

(8) B. Bluetooth 多個 piconets 共存之干擾模型在藍芽(Bluetooth)操作的頻帶(operating band)上，藍芽微網(piconet)間會互相干擾. 我們研究在多個微網共存的環境（multi-piconet environment）中，試著從數學機率分析的角度，來評估因為共同頻道干擾（co-channel interference）而導致封包碰撞（packet collision）所造成對整體網路效能的影響程度，以供藍芽網路佈署時有用之參考依據。在我們的分析中，考慮 N 個微網同時存在，且彼此間的距離非常接近，足以互相影響。因此，每一個微網有 N-1 個可能影響到其傳送的對手。而在任意的時間點上，若有兩個微網欲使用同樣的頻率作傳輸用，則視此傳送的封包被損壞。並假設λ1, λ3, λ5 為 1-slot, 3-slot , 5-slot 之封包的 arrival rates 。由於在多時槽的封包中，只有第一個時槽算入 arrival，則顯然λ1+3λ3+5λ5≦1。再將未使用的時槽視為許多單時槽(single slot)的封包，以λ0 代表其所佔的比率，則λ0=1-(λ1+3λ3+5λ5)。現考慮環境中同時存在兩個微網 X 與 Y，欲分析 X 所受到的干擾，推倒出 X 中，長度為 i 的封包之傳送成功率 Ps(i)，其中 i = 1, 3 ,5。接著我們將引入時槽分隔線(slot delimiter)的觀念來分析。就 X 中的任一時槽做考慮，由於 X 和 Y 不同步，則 Y 中可能有 1 或 2 個 slot delimiters 可能跨越(cross)此 X 的時槽。由於我們考慮的是連續的機率，此後的討論會將兩個 slot delimiters 跨越的機率忽略。以 X 中的一個 1-slot 的封包為例：此封包唯有在 slot delimiters 前後的封包接不造成干擾的情況下方能成功。故成功率依 Y 微網的傳送與否，可能為 1 ,(78/79)或(78/79)2。下圖為 slot delimiter 之示意圖：. i. B1,B3,B5:分別為為 1-slot, 3-slot, 5-slot 的封包的開頭。 ii. B3,B4:分別為 3-slot 的封包中，第二與第三時槽的開頭。 iii. B6,B7,B8,B9：分別為 5-slot 的封包中，第二，三，四，五時槽的開頭。 iv. B10：為未使用之時槽的開頭。顯然，B1 所出現的比率為λ1，B2,B3,及 B4 出現的比率皆為λ3；B5,B6,B7,B8,B9 出現，比率機各為λ5，而 B10 出現的機率為λ0。以函式的方式來表示 Bj 則得:λ(Bj)j=1 ..10。 X 中的一個封包，若被 Y 中一個類型 B1/B2/B5 為 delimiter 跨越(cross)，則 delimiter 前後各有一個封包，則 Y 中即有兩個封包有可能對 X 造成影響，反之，若是其他類型的 delimiter，則 Y 中只有一個封包會對 X 造成影響。在推導成功率 Ps(i)時，我們先推出一個特殊情況 Ps’ (i)。Ps’(i)只代表兩種情況的成功傳送比率。(1)Y 中第一個 delimiter 跨越 X 封包的型態為 B1/B2/B5/B10。(2)第一個 delimiter 前面的所有時槽並未發生干擾的情況。依此定義我們可以推導出 Ps(i)： 10. PS (i ) = ∑ λ ( B j ) • f ( j ) • PS ' (i − g ( j )) j =1. 其中 P0=(78/79) 4.

(9) 上面的式子中，我們定義 g(j)為 delimiter Bj (j=1..10)後剩餘的時槽數目。例如:g(1)=1,g(3)=2,g(7)=3 …。以下為推導的概念：λ(Bj)是第一個 delimiter 為 Bj 的機率。f(j) 是 X 在 delimiter 前後的封包都沒有發生碰撞(collision)的機率。在此之後，X 尚有 i-g(j)的時槽需要傳送，且對這 i=g(j)的時槽來看，第一個 delimiter 必為 B1/B2/B5/B10 中的一種。故我們可以套用 Ps`(i-g(j))。下圖為上述概念的例子：. (a)圖，X 中一個 3-slot 的封包，第一個 delimiter 為 B1，首先我們知道 X 中第一個時槽成功傳送的機率為 f(1)，則剩下的兩個時槽的成功機率為 PS’(2)，故對整個封包而言，成功的機率為兩者相乘，即 f(1) × PS’(2) (b)圖為一個 5-slot 的封包，所遭遇到的第一個 delimiter 為 B3，同理，其傳送成功的機率為 f(3) × PS’(3) 再者，可以用遞迴(recursive)方式來表示 Ps’(k)如下：. PS ' (k ) =. λ0. λ0 + λ1 + λ3 + λ5. λ3. λ0 + λ1 + λ3 + λ5. × PS ' (k − g (10)) +. × PS ' (k − g (2)) +. λ1. λ0 + λ1 + λ3 + λ5. λ5. λ0 + λ1 + λ3 + λ5. × PS ' (k − g (1)) +. × PS ' (k − g (5)). 而我們假設在 K≦0 時，Ps’(K)=1。這可以想做假設第 0 個時槽，或更早以前的時槽失敗的機率為零。上述的推導為兩個微網的環境，當環境有 N 個微網同時啟動的狀況下，延伸上述的推導，此時對微網 X 而言，有 N-1 微網成為干擾源(interference source)。由於這 N-1 個干擾源各自獨立，X 中第 i 個時槽的成功機率為 Ps(i)N-1 故單一網路 X 的生產量(throughput) T 為：. T=λ1 × Ps(1)N-1 × R1 +3 ×λ3 × Ps(3)N-1 × R3 + 5 ×λ5 × Ps(5)N-1 × R5 其中 R1, R3 和 R5 分別為 1-slot, 3-slot, 5slot 封包各自的資料負載量，其單位為(位元/時槽)，而整個網路的生產量(throughput)為 N × T。. C. Bluetooth 系統上電源模式之管理藍芽技術(Bluetooth)在智慧家庭(Smart Homes)是一個很重要的元件，在智慧家庭(Smart Homes)中，大部分的藍芽裝置都是可攜帶(portable)且裝電池(battery-operated)的，所以省電 5.

(10) (power saving)是個很重要的議題。我們研讀了藍芽技術管理的低耗電監聽模式(sniff mode)，從屬裝置(slave)是被允許可以週期性醒來(awake)的，一個很重要的議題就是去解決如何分配(schedule)每個從屬裝置的週期，使得網路流量需求(traffic requirement)和使用電量需求(powering saving requirement)能達成較好的平衡。我們發展出了一個調整監聽模式相關參數之策略(由於在藍芽規格裡主裝置與從屬裝置可以透過一些控制封包（control packet）來協調從屬裝置進入監聽模式之運作參數（sniff parameters），然而規格裡並無明確交代該如何設計這些參數)。讓主裝置(master)與從屬裝置 (slave)可以根據目前的數據流量(traffic)，在不犧牲工作效能的條件下，動態協調出一組合適的監聽參數，使得從屬裝置(slave)能進入低功率之監聽模式(sniff mode)以節省不必要的電池電源消耗，達到網路生命週期、效能提昇之目的。. 上圖顯示出我們監聽排程(sniff scheduling)的協定，主裝置有三個元件，評估器 (Evaluator)、排程運算器(Scheduler)、時槽資源庫(RP(resource pool))，評估器的工作是去評估衡量從屬裝置使用目前的監聽參數有沒有效率，如果不適合，會去觸發排者運算器重新調整監聽參數。時槽資源庫負責管理可用的監聽時槽(sniff-attempt slots)，這些可用的時槽可以給從屬裝置來使用。主裝置會週期性對每個從屬裝置執行評估器來評估狀況，如果有需要的改變的話，會產生一個訊息通知排程運算器，然後排程運算器會去時槽資源庫尋找適合這個從屬裝置的新的監聽參數。跟其它已有的協定比較，我們的排程考慮同時有多個從屬裝置在運作，而且我們的排程根據情況做更精確、更動態的調適。. D. IEEE802.11 為基礎之無線隨意多跳躍 (Multi-hop) 網路上的電源管理協定我們提出以下的策略來設計 multi-hop 隨意行動網路上的省電協定：. 1.. 2.. 發送更多的 beacon：為了避免鄰居資訊不精確的問題，我應該讓進入省電模式的行動主機多發送 beacon。因此，在每個 beacon interval，每個行動主機都會發送一次 beacon，即使別的行動主機已經成功發送 beacon。讓行動主機定期醒來的時間有所重疊：我們將重新設計行動主機的睡眠模式，使得行動主機即使在不同步的種況下，也能保證彼此醒來的時間相互重疊，因而讓彼此都能 6.

(11) 3.. 收到對方的 beacon。醒來時間的預測：當收到鄰居的 beacon 後，我們便可知道鄰居的存在，同時也可知道鄰居的時脈。此時，便可依據彼此的時間差來預估鄰居醒來的時間，以便在鄰居醒來時傳送訊息給對方。在此，我們只是去預測鄰居的醒來時間並不做同步的動作。. 依據以上的策略我們提出了三種省電協定，每種省電協定，其醒來的模式都不相同，但都能保證不同步的行動主機的醒來時間可以相互重疊。此外，我們也重新設計了 beacon interval 的結構。在每個 beacon interval 中包含了三個 window，即 beacon window、MTIM window 與 active window。Beacon window 是讓每個行動主機發送 beacon 的，MTIM window 是讓行動主機發送 MTIM(Multi-hop TIM)訊框的，MTIM 訊框是用來通知其他行動主機有資料封包要傳送給它，在 active window 中，行動主機會保持清醒，以便接收其他行動主機重送之訊息。在不同的省電協定中，其 beacon interval 的結構也會略有不同，我們將會在後面說明之。以下我們將分別說明三種不同的省電協定：. 1.. Dominating-Awake-Interval 協定要讓進入省電模式且不同步的行動主機醒來的時間互相重疊，最簡單的方法就是讓行動主機在每個 beacon interval 醒來的時間超過一半。但是，如此並不足以保證彼此都能收到對方的 beacon。因此，我們必須調整 beacon interval 的結構。我們將 beacon interval 分為奇數週期與偶數週期。在奇數週期 beacon interval 由 beacon window 開始，接下來是 MTIM window 與 active window；在偶數週期，則是由 active window 開始，接下來是 MTIM window 與 beacon window。如此一來，便可保證進入省電模式且不同步的行動主機彼此都能收到對方的 beacon。. 2.. Periodically-Fully-Awake-Interval 協定如果我們讓進入省電模式的行動主機每 T(T 是自然數)個 beacon interval 完全醒來一次，也可以保證進入省電模式且不同步的行動主機彼此都能收到對方的 beacon。當行動主機進入省電模式後，每三個 beacon interval 便會醒來一次，如此一來，每三個 beacon interval 便可收到鄰居的 beacon 一次。. 3.. Quorum-Based 協定我們依據分散式系統 quorum 的概念設計了這個省電協定，如上圖所示，我們將 n2 個 beacon interval 分為一群，每個進入省電模式的行動主機從 n2 個 beacon interval 中任意挑選一行與一列的 beacon interval 完全醒著，如此也能保證進入省電模式且不同步的行動主機彼此都能收到對方的 beacon。. 當我們能保證進入省電模式且不同步的行動主機彼此都能收到對方的 beacon 後，行動主機便可依據上述三種省電協定的清醒模式來預測鄰居醒來的時間，因而便能順利的傳送 7.

(12) 資料給對方。因此，所有的行動主機都進入省電模式，各種繞徑協定仍能正常運作。. E. 無線行動隨意網路上以電源為考量之通訊協定既有的通訊協定(protocol)大部分都只在某些層上去加進省電的功能，例如實體層 (physical layer)、媒體擷取層(MAC layer)、網路層(network layer)，但並非全部。所以我們發展出一套整合各層的能量感知(power aware)、位置感知(location aware)的通訊協定。我們的方法是利用行動裝置的位置資訊來達成所有協定層(protocol layers)的能量管理 (power conservation)，這些協定層包含實體層(physical layer)、媒體擷取層(MAC layer)、網路層(network layer)，是個完全能量感知(power aware)和位置感知(location aware)的通訊協定。類似手機網路(cellular network)，我們的方法首先也是把網路區域分割成正方形或六角形(hexagons)稱作方格(grid)，這個方法可以使得管理電源和管理移動性(mobility)變得比較容易處理。每個行動裝置要裝上全球衛星定位系統(GPS)使得可以分割網路區域變得一個個的方格(grid)。在路由層(routing layer)，為了選擇良好的路由路徑(routing path)，我們為每個方格(grid) 定義了一個聚集能量等級(collective energy level)，這個等級表示方格內所有可用電量的總和，等級愈高表剩餘電量愈多，路徑會依照這個等級選擇達成省電。電源管理(power management)方面，基本的概念，只有一些行動裝置需要醒來(awake mode)，其它的行動裝置可以進入休眠狀態(sleep mode)，可以達成省電的效果。我們的協定中，為了省更多的電，當在維持網路連接性(network connectivity)時，有一個電源管理機制 (power mode management mechanism)來減少作用中的行動裝置(awake hosts)，在方格內只有電量最高的醒著，這個裝置稱作領導者(leader)，其它都進入休眠模式。方格間的通訊是領導者跟領導者直接通訊。而在方格內通訊，領導者(leader)必需負責<1>維持本地方格的路由表(local grid’s routing table)。<2>轉送 RREQ、RREP、資料封包(data packets)。<3>維持監看本地格內其它的行動裝置。<4>暫存已經在休眠中的行動裝置的封包。當領導者(leader)要離開或電量不足時，必需從方格內選出一個新的行動裝置來當領導者(leader)。在媒體擷取層(MAC layer)，避免不必要的碰撞，我們把方格內和方格間的通訊用時間分開，隨著時間的不同來決定是否是屬性方格內通訊時間或方格間通訊時間。最後，在實體層(physical layer)我們的電源控制會隨著行動裝置的位置來做調整。因此我們的方法是結合了網路的各層，整體的做考量，以達成於隨意網路下電源管理的目的。. 8.

(13) IV. 研究成果本計畫依時程順利進行，並獲致豐碩成果，相關論文發表如下，論文全文詳如附錄：. [1] Y.-C. Tseng, C.-C. Shen, and W.-T. Chen “Mobile IP and Ad Hoc Networks: An Integration and Implementation Experience”, IEEE Computer, Vol. 36, No. 5, May 2003, pp. 48-55. (SCI, EI) [2] T.-Y. Lin and Y.-C. Tseng, “Collision Analysis for a Multi-Bluetooth Picocells Environment”, IEEE Communications Letters, Vol. 7, No. 10, Oct. 2003, pp. 475-477. (SCI, EI) [3] T.-Y. Lin and Y.-C. Tseng, “An Adaptive Sniff Scheduling Scheme for Power Saving in Bluetooth”, IEEE Wireless Communications, Vol. 9, No. 6, Dec. 2002, pp. 92-103. (SCI, EI) [4] Y.-C. Tseng, C.-S. Hsu, and T.-Y. Hsieh, “Power-Saving Protocols for IEEE 802.11-Based Multi-Hop Ad Hoc Networks”, IEEE INFOCOM, 2002. [5] Y.-C. Tseng and T.-Y. Hsieh, “Fully Power-aware and Location-aware Protocols for Wireless Multi-Hop Ad Hoc Networks”, ICCCN, 2002. [6] Y.-C. Tseng, C.-S. Hsu, and T.-Y. Hsieh, "Power-Saving Protocols for IEEE 802.11-Based Multi-Hop Ad Hoc Networks", Computer Networks, Elsevier Science Pub., Vol. 43, No. 3, Oct. 2003, pp. 317-337. (SCIE, EI). 9.

(14) 附錄 I. 國外差旅費使用情形本計畫之國外差旅費使用兩次，，兩次會議的行程如下: z. 林致宇，參加 IEEE BroadNets(First Annual International Conference on Broadband Networks)會議，2004/10/25 ~ 2004/11/1，San Jose，USA . 這一個研討會主要探討的研究主題是集中在寬頻網路(Broadband Network)的範疇，討論的議題包含了光學網路(Optical Network)的發展、無線網路(Wireless Networks)的發展及無線感測網路(Wireless Sensor Network)等等不同的領域，這些領域跟本子計劃的議題也是非常相關的。我們針對感測網路上的追蹤(Tracking)問題提出了一個省電的追蹤方法，並在會議中發表。. z. 王友群，參加 IEEE WICON (First International Conference on Wireless Internet)， 2005/7/10 ~ 2005/7/15，Budapest，Hungary . 本次會議所涵蓋的議題主要圍繞在無線網路中，包含有媒體層及網路層的協定設計(MAC Protocols & Routing Protocols)、基地台換手問題(Handover Issue)、無線感測網路(Wireless Sensor Networks)、網路傳輸品質議題(Quality of Services，QoS)、下一世代的無線網路服務(Next Generation Wireless Services)等相關議題。我們針對感測器佈置(Sensor Deployment)的議題提出一套節省成本的方法，並在本次會議中發表與討論。. 詳細的報告於附於下頁，會議中發表的論文則附在附錄 II。. 10.

(15) □ □ □ □. 赴國外出差或研習赴大陸地區出差或研習出席國際學術會議國際合作研究計畫出國. 心得報告. 計畫名稱無線行動隨意網路上之移動計畫編號支援與電源管理協定報姓. 告. 人名. NSC 93-2219-E-009-003-. 服務機構國立交通大學資訊工程及職稱學系博士班學生會議/訪問時間第 1 屆一年一期國際寬頻網路研討會 (First Annual International 地點 Conference on Broadband Networks) 2004/10/27~2004/10/19, San Jose, USA 會議名稱第 1 屆一年一期國際寬頻網路研討會(Broadnets 2004) Structures for In-network Moving Object Tracking in Wireless Sensor 發表論文題目 Networks. 林致宇. 一、參加會議經過：這一次研討會的舉辦地點是位於美國加州(California)聖荷西市(San Jose)的希爾頓飯店，會議中除了個人報告(Oral)以及張貼海報(Poster)外，還有邀請知名人士針對寬頻網路的發展做專題演講。我個人報告的時間是會議的第三天(當地時間 10/28)，這個 Session 是針對無線感應網路 (Wireless Sensor Network)做討論的子會議，每個人約有 15 分鐘左右的時間可以報告自己所做的研究，並預留 5 分鐘的時間讓其他的聽眾得以詢問問題。在這個子會議中，大家談論的議題包括在無線感應網路中感應器的佈置(Placement)方法、在感應網路中如何快速地形成一個個區域的群組(Local Clusters) 、感應器如何藉著週期性的關掉電源以達到節省能源的目地等，讓我對這個領域有著更深入的了解。二、與會心得：這一次研討會主要探索的方向是集中在寬頻網路(Broadband Network)的範疇，討論的議題包括有光學網路(Optical Network)的發展、無線網路(Wireless Networks)的發展及無線感應網路(Wireless Sensor Network)等等不同的領域，由於這些領域和我目前的研究習習相關，因此藉由聆聽別人對這些領域的研究報告，讓我了解到目前別人在這些領域中是朝哪些方向努力、並且發掘到哪些問題，相信這對我目前的研究有著極大的幫助。此外，藉由這次出國的機會，讓我得以和不同國家的人們交談並交換心得，除了能夠加強自己的語文能力外，也增加了自己的國際觀。三、建議與結語：在這次的會議中，有幾個與會的人員於會後針對我所做的報告詢問相關的問題，而我也於會後去和別的報告者交換心得，我覺得這是一個很不錯的經驗，我會建議學校或相關研究單位除了鼓勵學生出國於國際會議報告之外，更應鼓勵學生積極與其它學者做交流，這樣應更能提升出國報告的收穫。 11.

(16) 四、攜回資料： BROADNETS 2004 會議手冊：記載本研討會會議流程、報告人員、地點、及報告題目等與研討會相關的會議手冊，其中並包含了會議中所有的論文集。五、其他：無. 12.

(17) □ □ □ □. 赴國外出差或研習赴大陸地區出差或研習出席國際學術會議國際合作研究計畫出國. 心得報告. 計畫名稱無線行動隨意網路上之移動計畫編號支援與電源管理協定報姓. 告. 人名. NSC 93-2219-E-009-003-. 服務機構國立交通大學資訊工程及職稱學系博士班學生會議/訪問時間第 1 屆無線網際網路國際研討會(WICON 2005)，7/10 日-7/15 日在匈牙地點利布達佩斯(Budapest)-維斯格蘭德(Visegrád)的 Danubius Spa & Conference Hotel Visegrád 旅館舉行會議名稱第 1 屆無線網際網路國際研討會(WICON 2005) Efficient Deployment Algorithms for Ensuring Coverage and Connectivity 發表論文題目 of Wireless Sensor Networks. 王友群. 一、參加會議經過：本次研討會的舉辦地點是位在匈牙利布達佩斯(Budapest)-維斯格蘭德(Visegrád)的 Danubius Spa & Conference Hotel Visegrád 旅館，會議中除了以口頭報告方式的技術性會議(technical session)外，還邀請學術及產業界的知名人士就相關議題發表主題演說 (Keynote Speech & Tutorials)。我個人報告的時間是會議的第三天下午(當地時間為7/12)，是歸類在無線感測網路的子會議(Wireless Sensor Networks)中。該子會議的組別共有4組，每組由一人代表報告，每個人報告的時間約有20分鐘，並且預留5分鐘的時間讓其他與會聽眾發問。在我參加的這個子會議中，大家所討論的議題包含有感測器 (Sensors) 的部署 (Deployment)方式、網路傳輸資訊匯集(Data Aggregation)技術、以及省電的協定設計等。藉由別人的報告以及大家共同討論的問題當中，不僅讓我充分了解到自己研究的改進方向及空間、其他研究者在這個議題的研究方向、也同時讓我對這方面的領域有更進一步的理解。二、與會心得：本次會議所涵蓋的議題主要圍繞在無線網路中，包含有媒體層及網路層的協定設計(MAC Protocols & Routing Protocols)、基地台換手問題(Handover Issue)、無線感測網路(Wireless Sensor Networks)、網路傳輸品質議題(Quality of Services，QoS)、下一世代的無線網路服務(Next Generation Wireless Services)、以及其他相關議題等。我個人除了參加我本身所報告的子會議外，也同時參加了幾個我比較感興趣的子會議。藉由聆聽別人的技術報告以及詢問相關問題，讓我對其他領域目前的發展更有概念，也大致了解其他人的研究方向。此外，在我個人報告的部分當中，不僅讓我吸取了許多寶貴的經驗，在報告過程中以及報告之後、聽眾們的問題以及建議，也讓我對目前的研究方向有著更進一步的 13.

(18) 想法。另外，藉由出國參加國際研討會，不但能加強自己的語文能力外，與不同國家人們的討論當中，也能夠培養國際觀。三、建議與結語：我個人認為，能夠參加這次的國際研討會，對我來說可以是一個非常寶貴且收穫良多的經驗。因此，我建議學校能夠多鼓勵學生出國在國際研討會上報告自己的研究成果。這樣不但能夠開擴學生的視野、培養學生的國際觀、增進學生的語文能力、此外也能提升學校在國際舞台上的地位。四、攜回資料： WICON 2005 會議資訊：記載本次研討會的時間、地點、會議流程、報告人員、以及報告題目等與研討會相關的會議資訊。附件一：WICON2005網址 (http://www.wicon.org) 附件二：飛機行程(為當地時間) 7/9 22:00 台北 7/10 06:00 維也納 7/10 10:20 維也納 7/10 11:10 布達佩斯 7/20 09:00 布拉格 7/20 10:05 維也納 7/20 11:05 維也納 7/21 06:30 台北. 出發抵達出發抵達出發抵達出發抵達. 14.

(19) II. 發表論文全文. 15.

(20) 論文名稱：Mobile IP and Ad Hoc Networks: An Integration and Implementation Experience 作者：Y.-C. Tseng, C.-C. Shen, and W.-T. Chen 發表情況: IEEE Computer, Vol. 36, No. 5, May 2003, pp. 48-55. (SCI, EI).

(21) 1. Mobile IP and Ad Hoc Networks: An Integration and Implementation Experience Yu-Chee TsengÝ , Chia-Ching ShenÝ , and Wen-Tsuen ChenÞ Ý Department of Computer Science and Information Engineering National Chiao Tung University Hsin-Chu, 30050, Taiwan Email: yctseng, jcsheen@csie.nctu.edu.tw Þ Department of Computer Science National Tsing Hua University Hsin-Chu, 30050, Taiwan Email: [email protected]. Abstract— Mobile IP has been widely accepted as a standard to support IP mobility in a wireless Internet environment to keep a session connected when a mobile host roams from subnet to subnet. Another emerging wireless network architecture that is gaining more and more popularity is the mobile ad hoc network (MANET), which can be flexibly deployed in almost any environment without the need of infrastructure base stations. In order to move to an all-IP environment, there seems to be a growing demand to integrate these two architectures together. Typically, mobile hosts are served by access points that can connect to them directly (in one hop). In this paper, we propose to extend access points to multiple MANETs, each as a subnet of the Internet, and discuss how to support Mobile IP in such environment. Such integration is beneficial to both societies. From Mobile IP’s prospective, Foreign Agents’ service areas are not limited to hosts within a single (wireless) hop any more. From MANET’s prospective, mobile hosts can immediately enjoy tremendous services already existing on the Internet through Mobile IP. This article reports our integration and implementation experience based on IEEE 802.11b wireless LANs. Issues such as overlapping of MANETs, dynamic adjustment of mobile agents’ service coverages, support of local broadcast and various communication scenarios are addressed. Discussion also covers required adjustments of Mobile IP to support such architecture. Index Terms— ad hoc network, mobile computing, Mobile IP, mobility management, routing, wireless Internet.. I. I NTRODUCTION Wireless communications and mobile computing are gaining more popularity in recent years. Wireless communication devices have become standard features in most portable computing devices, such as laptops, PDAs, and. handsets. People are becoming used to carrying computers while traveling around to enjoy the tremendous services on the Internet. Ubiquitous computing has added a new feature, mobility, to the world of computing and communications. We have observed two strong growths of interests related to this trend. The first one is Mobile IP [15], which supports mobile hosts roaming from subnet to subnet without need of changing IP addresses. Mobile (home and foreign) agents are used to support seamless handoffs. The next generation IPv6 will include features of Mobile IP as inherent functionality. Another emerging wireless network architecture is the mobile ad hoc network (MANET), which can be flexibly and conveniently deployed in almost any environment without the need of infrastructure base stations. MANETs have received intensive attentions recently [6], [11], [18], [20]. In the literature, most works are based on IEEE 802.11-like network interface cards to build a MANET. The recently proposed wireless sensor networks also have a similar architecture to the ad hoc networks. In the trend of moving to an all-IP environment, there seems to be growing demand to integrate these two architectures together. In this paper, we propose to extend the typical wireless access points to multiple MANETs, each as a subnet of the Internet, and discuss how to support Mobile IP in such environment. Such integration is beneficial to both societies. From Mobile IP’s prospective, Foreign Agents’ service areas are not limited to hosts within a single wireless hop any more. From MANET’s prospective, mobile hosts can immediately enjoy tremendous services already existing on the Internet without worrying about disconnection due to mobility. With such combination,.

(22) 2. macro mobility is supported by the former, while micro MH s Home mobility is supported by the latter. network This article reports our integration and implementation CH experience based on IEEE 802.11b wireless LANs. VariHA ous routing scenarios involving Mobile IP and MANET are discussed. One fuzzy area is the possibility of Data packet some MANETs overlapping with each other to form a Internet larger MANET (which is sometimes inevitable), makEncapsulated header Data packet ing the service boundaries of home/foreign agents vague. FA The dynamics of MANETs also necessitates redefining the service coverage of AGENT ADVERTISEMENT and Foreign Data packet AGENT SOLICITATION messages in Mobile IP so as to network MH adapt to constant topology changes of MANETs. The support of local broadcast and various communication scenarios and issues like TTL, ARP, and registration, are ad- Fig. 1. The transmission scenario of Mobile IP. dressed. Discussions also cover required adjustments of Mobile IP to support such architecture. Four roles are defined in Mobile IP: mobile host (MH), corresponding host (CH), home agent (HA), and foreign agent (FA), as illustrated in Fig. 1. A MH is a host or a II. P RELIMINARIES router that may change its point of attachment from subA. Mobile Ad Hoc Networks (MANETs) net to subnet. When a CH, which is a host in the Internet, A MANET is a network consisting of a set of mobile sends an IP datagram to a MH, it will be delivered to the hosts which may communicate with one another and roam MH’s home network. When the MH is away from home, around at their will. A routing path may consist of a the datagram will be tunneled to the foreign network. The sequence of wireless links without passing base stations HA will encapsulate the datagram with an IP header car(i.e., in a multi-hop manner). This requires each mobile rying FA’s IP address or the MH’s co-located CoA (carehost to serve as a router. Applications of MANETs oc- of-address). In case of using FA’s address, the FA should cur in situations like battlefields, outdoor assemblies, and de-capsulate the datagram and forward it to MH. CoA can emergency rescues, where base stations or fixed network be dynamically obtained as a temporary address, through infrastructures are not available, but networks need to be such as DHCP address configuration procedure. If the codeployed immediately. located CoA is used, the MH itself serves as the endpoint Extensive efforts have been devoted to the routing is- of the tunnel and performs decapsulation locally. In our sues on MANET. Routing protocols can be classified as implementation, we follow the former solution. proactive and reactive. A proactive protocol (such as the HA and FA need to advertise their services by periodDSDV protocol [12]) constantly updates routing informa- ically sending AGENT ADVERTISEMENT. A MH untion so as to maintain a (close to) global view on the aware of any local mobile agent may inquire by sending network topology. On the contrary, a reactive protocol AGENT SOLICITATION. From time to time, MH needs searches for a path in an on-demand manner. This may to register with its HA its current CoA. HA keeps track be less costly than a proactive protocol when host mobil- of the mapping between each residential MH’s permanent ity is high. Representative reactive protocols include DSR address and its CoA in a location dictionary (LD). Further [5], ZRP [3], CBR [4], and AODV [11]. A review of uni- extensions of Mobile IP also exist, such as smooth handoff cast routing protocols for MANET is in [18]. Multicast is [14] and extension for IPv6 [16]. studied in [1], [2]. Broadcasting issues are studied in [7], [8], [9]. C. Related Work Cellular IP [21] and HAWAII [17] are two Internet protocols to support IP mobility. Cellular IP separates The Mobile IP is defined by IETF to support IP mobility macro mobility from micro mobility. Originally, mobile [15]. Transparent to TCP and UDP, it allows sessions to IP is to support macro mobility. To reduce frequent regremain connected when mobile hosts roam from subnet to istrations to HA as a MH is roaming around, Cellular IP adopts a hierarchical approach. A FA can provide sersubnet without the need of changing IP addresses. B. Mobile IP.

(23) 3 MANET1 vices to multiple base stations. As long as a MH is covMANET2 A ered by base stations belonging to the same FA, no reregistration is required. In this case, handoff delay may be significantly reduced. As such, micro mobility to support seamless handoff is achieved. HAWAII (HandoffN=2 G1 G2 N=3 Aware Wireless Access Internet Infrastructure) adopts a Internet domain-based approach to support mobility. Base staG3 G4 tions can be connected as a tree. It uses specialized path N=3 N=3 setup schemes which install host-based forwarding entries MANET3 MANET4 B in specific routers to support intra-domain routing. This D C results in the same advantage of supporting micro mobility and fast handoff as in Cellular IP. Compared to Cellular E IP, HAWAII breaks the tie between gateway and FA, and thus is more tolerant to failure of gateways and simplifies Fig. 2. The proposed network architecture, which extends each access the design of gateways. point to a MANET. References [10], [13] also address the construction of MANET by providing continuous Internet access based ther, to enjoy the flexibility of MANETs, direct communion Mobile IP. How to extend Mobile IP to allow MHs to cations between hosts under two FA’s coverages, through use CoA even if they are more than one hop away from MANETs’ links, are possible. FAs is addressed. The conflict between the management of routing tables in Mobile IP and MANET is resolved. III. N ETWORK A RCHITECTURE AND Implementations on both OS/2 and AIX are reported. In C OMMUNICATION S CENARIOS particular, two separate daemons are used by Mobile IP and MANET. To coordinate these two daemons, a route A. Network Model manager is used to control the system’s routing table. We consider the network consisting of multiple. Compared to [21], [17], our network architecture does not rely on hierarchical (wireline) routers. Instead, following the basic idea of ad hoc networks, mobile hosts are used as routers to extend the coverage of FAs. Thus, our framework also support micro mobility as well as macro mobility. While Cellular IP and HAWAII restrict mobile hosts be resident in one (wireless) hop from the base stations, our framework allows mobile hosts in multiple (wireless) hops from the base station. Also, our work is compatible with current design of MANET, thus easily extending MANET for IP mobility support. In addition, since the topologies of MANETs may change dynamically, the service ranges of FAs may also change accordingly. An advantage is a higher fault-tolerant capability — if one FA crashes, a mobile host may rely on MANET’s routing capability to connect to neighboring FAs. Compared to [10], [13], which considers only one single MANET, we consider the existence of multiple MANETs in a vicinity area. Specified by the local FA, the service range of a FA may be dynamically adjusted. Negotiation between mobile agents and mobile hosts on FAs’ service ranges is possible. This may greatly improve the flexibility of MANETs and reduce the service overhead of mobile agents. Also, MANETs may overlap with each other, and thus can support each other and offer a higher fault-tolerant capability in terms of Internet access. Fur-. MANETs, each of which has a point of attachment to the backbone Internet. The host connecting a MANET to the Internet is called the gateway. We use gateways to define the ranges of MANETs. Each gateway has two network interface cards (NICs), one wireless and one wireline. Gateway hosts have no mobility since they have fixed links. However, non-gateway hosts can roam around freely, and thus the definition of MANETs actually changes by time. Several MANETs are shown in Fig. 2. Gateways are responsible of intertworking MANETs with the Internet by forwarding/relaying packets. To support Mobile IP, each gateway also serves as the FA in its local MANET. So it should periodically broadcast AGENT ADVERTISEMENT messages to announce its service to members of its MANET. (In our discussion, we may interchangeably use gateway and FA according to the context.) Since members of MANETs are mobile, it is likely that a MANET is partitioned into multiple MANETs, or some MANETs may join and overlap with each other. In such cases, the boundaries between MANETs become vague, making the service ranges of FAs unclear. We propose to define the service ranges of gateways by associating with each gateway a parameter . Any wireless hops from the gateway mobile host within can join the MANET served by the gateway. This is. . .

(24) 4. . MANET1 achieved by specifying a TTL = in each gateway’s C MANET2 AGENT ADVERTISEMENT. For example, in Fig. 2, host A, though connected to MANET2, can not be a part B A of the network. G1 G2 In case that a host is within the service ranges of multiple gateways, it can choose the shortest-distance one as Internet Data packet its default gateway. By so doing, the boundaries of subEncapsulated header Data packet G3 nets are clearly defined even if MANETs are overlapping G4 Data packet with each another. For example, in Fig. 2, host C belongs CH to MANET3, while host B belongs to MANET4, and their HAs will tunnel IP datagrams accordingly from the proper D gateways. Also, note that each gateway can define its MANET4 MANET3 D own independently based on its willingness/capability to provide services. Fig. 3. Intra- and inter-MANET routing scenarios. When MHs move around, it is even possible that a MH is disconnected from its gateway, but still remains connected to other MANETs. For instance, in Fig. 2, ¯ Inter-MANET communication (direct): For any if the link between G3 and D breaks, hosts D’s and packet whose destination is not listed in the kernel E’s connections to the Internet will become broken berouting table, it will be forwarded to the gateway of cause they are beyond the service range of G4. To the local MANET. The gateway will then forward the dynamically adjust a gateway’s service range, we propacket to the Internet. The transmission from A to pose that a MH, on missing AGENT ADVERTISEMENT C in Fig. 3 is such an example. Packets travel on for a certain period, may broadcast or multicast an MANET1 based on DSDV, then on the Internet to AGENT SOLICITATION message with a TTL = ¼ . The G2 based on IP routing, and then on MANET2 to B value of ¼ can be gradually increased to avoid the broadby DSDV again. cast storm problem [7] caused by flooding. The solicita¯ Inter-MANET communication (with Mobile IP): A and the MANET is contion can be heard if ¼ MH may roam away from its home network. In this nected. On receiving the AGENT SOLICITATION, the case, Mobile IP will be involved to forward packets gateway may decide, based on its willingness, whether to between MANETs. In the transmission from CH to or not. In the above example, if host E’s increase its D in Fig. 3, packets will arrive at G4 by IP routing. AGENT SOLICITATION has an ¼ , G4 will receive These packets will be encapsulated and tunneled, by the request, and may increase its service range to cover D Mobile IP, to G3, which will then forward them to D and E. by DSDV. To support such scenario, MHs have to monitor any B. Some Communication Scenarios existing AGENT ADVERTISEMENT. Registration Based on the above network architecture, several differand deregistration procedures in Mobile IP should be ent communication scenarios may exist. In the following, followed. The routing of these packets will be supwe discuss the possible combinations and the correspondported by DSDV. HAs should maintain the current ing routings. In the discussion, we assume that routing locations of its MHs. FAs should maintain the visin MANETs is supported by DSDV (however, any proper iting MHs in their MANETs. HAs should execute routing protocol for MANETs is applicable). proxy ARP for roaming MHs. ¯ Intra-MANET communication: The communications ¯ Inter-MANET communication in overlaid MANETs are supported by DSDV. In the DSDV protocol, hosts (direct): When two MANETs overlay with each will exchange routing information periodically and other, a MH may be aware of a route to another MH compute the next hop to reach the destination with that belongs to a neighboring MANET. This is made the least metric (such as hop count). Proper route possible by the frequent exchange of routing inforentries will be written into the kernel routing table mation by DSDV. In this case, directly routing beby system calls. So whenever a route entry leading tween these MANETs is allowed. For example, in to the destination is found, the packet is directly forFig. 4, since A has a route entry leading to B, direct warded to the next hop. The transmission from A to inter-MANET transmission is possible. B in Fig. 3 fits into this category. To support such scenario, we propose to associate. . . . . . . .

(25) 5 MANET2. Internet. B. A. Gateway/Agent. text. MANET1 G1. dest IP=Gateway. G2. Encapsulated header. Broadcast packet. Broadcast packet. Internet M=5 G3. G4. HA of D. C. A MANET5 D. MANET3. MANET4. Fig. 4. Inter-MANET routing scenarios in overlaid MANETs.. . with DSDV a parameter , which reflects the service range of DSDV. I.e., a MH always collects/propagates routing information for MHs that wireless hops from itself. As a reare within sult, hosts in different, but connected, MANETs may communicate with each other directly. The routing, tunneling, and encapsulating overheads can be reduced by such optimization. Note that it is manda so that routing information leading tory that to the local gateway is always known by a MH. Inter-MANET communication in overlaid MANETs (with Mobile IP): Contrary to the above scenario, when two MHs are resident in connected MANETs hops, their communibut away by more than cations should be routed through the Internet. In the transmission from C to D in Fig. 4, assuming , host C will not be aware of any route (although existing) leading to D. In this case, its IP datagrams will be forwarded to the local gateway G3 (by DSDV), which will in turn forward the datagrams to D’s HA (by IP routing), which will encapsulate the datagrams to D’s current FA (by Mobile IP), which will forward the datagrams to D (by DSDV). and should As can be expected, the values of be properly tuned to reduce overheads and improve efficiency, which may be directed to an interesting research problem. Broadcast: Broadcasting is useful in many circumstances. In wireline communication, the scope of broadcast is well defined — a broadcast message is typically flooded to the physical range covered by a subnet. In wireless communication, due to the radio transmission property, the range that should be covered by a broadcast is usually not well defined. This is particularly true for ad hoc networks, where. . . ¯. . . . ¯. . Fig. 5. Routing scenario of broadcasting.. each MH has its own radio coverage. Note that if we directly adopt a TTL value to a broadcast packet, each mobile host’s broadcast range will be distinct (depending on its current location). We propose to define the coverage range of a broadcast as the service range provided by the local gateway where the broadcast is issued. As a result, the range of a subnet matches with the range of a MANET. The detailed routing is conducted as follows. When a MH wants to send a broadcast datagram, it first encapsulates the packet as a unicast by identifying the gateway as the destination host. When the unicast packet is tunneled to the gateway, the gateway will decapsulate the packet and find that it is a broadcast packet. Then the gateway broadcasts this packet on behalf of the original source with a TTL = . For example, Fig. 5 shows how A’s broadcast datagram flows. Note that to detect duplicate broadcasts, each MH should maintain a list of broadcast IDs that it has received recently. The “source IP address” and the ”IP identification” fields in the IP header can serve as a unique identity.. . IV. I NTEGRATION AND I MPLEMENTATION I SSUES Based on the proposed network architecture, we have developed a prototype which integrates Mobile IP and MANET together. Below, we report our integration and implementation experiences. The following modifications are required: ¯ TTL in IP Packets: Each IP datagram has a TTL field to control its lifetime on the Internet. In the original Mobile IP, each AGENT ADVERTISEMENT should have TTL = . We dynamically tune TTL to control our AGENT ADVERTISEMENT, AGEN SOLICITATION, and broadcast packets..

(26) 6 ¯. Routing inside MANET: Our implementation is based on the DSDV protocol [12]. Each host maintains a forwarding table containing a list of all available destinations together with the next hop leading to each destination. This forwarding table is used to update the kernel’s routing table. Control packets are used to exchanges distance vectors between neighboring hosts. Each route entry is tagged with a sequence number originated by the destination host. These control packets have a destination address of 224.0.0.1 (all-systems multicast address) with TTL = because they need not to be rebroadcast. In our protocol, several modifications are required. First, since we allow MANETs to overlap with each other, to avoid the amount of information being exchanged becoming too large, only route information hops is registered and propagated. that is within should be at least as large as Second, recall that used by the local gateway; the value of should be broadcast together with the gateway’s control packets. Third, each gateway should identify itself as a gateway by associating a gateway bit in its control packets. Each MH should set its default router to be the host that leads to the gateway host with the least metric. Forth, if a MH also has a CoA, it has to advertise through DSDV’s control packets its original IP address as well as its CoA. This is similar to having two IP addresses by a host. This can be easily achieved by providing two entries in the control packets. So the MH can be reached both by its permanent IP address directly (in the MANET sense) and by its CoA (in the Mobile IP sense). Agent Advertisement: In the original Mobile IP, AGENT ADVERTISMENT has TTL = . Due to the multi-hop nature of MANETs, the TTL should be set to , and the value is decreased by one each time it is rebroadcast. No rebroadcast is needed when TTL = . The destination field should be 255.255.255.255. Agent Solicitation: A MH can multicast AGENT SOLICITATION to find a nearby mobile agent. The destination field should be the all-routers multicast address 224.0.0.2. We recommend that its TTL field be doubled each time when the solicitation process fails. Intuitively, doubling the TTL can reach approximately four times the hosts that can be reached in the previous round. In addition, since the value of TTL will be decreased as the packet travels more hops, the original TTL value should be recorded in the packet’s payload so that when the gateway receives the packet, it can recover its distance to the requesting MH. By comparing this. . ¯. . . . ¯. ¯. . . . ¯. . ¯. . . . . ¯. value to , the gateway can decide whether it needs to enlarge its service range or not. ARP: In the original Mobile IP, ARP should be disabled when a MH visits a foreign network. The MAC-to-IP address mapping is registered when AGENT ADVERTISEMENT is received. Under our network architecture, to allow peer-to-peer communication inside a MANET, ARP still needs to be enabled in foreign networks. ARP requests and replies should be sent as usual. Since many nomadic hosts may exist in a MANET, the concept of subnet mask should not be used and packets of any destination should be relayed. Broadcast: We design a broadcast daemon to support the scenario in Fig. 5. Whenever a broadcast datagram with destination address = 255.255.255.255 and TTL = , source address = myself is intercepted, the daemon will encapsulate this packet as a unicast destined for the local gateway. On receiving the packet, the gateway will decapsulate and broadcast it with TTL = . However, one potential problem is that the broadcast datagram may loop back to the source host. To resolve this problem, the broadcast daemon should also record the recent broadcast datagrams that it has encapsulated recently. Destination address and TTL: Recall that the used used by MANETs should be at least as large as and , we can conby Mobile IP. By adjusting trol the amount of traffic flowing into and out of , which a MANET. We recommend that guarantees that intra-MANET communication can always be done directly without encapsulation (in the worst case, an intra-MANET packet may need to be sent to the gateway first and then forwarded to the destination). Also, inter-MANET communication between nearby MANETs are likely to be done without going through Mobile IP (and thus the encapsulation procedure). Configuration of IEEE 802.11b NICs: In our implementation, all wireless NICs are set to the peer-topeer (ad hoc) mode. All mobile hosts use the same ESSID and the same channel number so as to communicate with each other. To increase channel reuse (and thus communication bandwidth), it is possible that FAs can use different channels. In most current products, a NIC will automatically scan the available channels only when its current connection is broken (i.e., active scanning). So a host may not be able to discover all its neighbors if they are operating at different channels. Under our framework, the network should function correctly, except that some. .

(27) 7. Broadcast Daemon. MIPd. Other Applications. R EFERENCES DSDVd Forwarding table. TCP/UDP. ARP. Encapsulation and Decapsulation. IP. System routing table. Data Link Layer. Physical Layer. Fig. 6. System architecture of our implementation.. routes may not exist even if some hosts are physically neighbors. Our system is developed based on Linux Redhat 2.2.16. Two daemons, namely DSDVd and MIPd, are implemented. Conceptually, both daemons are network-layer programs. However, they are actually implemented on the application layer and interact with system kernel through socket interfaces. The concept is shown in Fig. 6. DSDVd periodically multicast UDP packets to help maintain hosts’ forwarding tables. Proper entries from the forwarding table are written into the kernel’s routing table by system calls. In MIPd, we use RAW sockets for advertisement, encapsulation, and decapsulation, and normal sockets for registration. Proxy ARP is done by UNIX system calls. Also, the IP forwarding option at each mobile host must be turned on. V. C ONCLUSIONS In this paper, we have investigated the related issues to integrate MANETs with Mobile IP. Hence, traditional access points can directly enjoy the flexibility of MANETs and widen their coverage ranges. In view of the worldwide explosive deployments of IEEE 802.11-based access points, such extension would help make our dream of ubiquitous broadband wireless access come true. Details of our prototyping and implementation experiences are reported. Some performance test results are available in [19]. The discussion in this paper is based on IP version 4; it will be interesting to investigate the related issues on IP version 6. ACKNOWLEDGEMENTS This work is co-sponsored by the MOE Program for Promoting Academic Excellence of Universities under grant numbers A-91-H-FA07-1-4 and 89-E-FA04-1-4.. [1] C.-C. Chiang, M. Gerla, and L. Zhang. Forwarding group multicast protocol (fgmp) for multihop, mobile wireless networks. ACM-Baltzer Journal of Cluster Computing, 1(2), 1998. [2] J. J. Garcia-Luna-Aceves and E. L. Madruga. A multicast routing protocol for ad-hoc networks. In INFOCOM, 1999. [3] Z. Haas and M. Pearlman. ZRP: A Hybrid Framework for Routing in Ad Hoc Networks (a book chapter in Ad Hoc Networking, Ed. C. E. Perkins, Chapter 7). Addison-Wesley, 2000. [4] M. Jiang, J. Li, and Y. Tay. Cluster based routing protocol (CBRP) functional specification (internet draft), 1998. [5] D. B. Johnson, D. Maltz, and J. Broch. DSR: The Dynamic Source Routing Protocol for Multihop Wireless Ad Hoc Networks (a book chapter in Ad Hoc Networking, Ed. C. E. Perkins, Chapter 5). Addison-Wesley, 2000. [6] J. Macker and M. Corson. Mobile ad hoc networking and the IETF. ACM Mobile Computing and Communications Review, 2(1):9–14, Oct. 1998. [7] S.-Y. Ni, Y.-C. Tseng, Y.-S. Chen, and J.-P. Sheu. The Broadcast Storm Problem in a Mobile Ad Hoc Network. In MobiCom, pages 151–162, 1999. [8] E. Pagani and G. P. Rossi. Reliable broadcast in mobile multihop packet networks. In MobiCom, pages 34–42, 1997. [9] E. Pagani and G. P. Rossi. Providing reliable and fault tolerant broadcast delivery in mobile ad-hoc networks. Mobile Networks and Applications, 4:175–192, 1999. [10] C. Perkins. Mobile-IP, ad-hoc networking, and nomadicity. In COMPSAC, 1996. [11] C. Perkins. Ad-hoc on-demand distance vector routing. In MILCOM, 1997. [12] C. Perkins and P. Bhagwat. Highly Dynamic DestinationSequenced Distance-Vector (DSDV) Routing for Mobile Computers. In ACM SIGCOMM Symposium on Communications, Architectures and Protocols, pages 234–244, Sep. 1994. [13] C. Perkins and H. Lei. Ad hoc networking with mobile IP. In EPMCC, 1997. [14] C. Perkins and K. Wang. Optimized smooth handoffs in mobile IP. In IEEE Symp. on Computers and Communications, 1999. [15] C. E. Perkins. Mobile IP Design Principles and Practices. Addison-Wesley, 1997. [16] C. E. Perkins and D. B. Johnson. Mobility support in IPv6. In Mobile Computing and Networking, pages 27–37, 1996. [17] R. Ramjee, T. F. L. Porta, S. Thuel, K. Varadhan, and S. Y. Wang. HAWAII: A domain-based approach for supporting mobility in wide-area wireless networks. In ICNP, pages 283–292, 1999. [18] E. Royer and C. Toh. A review of current routing protocols for ad-hoc mobile wireless networks. IEEE Personal Communications, Apr. 1999. [19] C.-C. Sheng. Mobile IP and ad hoc networks: An integration and implementation experience. Master Thesis, National Chiao Tung University, Dept. of Computer Science and Information Engineering, 2002. [20] Y.-C. Tseng, S.-L. Wu, W.-H. Liao, and C.-M. Chao. Location awareness in ad hoc wireless mobile networks. IEEE Computer, 34(6):46–52, June 2001. [21] A. G. Valko. Cellular IP - a new approach to internet host mobility. In ACM Computer Communication Review, 1999..

(28) 論文名稱：Collision Analysis for a Multi-Bluetooth Picocells Environment 作者：T.-Y. Lin and Y.-C. Tseng 發表情況: IEEE Communications Letters, Vol. 7, No. 10, Oct. 2003, pp. 475-477. (SCI, EI).

(29)

(30) .

(31)

(32)

(33) !"" # . . .

(34)

(35) !

(36) " # "# "# $ # %

(37) % & #

(38) # % '()

(39) #

(40) #

(41)

(42)

(43) ! # % #

(44) #

(45) # " "

(46) !

(47) ! % * # ! % ! # . .

(48) ! " #$%& '() * +, + - * ' * . * + * / . 0 ( + / + . + * 1. 23 + / * * Æ ( .

(49) . * * 1/443 +, 10%3 " / + * * * 5 , + 67 " +, ' , 8 '4 * / * . +, ' + . 67 +, ' * * * / & *9 . 2 0 +, + : * ( . ' + +, ! * * + * + Æ * + . 2 * + * ' ; ; 2 . 0 * <= / * + 1.3 > .1 ; ;23 . ?8 , + + + + : * 13 + > . 2 + < = @ ! 8 % * + + 0 + . + + + + + + 8 . 1 3 + 6AB67 + 4 13 * : 1+ 0 .39 ¯ 9 + . 2 * ¯ 9 + + *.

(50) B1. B2. 1-slot. B3 3-slot. B4. B5. B6. packet. B7. B8. 5-slot. B9. ;; ;; ;; ;;;;;;;;;. B 10. packet. X.

(51) . Y. B1. 9 + + :+ + 2 * ¯ 9 + ' + C + + C + + C + 0 + * + 1 3 > .. .D $* : 1 3 + + 0 1.3 > . 13 > " 163 > 1.D3 > . ' * + B B 1+ - +, 3 + + + 8 ! + + + 13 + > . 2 * + / : + $* + :

(52) 1 3 + + + + + 1 ;.3 + + + + BB B 0 D 1 * : 3

(53) 1 3 > . ' *

(54) 1 3 + + 8 + : + * : : 13 - + + 8 (: 13

(55) 13 +9. ¯. 13 >. 1 3 1 3

(56) 1 1 33 . . . 1. 3 ; 1 3 > 1. 3 ; . . L(5-g(3)) = L(3). X. Y. B3. (b).

(57)

(58)

(59)

(60)

(61) . 0 0 "13 . / : + / + : 1.3 > 1. 3 ; ' * + + + 1+ 3 C + / * + + + *

(62) 1"3 + 2 0 "13 / : 2 + ( + + 13 0

(63) 13 ( + 2 13

(64) 13 / + 8 *

(65) 1 3 + + BBB ( + * * + 1 D39

(66) 1 3 > ; ; ;

(67) 1 1.D33 ; ; ; ;

(68) 1 1.33 ;

(69) 1 1"33 ; ; ; ; 1.3 ; ; ;

(70) 1 1233 1"3 . . . ' : +

(71) 1 3 > . + D * 0 . *. + ( + + 13 ( + 9 > 1.3 ; 13 ; 2 123 13 1B3 + . 2 * 1+ +. + > . " 2 + > .D. ' , > .. .D + : 8 / 1 3 0 1 3 * 13 + + : 13 + 8 ' + 1 3 + 8 + : + 1 3 1 + BB B 3 / E +

(72) 1 1 33 . (a). f(3). 5-slot packet. . . L(3-g(1)) = L(2). f(1). 3-slot packet. empty slot. . ".