適於DTN環境下的適地性內容服務之實踐議題研究

行政院國家科學委員會專題研究計畫 期末報告

適於 DTN 環境下的適地性內容服務之實踐議題研究

計 畫 類 別 ： 個別型

計 畫 編 號 ： NSC 101-2221-E-004-005-

執 行 期 間 ： 101 年 08 月 01 日至 102 年 09 月 30 日

執 行 單 位 ： 國立政治大學資訊科學系

計 畫 主 持 人 ： 蔡子傑

計畫參與人員： 碩士班研究生-兼任助理人員：林宇軒

碩士班研究生-兼任助理人員：林煜泓

碩士班研究生-兼任助理人員：韓建淳

碩士班研究生-兼任助理人員：詹筱璇

碩士班研究生-兼任助理人員：王凱柔

博士班研究生-兼任助理人員：詹賀翔

報 告 附 件 ： 出席國際會議研究心得報告及發表論文

公 開 資 訊 ： 本計畫涉及專利或其他智慧財產權，1 年後可公開查詢

中 華 民 國 102 年 12 月 28 日

中 文 摘 要 ： 近年來，無線網路技術不斷推陳出新，政府也不斷建設無線

網路的基礎建設，在都市每個角落皆能收到無線網路的訊

息，使用者擁有行動裝置即能連上網路，將不被地區及時間

所限制。

然網路連線常遇壅塞，或有些使用者並無 3G 上網能力，而

通常又需要適地性內容服務(Location and Content Based

Service)的 Data query/upload 的需求。因此，在此情況

下，耐延遲網路(Dealy/Disruption Tolerant Networks,

DTN)架構的可能應用價值更能呈現。

但上述 DTN 關鍵成功因素為 Data Forwarding 設計、轉發

誘因、以及是否可提升優先轉發策略等。因此將針對這些議

題加以研究，並部份功能實踐於 WiFi Direct 行動裝置上評

估效能，預期這將是一個完整實用性的技術與應用之關鍵研

究。

本研究的議題如下：針對在耐延遲網路下提出行動定位資料

搜尋方法、路徑選擇獎勵機制及訊息優先權演算法。依據耐

延遲網路訊息傳遞的特性儲存後轉發，各別提出方法來解決

使用者無法連上網路的困境。最終將所提的方法能在

Android 上部份系統實作，驗證其設計結果的可行性。

中文關鍵詞： 適地性內容服務、耐延遲網路、Android、P2P、獎勵機制、

訊息優先權

英 文 摘 要 ： Recently, the technologies of wireless networking

advance rapidly. In additions, users can almost

everywhere in the metropolitan area receive/access

wireless messages due to government＇s continuous

infrastructure deployment.

However, sometimes network would suffer congestion

due to traffic concentration or hot spot. Or some

users would not have subscriptions to 3G data

services. These users still have the needs for

Location and Content Based Services which often

require to query data and to upload them. In this

situation, DTN (Delay/Disruption Tolerant Networks)

will be more suitable and increase the application

value.

To be a success, the key challenging issues will be

data forwarding design, forward incentive, and

differentiated message priority. We will study these

issues and practice part of the developed features

into an integrated system using WiFi Direct enabled

mobile devices to do field test. It is expected as a

crucial study for the practice of location and

content based service for DTNs.

The research issues of this project will be:

location-based content search approach, message

forwarding strategies with incentive mechanism, and

message priority design. Finally, proof of concept

will be made through by Android mobile devices.

英文關鍵詞： Location and Content Based Services, Delay Tolerant

Networks, Android, P2P, Incentive, message priority

行政院國家科學委員會補助專題研究計畫成果報告

適於 DTN 環境下的適地性內容服務之實踐議題研究

計畫類別：■個別型計畫

計畫編號：NSC 101－2221－E－004－005

執行期間： 101 年 8 月 1 日至 102 年 9 月 30 日

計畫主持人：蔡子傑

共同主持人：

計畫參與人員： 李欣諦、林昶瑞、李英明、林煜泓、林宇軒

成果報告類型(依經費核定清單規定繳交)：■期末報告

本成果報告包括以下應繳交之附件：

□赴國外出差或研習心得報告一份

□赴大陸地區出差或研習心得報告一份

■出席國際學術會議心得報告及發表之論文各一份

□國際合作研究計畫國外研究報告書一份

處理方式：除產學合作研究計畫、提升產業技術及人才培育研究計畫、

列管計畫及下列情形者外，得立即公開查詢

□涉及專利或其他智慧財產權，■一年□二年後可公開查詢

執行單位：國立政治大學 資訊科學系

中 華 民 國 102 年 12 月 21 日

行政院國家科學委員會專題研究計畫成果報告

適於 DTN 環境下的適地性內容服務之實踐議題研究

計畫編號：NSC 101-2221-E-004-005

執行期限：101 年 8 月 1 日至 102 年 9 月 30 日

主持人：蔡子傑 國立政治大學 資訊科學系

計畫參與人員：李欣諦、林昶瑞、李英明、林煜泓、林宇軒

[2] Tzu-Chieh Tsai, Chan-Juei Lin, “Mobile Trusted Bank and Incentive Strategy Design in Delay Tolerant Networks”, in preparation for submission.

[3] Tzu-Chieh Tsai, Chieh-Cheng Chen, “Popularity Spray and Utility-based Forwarding Scheme with Message Priority Scheduling in Delay Tolerant Networks” , in preparation for submission.

本計畫原為申請二年期的計畫，但只核准了一年。 但是原本計畫書二年要研究的議題幾乎都完成了， 除了實作的手機行動系統之外(原預計就是第二年)。 一、A b s t ra c t 近年來，無線網路技術不斷推陳出新，政府 也不斷建設無線網路的基礎建設，在都市每個角落 皆能收到無線網路的訊息，使用者擁有行動裝置即 能連上網路，將不被地區及時間所限制。 然網路連線常遇壅塞，或有些使用者並無 3G 上 網 能 力 ， 而 通 常 又 需 要 適 地 性 內 容 服 務 (Location and Content Based Service) 的 Data query/upload 的需求。因此，在此情況下，耐延 遲 網 路 (Dealy/Disruption Tolerant Networks, DTN)架構的可能應用價值更能呈現。 但 上 述 DTN 關 鍵 成 功 因 素 為 Data Forwarding 設計、轉發誘因、以及是否可提升優 先轉發策略等。因此將針對這些議題加以研究，並 部份功能實踐於 WiFi Direct 行動裝置上評估效能， 預期這將是一個完整實用性的技術與應用之關鍵研 究。 本研究的議題如下：針對在耐延遲網路下提 出行動定位資料搜尋方法、路徑選擇獎勵機制及訊 息優先權演算法。依據耐延遲網路訊息傳遞的特性 儲存後轉發，各別提出方法來解決使用者無法連上 網路的困境。最終將所提的方法能在 Android 上部 份系統實作，驗證其設計結果的可行性。 關鍵詞：適地性內容服務、耐延遲網路、Android、 P2P、獎勵機制、訊息優先權 二、緣由與目的、結果與討論 智慧型手機及平板電腦方便攜帶的優點及現今的行 動網路技術研究趨近成熟，如 3G(UMTS、CDMA2000)、 3.5G(HSDPA)甚至是號稱 4G 的 WiMax 都已經有商業 化的產品出現，其中又以 3G、3.5G 上網情況較為 普及。使用者能隨時隨地透過手機及平板電腦連上 能網路查詢使用者需要的資訊，這樣的使用行為也 延 伸 出 新 興 的 研 究 議 題 ， 名 為 適 地 性 服 務 (Location based service，LBS)，適地性服務應 用層面相當廣泛，例如尋找鄰近資訊、行動導航、 社群互動、區域廣告、折價券廣播等服用應用。當 使用者移動到陌生的環境中，對於當地有那些特色 的資訊是不太清楚的狀況時，使用者能藉著手機或 是平板電腦等行動裝置來獲取由當地商家提供的資 訊或是與其他使用者通訊來交換現有的資訊。然而 這些服務大多必須能連上網路才能獲得當地的資訊， 若使用者無法連上網路就無法獲取當地的資訊。 在行動網路技術仍有一些挑戰必須解決，當某一個 區域內同時連上網路的用戶過多而頻寬卻有限制的 情況下，很容易造成網路的延遲(Delay)，甚至造 成無法連線，尤其是在人口密集處時，這樣的狀況 會更加嚴重，例如都會區、遊樂區、觀光景點區等 等，都很容易發生此現象，因此，因而延伸新型態

的 網 路 架 構 稱 為 耐 延 遲 網 路 (Delay/Disruption Tolerant Networks, DTN)。 此類型的網路架構特性，節點之間通訊連線並非同 時存在，而是間歇建立。會有此網路特性是因為節 點具有移動性、使用者手持裝置的電力問題或者是 使用者當下環境造成通訊節點連線不穩定無法與通 訊節點維持通訊路徑而斷線。而在 DTN 上採取儲存 後攜帶轉發(store, carry and forward)的機制來 幫忙轉送訊息，幫忙轉送的節點會先將訊息儲存在 本身的 buffer 中，遇到合適節點才會決定將訊息 傳送出去，所以當使用者無法連上網路的時候，使 用者可與鄰近的使用者通訊，交換彼此本身擁有資 訊，這樣使用者即使沒辦法連上網路獲取當地資訊， 還是能藉由與鄰近使用者通訊來獲得當地得資訊。 但由於無法確定會在什麼時機與其他節點相遇，亦 無法確定訊息什麼時候會送到目的端，這是造成延 遲時間過長的主要因素。

而 WI-FI 聯盟已在 2010 年正式提出 WI-FI Direct 的標準[19]，是屬於點對點的協定。具有 WI-FI 的 行動裝置，將可以不必透過無線網路基地台即可與 另一個具有 WIFI 的行動裝置連線，雙方能夠互相 交換資料。一開始行動裝置會先掃描通訊範圍內的 行動裝置，若行動裝置是處於 listen 狀態則可以 被其他裝置發現，此時 channel 相同則 search 狀 態的使用者可以向 listen 狀態的使用者發送要求 連線的訊息，若對方給予回應則雙方就能夠開始互 相傳遞資料。目前行動裝置的作業系統為 Android 4.0 已經有支援 WiFi Direct 的功能，市面上的產 品 有 :Samsung GALAXY Nexus S 、 GALAXY S II 、 GALAXY Tab 10.1、GALAXY Note、HTC Sensation 、 HTC Sensation XL 及 HTC EVO 3D 等。 由於智慧型手機及平板電腦的快速發展，行動裝置 都含有 GPS 可幫助使用者來定位，藉此能尋找 當地熱門商家或其它當地相關資訊。而有時候現有 的資料並非使用者需要的資訊，使用者能依照各自 的需求，發送查詢的訊息傳送給附近的使用者，以 等待其他使用者能給予發送查詢的使用者想得到的 解答。使用者可透過行動敘事平台[我們之前所發 表的論文以及 APP 系統] 根據其所在位置，利用該 系統提供之編輯方式即時分享所見所聞、心得感觸， 同時也能瀏覽其它遊客在此地所分享之心得故事， 但若在取得或分享資訊的同時沒有網路連線，也就 是無法連上網時，遊客則必需利用其它方式，將資 訊傳遞出去，例如可把欲傳遞之資訊傳遞給相遇的 節點，若該節點無所需之資訊時，再幫忙轉傳給其 它節點，直至遇到可給予相對應行為之節點為止， 如此一來，則可在沒網路連線時也能取得自己所需 之資訊。 在上述有提到 DTN 網路特性採取儲存後攜帶轉發， 找到合適的節點來幫忙做轉送，能傳送到目的端。 研究 DTN 的學者，多半是假設節點都是願意幫忙傳 遞，但實際的情況很可能並非每一個節點都會願意 幫忙，節點考慮到自身的情況而選擇不幫忙傳遞， 這些不幫忙傳遞的節點稱之為自私節點，自私節點 的存在，將會降低整體網路的效能，導致訊息無法 傳送到目的端。而針對自私節點，有學者提出獎勵 機制來吸引自私節點來幫忙傳遞，獎勵機制必須是 吸引人並且是公平的，才能提高自私節點的幫忙意 願，藉此能提高訊息的傳遞率。 DTN 其中的特性之一是節點之間的通訊不必依賴基 地台等基礎建設，而當發生大自然及人為災難，例 如:地震、海嘯及火災時，有可能對現有的通訊基 地台或骨幹網路造成一定程度的損壞，導致無法提 供正常的通訊服務。在此惡劣環境下，使用者能透 過 DTN 將訊息傳遞出來，但在同個區域很多使用者 想傳送的訊息並非各個訊息都擁有非常急迫的性質， 若以一般的傳送方法可能會影響到救援的黃金時機。 若能將訊息的重要程度依照一個基準來分類，再依 據訊息的重要程度來選擇是否先行傳送，由於節點 之間的相遇時間可能很短，無法將所有訊息一次傳 完，但若重要訊息能在節點一相遇時將訊息傳送給 其他節點就能較快速地送達其目的節點。綜上所述， 我們相信如果能有效的組織與排序這些緊急程度不 同的訊息，除了能夠適時的依重要(優先)程度傳遞 訊息，再搭配著適用於緊急訊息通訊路由廣播機制， 將能夠有效減少重要的訊息傳送時可能造成的頻寬 壅塞(Congestion)、封包碰撞(Collision)以及減 少訊息傳遞的延遲時間(Delay time)，並提高訊息 的傳遞成功率(Delivery ratio)。 基於上述，本研究的目的為:在 DTN 行動網路環境 下，利用使用者位置定位來查詢當地的熱門資訊， 並且發送查詢的節點會給予幫忙轉送及回覆的節點 獎勵，來解決使用者在當地環境能根據查詢得到回 覆並且提高節點幫忙的意願，希望這樣的設計能提 高訊息傳遞率，並且貼近實際情況的機制。 本研究的作法是使用者會根據手持行動裝置定位了 解週遭環境，當使用者想要查詢當地的資訊時，即 便使用者位於無網路連線狀態，亦能在最快的時間 得到查詢結果的回覆。本研究會依據資料複製的策 略、查詢複製的策略、取得資料回傳的策略及節點 將資料同步於伺服器的方法來做為在 DTN 環境下機 制的方法，藉此能達到訊息快速的傳播及得到回應。 而實際的情況下，當查詢節點需要其它節點的幫忙 傳送或回覆查詢時，節點並非都願意幫忙傳送訊息， 在實際的環境下陌生的節點收到傳輸範圍的節點所 發送的訊息，因為彼此不認識且幫忙節點來轉傳訊

息或回覆訊息從中無法獲得好處，收到訊息的節點 很可能就會選擇丟棄訊息，所以在這我們提出獎勵 機制的方法。當來源節點發出請求，希望其它節點 能給予幫助幫忙傳送訊息或回覆訊息時，來源節點 將會提供獎勵回饋給幫忙傳送訊息的節點及回覆的 節點，來鼓勵節點之間的互相幫忙傳送訊息。經由 我們所設計獎勵機制的方法，提升節點幫忙的意願， 藉此能提高訊息送達率，減少訊息在傳送中丟棄的 可能性。 在同個時間下，有可能節點會接收到來自四面八方 的訊息情況，而這些訊息可能會有時效性，需要優 先被快速傳送到目的節點，本研究提出能根據訊息 的重要程度來做為訊息傳送的優先順序，幫忙傳送 的節點能明確了解此訊息的重要程度來立即傳送， 進而能有最小的延遲時間、較高的訊息送達率，而 這項服務將會根據環境的狀態來做調整。 因此綜觀上述，要使得 DTN 在 Location/Content-based Services，能實踐成功的關鍵議題，包括資 訊擷取、轉發策略、幫忙轉發誘因、優先轉發機制 等。本研究的議題將包括以下所列項目: 1. 依據使用者位置查詢當地資訊及回覆研究 2. 節點幫忙傳送及回覆得到獎勵機制演算法 3. 訊息依重要程度區分的品質服務演算法 我們一年之內就完成了上述三大研究議題，且有一 篇已經發表，並同時被推薦收錄於另一期刊中。另 外兩篇有一篇在整理，一篇已被接受在修改中。

接下來就分別就三篇論文/報告作成果之摘要

重點：

1. A Location-based Content Search Approach in Hybrid Delay Tolerant Networks

1.1 Abstract

In Delay Tolerant Networks (DTNs), the offline users can, through the encountering nodes, use the specific peer-to-peer message routing approach to deliver messages to the destination. Thus, it solves the problem that users have the demands to deliver messages while they are temporarily not able to connect to Internet. Therefore, by the characteristics of DTNs, people who are not online can still query some location based information, with the help of users using the same service in the nearby area. In this paper, we proposed a Location-based content search approach. Based on the concept of three-tier area and hybrid node types, we presented four strategies to solve the

query problem, namely, Data Replication, Query Replication, Data Reply, and Data synchronization strategies. Especially we proposed a Message Queue Selection algorithm for message transferring. The priority concept is set associated with every message such that the most “important” one could be sent first. In this way, it can increase the query success ratio and reduce the query delay time. Finally, we evaluated our approach, and compared with other routing schemes. The simulation results showed that our proposed approach had better query efficiency and shorter delay.

1.2 Main Results

In our research scenario, all nodes in the networks are categorized into two types of statuses: (1) Online, and (2) Offline. Online-node can connect to the Internet to access the remote server, and Offline-node cannot connect to that.

To sum up, our goal is to let Offline-nodes get the information they need by help of other nodes soon and efficiently. In order to accomplish this, we proposed four strategies: (A) Data replication, (B) Query replication, (C) Data reply strategic, (D) Data synchronization and update strategies.

(A) Data replication

When two nodes A and B encountered, the selection rule is shown in Equation (2). If the distance

d between Node A and Area Central Point (CP) is less than or equal to radius , it indicates Node A is in the Inside Area. When Node A encountered Node B, it will add all the Data messages into the Data dataset D_set to prepare to send to the Node B. The main purpose is to let the Inside Area be filled with related Data messages. If any node wondered to query in this area, it can get the Data messages quickly and has high query success rate. However, the encounter time of the two nodes and the transmission are limited, two nodes have to change itself metadata before transport. Therefore, Node A skips the identical Data messages, and adds the others into D_set. If the distance d between Node A and CP is greater than radius and less than , it indicates Node A is in the Border Area. According to Node B’s past path, if , it indicates Node B will likely enter to the Inside Area, then, Node A add the Data messages into the D_set. If the distance d between Node A and CP is greater than , it indicates Node A is in the Outside Area, then, Node A should not do anything.

We use to predict whether Node B’s will enter to the Inside Area or not.

( ) { ( ) ( )

In order to avoid data overflow, we set all the Data messages a storing time TTS (Time-to-Store). TTS means the storing time of the message from being created or replicated. When TTS expired, the messages will be dropped.

When the encountered node, Node B, is an Online-node, Node A doesn’t need to send Data messages to Node B. Because the Online-node can connect to the Internet to get the messages by itself, it doesn’t need to waste resource to send messages.

(B) Query replication

The goal of this strategy is to let user’s query spread appropriately if not yet having data reply. It is a three-tier Query strategy, namely, from Node A’s local database to Node B’s local database then the remote server. When Node A receives query messages, it has to check the database whether the query messages have the data already. If it has, it doesn’t need to do Query replication; if not, it reacts by the rule: (1) Node A is an Online-node. It does nothing. (2) Node A is an Offline-node. When Node A is in the Inside Area, and then adds all the related Query messages to the Query dataset Q_set; When Node A is in the Border Area, then Node A has to predict Node B’s direction ( ∆N_b.d , see details in Section 3.1, and also shown in Equation (3)). If ∆N_b.d less than 0, it indicates Node B is likely to enter to the Inside Area, then, add the related Query messages to the Q_set. If not, then do nothing. The main purpose of the Data replication strategy will centralize the related Data messages in the Inside Area. Therefore, if we send the queries to the Inside Area, and there are more opportunities to facilitate query success. The reason is twofold. There are many related messages in the Inside Area. And, if any node leaves from Inside Area, we expect it could carry much more related messages; when Node A in the Outside Area, it shouldn’t do anything.

In addition, if the query represents it had been queried (query.isQueried=true), then we don’t have to add this query into the Q_set.

(𝑞 𝑦 𝑄 ) { ( ) 𝑏 (𝑞 𝑦) 𝑞 𝑦 𝑄

(C) Data reply strategic

Node A received Node B’s metadata, it will check all query messages of Node B’s metadata. If Node A has the match data, it will generate a Reply data R_data of the Query message and change the Query message attribute isQueried to TRUE, then we can call Node A as a Replier.

It is difficult to send the Reply data to the Querier. Because all nodes in the network are keeping moving, we can’t use the original packet delivery path to send back to the Querier. Therefore, we have to use the other rule to send Reply data.

Replier (Node A) can know the Querier’s location from Query message, and send the Reply data to that location. But, as time goes by, Querier may not still stay there. So, when Replier (Node A) encounters other node (Node B), it will check the Encounter Table of Node B whether it had encountered Querier or not. If it had, we will check the encounter time of whom is closer to the recent time. If Node B’s encounter time is closer, we will update the Querier position of Replier’s (Node A) metadata.

𝑝 (𝑚 𝑞 𝑙𝑜𝑐)

{ (𝑄 ∈ 𝐸𝑁𝑠 𝑙𝑒) ( 𝑞 𝑖𝑚𝑒 𝑞 𝑖𝑚𝑒)

(5)

Then we have to distinguish which Node will move toward to the Querie. If Replier (Node A) won’t move toward to the Querier but Node B will do, it would add the R_data to the Reply dataset R_set. Otherwise, Replier (Node A) will move toward to the Querier but Node B won’t, it just keeps the R_data and does nothing. (𝑅 𝑅 ) { (𝑞 ) (𝑞 ) (𝑞 ) (𝑞 ) 𝑞 & ( ) 𝑞 & ( ) ′ _𝑞 (6)

If Replier (Node A) and Node B are moving the same direction and forward to the Querier, we have to calculate whose moving path will be close to the Querier. We can use the position Pnow, the past position Pt-1 to determine this node’s moving direction, then we add the Querier position Pq and use the basic Triangle Area formula to calculate the area value A. As we know the distance between Pnow and Pt-1, we can get the distance D to Querier. If Da is greater than Db, it indicates Node B will move closer to the Querier, then we add the R_data to the R_set and wait for transmission. Otherwise, If Da is less than Db, it indicates Replier (Node A) will move closer to it, then, we just keep the R_data.

(D) Data synchronization and update strategies Either Online-node or Offline-node, it can create data messages anyway. Besides keeping the data in local database, it has to upload this data messages to the remote server to share this information. When nodes received messages, the first step is to distinguish the type of the node ifself. If it is Online-node, it should directly upload all the data messages which do not sync to the remote server. Then it changes the Sync attribute of the data message to TRUE. Otherwise, if it is Offline-node, it shouldn’t do anything at this time point.

In Figure 1, we can see our approach and Locus have a similar performance in query-reply success ratio, because both are using the region concept to centralize the messages. But, as the number of nodes increase, Locus will perform well. It is because our region concept is based on local area. If there are many nodes in this area, it will have many messages (data, query, reply) in every node. Although nodes have rich data source, the transmission rate is fixed, and nodes are intermittently connected, they can’t send all the messages. So, some messages might be ignored, and the performance is not well. And we want to see the important part of our approach, we modify it in two points. (1) We use epidemic routing to replace the data replication strategy, and (2) we use epidemic routing to replace the query replication. The result shows the (1) type got worse performance. Because the data messages couldn’t be centralized to the inside area, the messages would spread to the whole network. It is difficult to query unique data message in the network. On the other way, in order to compare fairly, we present a scenario with only Offline-node in our approach. Although the success ratio is worse than Locus, but the overhead and latency are better than Locus.

Figure 1: Query-Reply Success Ratio (Node Density) Figure 2 is overhead, and we can see the result of Locus. Although the query success ratio of Locus is

higher than LCS, but its overhead is larger than LCS. It is because Locus centralizes the same data in the same area. If Querier is close to this area, it can use a few query replicate to look up data quickly. But, if Querier is far away from this area, it has to replicate more message in delivery. So the overhead will be higher. And LCS is a wide range of the local area, it spreads all the messages in the area, thus, LCS has higher opportunity to spend lower cost to get the data.

Figure 2: Overhead (Node Density)

Figure 3 is Query Latency. We can see the result of our approach LCS is much lower than Locus and Epidemic. Because all the data will be spread in the area, if node in the Inside Area or Border Area, it will quickly get the data. Although Locus also uses the area concept, if nodes want to query the data which are not near the area, it will take much time to query data.

Figure 3: Latency (Node Density)

2. Mobile Trusted Bank and Incentive Strategy Design in Delay Tolerant Networks

2.1 Abstract

DTN (Delay Tolerant Networks) is a network structure without need to use any infrastructure. In DTNs, wireless connections between nodes do not always exist, i.e., nodes are connected intermittently. Due to the mobility, power issues, or surrounding environment of nodes, connections between nodes may be disruptive occasionally or randomly. In a DTN, nodes usually

transfer the message to the encountering nodes. By this way, the messages are stored, carried, and forwarded to the next nodes, possibly the destination. However, in reality, nodes may not be willing to help each other for the message forwarding. More specifically, there are “Selfish nodes” which refuse to forward messages due to issues such as energy and network bandwidth. Thus it will prevent messages from being forwarded to its destination.

In order to solve the problem of message forwarding failure caused by selfish nodes, we proposed an Incentive Strategy called “MTBIS (Mobile Trusted Bank and Incentive Strategies)”. We construct a rewarding scheme called “Incentive Credit” for nodes who forward the messages for the source node. In addition, to increase the intention of the selfish nodes for forwarding messages, we add two more mechanism called SI(Social Incentive) and DGT(Dynamic Grudger Threshold). The DGT uses credits to solve selfish node problems from an economics point of view to enhance MTBIS to work with existence of selfish nodes. In this paper, we emphasis on the simulation of the behaviors of selfish nodes, using four different types of nodes: the “Sucker”, the “Cheater”, the “Grudger”, and the “Ecci”. These selfish nodes will decrease the success rate of message forwarding, and even make the DTN unusable. We simulated the behaviors of these selfish nodes, using the rewarding scheme we proposed. From the results of our experiment, we see that the Grudger can effectively deal with performance issues caused by selfish nodes, and the system can gain 34% in performance compared to the traditional algorithms. 2.2 Main Results

In traditional incentive techniques, there is a centralized (or distributed) Trusted Bank in system. Trusted Bank is responsible for managing all nodes Incentive Credit, when a node requested to other node help forwarding, after more than one node to help successfully forwarded the message to the destination node, all of the intermediate nodes will get the reward called “Incentive Credit”. We construct a rewarding scheme called “Incentive Credit” for nodes who forwards the message from the source node. Trusted Bank will deduct from the source node’s incentive credit, and assigned incentive credit to nodes who participate in forwarding. In this paper, we focused on Mobile Trusted Bank.

Assumed in a city, all nodes can be divided into two major categories the bus and pedestrian. Bus nodes have a fixed moving path, moving faster and larger transmission range, and larger buffer size characteristics of bus. General node which is pedestrian with a handheld device (such as a PDA or smart phone),

moving slower, smaller transmission range, and smaller buffer size characteristics of pedestrian, most of the nodes are moved within a certain range, so we take advantage of these features to the design system. Based on above characteristics, we choose bus node as MTB(Mobile Trusted Bank). MTB just like a bank responsible for managing all nodes incentive credit and credit clearance and help forwarding messages. The nodes will get the qualifications after forwarding messages. When nodes encountered MTB in the future, nodes can use the qualifications we called “Receipt Data” to get the reward we called “Incentive Credit” from MTB, above process called “Credit Clearance”. Nodes can use these incentive credit to request other node to forward messages. If node s help more times, nodes will get more incentive credit. Selfish nodes are stimulated to help forward messages with credit-based incentive strategy. The strategy can further stimulate DTN nodes to improve the DTN’s performance in terns of delivery ratio.

In order to implement selfish nodes environment in simulator, we used following arguments. ESC denote as a node encountered selfish node count(maintain by Sucker), when a node encountered a selfish node, ESC will add one. Nodes will be died when ESC over ESC Threshold (maintain by sucker), it means nodes has been cheated too much times, and don’t help forwarding for every nodes anymore. Grudger Threshold represents threshold of number of times a Grudger node can be cheated, when ESCi (denote encountered selfish node count for node i, maintain by Grudger) is over Grudger Threshold(maintain by Grudger), then Grudger don’t help node i forwarding in the future. CCi represents Ecci cheated count for node i (maintain by Ecci), when CCi over Grudger Threshold, Grudger wouldn’t help Ecci, so Ecci ‘s behavior turn to Grudger’ behavior to avoid Grudger discover. And each type of nodes behavior as following :

Sucker and Grudger’s Behavior.

As above mentioned, Grudger can effectively restrain destruction of the selfish nodes. So we design MTBIS based on Grudger’s behavior, include characteristics of Grudger : Grudger will not help a node forwarding if has been cheated over Grudger Threshold , and also include SI and DGT. In forward period when a node in transmission range , MTBIS determine whether a node is friend first (part of SI), then determine the messages are paid or not. When the conditions is positive, node will help forwarding message, if is negative, then determine a node has been cheated over Grudger Threshold, if negative then help forwarding, and don’t help forwarding if positive. During the moving period, nodes will be based on the nodes that have been encountered selfish nodes in the past statistics (maintained in History List) to calculate Grudger Threshold dynamically after forwarding (part of DGT). Nodes will get incentive credit from MTB (called Credit Clearance ) after forwarding.

Since adding SI(Social Incentive), Cheater are not like before that don’t help forwarding for every nodes, they begin to help forwarding message when they are friend, otherwise, drop message when they are not friend such as before. And since begin to help forwarding message, they have qualifications to get incentive credit from MTB.

MTB have a fixed moving path, moving faster and larger transmission range of characteristics, the delivery ratio more higher than other nodes. Since those reasons, MTB will help to all nodes forwarding message as possible.

We did some experiments is to show the performance of protocol under various number of Cheater. In terms of impact of selfishness, the simulation results show, as expected, that the number of the successful message delivery ratio decreases directly proportional with the number of selfish nodes. (for this simulation, the number of Ecci, Grudger and Sucker nodes are fixed, Cheater is variable) After

adding Dynamic Grudger Threshold, Grudger adjust Grudger Threshold dynamically. As represented in Fig. 3, MTBIS always has higher delivery ratio than traditional protocol (Epidemic, PROPHET and Spray and Wait), this is important shown incentive strategy that stimulate nodes to help forwarding. When the number of Cheater nodes reaches 160 ( the number of Cheater is 4 times than Grudger), Grudger is difficult to restrain Cheater, so delivery ratio begin to degrade. In traditional protocol, Epidemic has higher delivery ratio than other protocol, because Epidemic use flooding manner to forward, a node continuously replicates the messages it has and sends all of them to all the nodes it encounters, if they don’t already have it. In hypothesis, Epidemic has most performance, but not in selfish node environment, since the selfish nodes could jeopardize the performance severely. On the contrary, the delivery ratio of Spray and Wait is the best in traditional protocol, it limited the number of messages copied, reduce the chance that selfish nodes jeopardized. PROPHET use history to predict probabilities for the successful delivery for each node, but in selfish node environment, it’s more difficult to predict, therefore, it has worst delivery ratio.

The result also shows that when MTBIS encounters a large number of selfish nodes (number of Cheater 160 to 280), it can resist to being cheated by selfish nodes effectively, and unlike other protocols, its performance do not degrade dramatically in such environment.

Another interesting finding is that when a DTN has a small number of selfish nodes (around 40 to 80), its performance actually seems to increase in contrast. This is because, in a completely non-selfish nodes configuration, the number of message exchanges is high, so the node’s buffers tend to overload more quickly and thus more messages are dropped and never reach their destination. Small number of selfish nodes can relieve the situation a node’s buffer size is full that can not store more message.

In order to mesure the impact of selfishness, as represented in Fig. 4, all of protocol has lose least 29% delivery ratio compare with unselfish node environment. In original MTBIS (without SI and DGT), has most impact of delivery ratio loss from selfish nodes (-41%). After adding SI the delivery ratio are improved reaching -36%, since it stimulate selfish node to forwarding message. After adding DGT are improved reaching -29% further compare with unselfish node environment, since DTG can calculate Grudger Threshold dynamically according to the number of selfish node from history list. The results represented the delivery ratio of traditional protocol degrade dramatically (more than -46%), this is why the reason we must consider the selfish node in order to implement in real world.

Fig. 5 The Delivery Ratio results of Unselfish and Selfish Environment.

3. Popularity Spray and Utility-based Forwarding Scheme with Message Priority Scheduling in Delay Tolerant Networks

3.1 Abstract

Delay Tolerant Networks (DTNs) use the “Store-Carry-then-Forward” approach to deliver the message to the destinations. It relies on the intermittent link that occurs when two nodes contact with each other due to mobility. In this paper, we propose a three-phase algorithm (SFMS: Spray and Forwarding scheme with Message Scheduling) that integrates the concepts of flooding-based and forwarding-based protocols, and considers message priority. The idea of SFMS is to periodically predict the contact popularity and contact association among nodes, such that we can determine the fast message spraying and efficient forwarding strategy. Furthermore, we propose a message scheduling mechanism to enhance the resource allocation. Simulation results show that our scheme

has a better performance for delivering message. Besides, it also achieves a differential delivery performance for different priorities of message while maintaining a better resource allocation.

3.2 Main Results

In view of most existing routing protocols in DTNs, it still leaves some issues to be further investigated. Here, we will focus on improving some of these problems, and propose a routing protocol by integrating the improvements we achieve. Our scheme could be divided into three parts, (1) Popularity Spray Phase, (2) Utility-based Forwarding Phase, (3) Message Forwarding with Priority Scheduling Phase. In the following, we will introduce in detail each of three phases, including how to improve the problems of the that existing protocols, and how to achieve a differential performance for message with different priority.

[Popularity Spray Phase]

When nodes are moving with a specific mobility pattern, they would have their own predefined attributes. The Popularity Spray will redistribute the N copies of a message that is held by the sender and the receiver according to their total counts of contacted nodes in the last period of time. The spraying formula could be modified to ⌈( 𝑁) ⌉ copies for the sender and ⌊( 𝑁) ⌋ copies for the receiver, where CCi and CCj means the contact counts of nodei

(sender) and nodej (receiver) in last time period,

respectively.

The idea in Popularity Spray is to let a node which is more popular in the past will keep more message copies to spray, because it could spray these message copies faster than the node which is less popular within the same time in a regular mobility pattern. In this phase, the less message spraying delay time, the better performance (Latency) could be achieved

[Utility-based Forwarding Phase]

If the message could not be delivered to their destination during the spraying process, the message will be switched to utility-based forwarding phase. This means each of the N nodes that have a message copy would stop duplicating the message to other nodes unconditionally, instead, a directional way to guide to its destination. The design philosophy in this part is to let the message yet to be successfully delivered could have chances to be further forwarded to the nodes which have a higher delivery utility

(weighted value) to the message’s destination. Through the utility guidance instead of blind flooding message to other nodes, we can decrease the delivery overhead while increasing the throughput of message delivery. Therefore, how to design a proper utility function in this phase is our main work. Four utility functions have been commonly used in many researches: contact frequency, contact duration, encounter aging and location. In our scenario, the GPS is not considered to be used for an auxiliary tool, the location and moving speed related information is unknown. Because the utility of contact duration has been proved that it has a higher accuracy than the utility of contact frequency, we utilize this utility concept to design a more efficient message forwarding approach. Each node in our system will hold a Node State Table (NST), shown as the following Table.

The calculation of Delivery Utility is as following steps: ( )

_{( )}

̃ ( ) ( ) ( )

( ) ( ( ) _∈𝑁 ( ̃ )) ( )

The first formula shows the direct contact delivery utility between node(i) and node(j), where CD(i,j)

indicates the total contact duration in a time period, and T indicates the time period we predefined. The second formula shows the indirect contact delivery utility between node(i) and node(j). It means that node(i) and

node(j) could indirectly contact through another node.

The third formula shows that the final delivery utility is determined by choosing the highest utility from direct contact and all other indirect contacts. The delivery utility will be periodically (every T) updated. In order to make the delivery utility more accurate for reflecting the network situation, we put both the old utility and the new utility into consideration in every update period shown as:

( ) ( ) ( ) ( )

where is used to represent the network state, we use the contact counts to evaluate , and is equal to

.

Figure 6: An example of message forwarding. Whenever two nodes contact with each other, they will first exchange the NST. By consulting the NST, a node knows which messages can have a better delivery performance if they are carried by the contacted node and will be chosen to be further forwarded. As shown in Figure 6, node(A) knows message_08 has a better

delivery performance by node(B), and node(B) knows

currently there is no message better for delivering by node(A).

[Message Forwarding with Priority Scheduling Phase] In DTNs, the contact among nodes may not last a long time, probably very short and unstable due to node mobility. Thus, during each contact, a node probably does not have enough time to deliver all the selected messages to the contacted node. Hence, the message delivery sequence could directly affect the ratio of successful delivery to the destination. Therefore, we propose an approach to schedule the message forwarding sequence according to the cost to the destination along with a contention mechanism based on the message priority. In our protocol, we divide message into four priorities. A message will be automatically assigned a priority when it is created. We wish to take both message priority and the cost to its destination into consideration, making it possible of a differential delivery performance for different message priorities in DTNs. we use the elapsed time (ET) since the destination last met as the cost for a node delivering message to the contacted node. The longer the time elapsed indicates the more distance to the destination. If delivering a message which has a longer elapsed time for the contacted node may cause the contacted node still hard to deliver the message to its destination. Therefore, the basic concept is that the less the elapsed time that a message has for the contacted node, the more advanced that a message could be transmitted to it. Besides, to achieve a differential performance among the four priorities and to avoid the transmission opportunity strictly on high priority message, we apply the contention mechanism that derived from the backoff time of EDCA in IEEE 802.11e to sort the forwarding sequence of message. An example of this phase is illustrated in Figure 2.

<Node State Table of node B>

Node Delivery Utility ET CC BufferId DeleteId.

C 0.7 1080 15 03 07 11 15 01 02 05 09 D 0.8 558 E 0.7 1320 G 0.1 1907 K 0.6 89

Figure 7: An example of calculating message forwarding sequence

For simplicity, we demonstrate message forwarding process for one message in Figure 7. There are five messages (red color) which is decided to be transmitted to node B. The sequence for delivering is calculated by considering message’s Cost and Weight Range of priority, using message’s ET multiplied by a weighted value randomly chosen from its Weight Range. We can get the FS (forwarding sequence) value of a message, then sort all the FSs increasingly. the smaller the FS value, the earlier the corresponding message can be sent. Note, because a message may be in Spray Phase or Forwarding Phase. In the whole delivery process, a newly created message to be sprayed to distinct N nodes is the first step, hence, the messages in the Spray Phase are always be sent before the messages in the Forwarding Phase.

[Simulation Result]

From Figure 8 to Figure 10, our SFMS algorithm has a better delivery ratio among all the compared algorithms while maintaining a very low overhead ratio. Epidemic suffers from huge redundant messages copies, it would also cause too many messages be dropped so that the messages could not efficiently be carried to their destination. Although PROPHET uses the history of contact frequency, it may cause the accuracy not enough to be a good forwarding indicator. Especially PROPHET still suffers from heavily overhead ratio, and in response to the delivery ratio is also a worse performance. Spray And Wait has a medium delivery ratio, and the lowest overhead ratio because it restricts the number of a message that could be copied. UDM has a similar routing step and forwarding strategy with SFMS, hence, it has almost the same the low overhead ratio with SFMS. But in SFMS we adopt a popularity spray strategy that could more efficiently perform the distribution of the N message copies, and in the forwarding process we import the aging of contact to more precisely guide the transmission sequence. Therefore, SFMS could achieve a better delivery ratio than UDM through all the buffer size in the simulation.

Note that comparing with UDM, SFMS has a better performance of delivery ratio that is more obvious in the condition of small buffer size, but slightly better than UDM in big buffer size. It is because the bigger buffer size could store more messages and the chance for the message to be dropped would also become smaller. Hence, the message could be carried longer in the process of delivering. Therefore, our method may not benefit the delivery performance that much.

Figure 8: Delivery ratio vs. Different buffer sizes

Figure 9: Overhead ratio vs. Different buffer sizes

Figure 10: Delivery delay vs. Different buffer sizes

三、計畫成果自評 本計畫的主要的研究主軸為 DTN 環境下的適地 性內容服務的實踐議題。主要的貢獻是研究出了一 個適地性的內容搜尋機制，包含 Query 的傳遞，以 及 Reply 的回傳等機制，如何用最少的複製達到最 好的查詢/回複的成功率，是這個研究很重要的成 果之一。本篇已經發表在國際會議論文了，之後又 被推薦於期刊的收錄，即將在明年一月出刊。可見 此部分的研究成果極受肯定。 另外的重要研究議題之一是 incentive 相關的 研究。這方面是關於人的行為以及動機的研究，為 0 1 5 25_{Buffer Size (MB)} 50 100

Delivery Ratio

Epidemic SAW Prophet

0 100

5 _{Buffer Size (MB)} 25 50 100

Overhead Ratio

Epidemic SAW Prophet

3000 13000 5 25 50 100 Sec Buffer Size (MB)

Delivery Delay

啥要在 DTN 的環境中幫別人 forwarding，這對”實 踐”DTN 是很重要實務的問題。我們也提出了一個非 常實務的模型，以及如何處理 selfish 的議題，搭 配的路由方法，因其確實考慮此因素，因此可以比 傳統的方法有較高的成功傳遞率。這部分的研究可 說是 DTN 能否在實際社會中能實踐的重要參考價值。 第三部分的研究則在同時考量 priority 以及傳 送成功率之相對成本，提出一個綜合的 forwarding 機制。結果也證明確實能提高不少的成功傳遞率， 同時也支援了不同的服務等級的 priority，讓在某 些特定的應用，使得較緊急或優先權較高的應用， 也可在 DTN 環境中得到較好的服務品質，使之成為 可能。 綜觀這三個面向的研究，完整涵蓋的 DTN 的實 踐所需考量研究的議題，我們在本計畫中提出了完 整的全套方法與機制，相信對日後 DTN 的實際運作 系統的建置，提出了不錯的參考依據，作為其扎實 的 基 礎 理 論 。

1

國科會補助專題研究計畫項下出席國際學術會議心得報告

日期：102 年 12 月 28 日

一、參加會議經過

本次會議在桂林舉行，地點選在桂林賓館，會議由 keynote speaker 開始，共有二

場。第一場，由德國的 University of Duisburg-Essen 的 Prof. A. J. Han Vinck

開場主講，講題為”Noise Models and Noise Mitigation”。他提到了目前數位通訊

系統，使用的 scramble 技術，如何對抗 noise 以及 error。講到的議題包含 error

coding, transform, receiver detection 等。提及了目前的技術以及挑戰，對下一

代的通訊系統的發展，提示了重要的研究方向。

計畫編號

NSC 101－2221－E－004－005

計畫名稱

適於 DTN 環境下的適地性內容服務之實踐議題研究

出國人員

姓名

蔡子傑

服務機構

及職稱

政治大學資訊科學系

副教授

會議時間

102 年 08 月 14 日

至

102 年 08 月 16 日

會議地點

Guilin, China

會議名稱

(中文)

(英文)IEEE 8th International Conference on Communications and

Networking in China (IEEE Chinacom 2013)

發表論文

題目

A Location-based Content Search Approach in Hybrid Delay

Tolerant Networks

2

第二場為來自義大利 Univeristy degli Studi di Milano 的 Prof. Vincenzo Piuri

主講，講題為 Dependability in Cloud Computing。雲端議題可說是目前最夯的議

題了，所以大家都洗耳恭聽，聽眾此時變得很多。他講的雲端議題，主要是圍繞在

security 以及 fault-tolerance 相關的議題。強調在 deployment 時的動態任務，包

含時間的限制，以及 task 有先後順序的限制等的挑戰。最後結論出雲端是最安全受

保護的骨幹，可提供容錯的服務給使用者，以 deploy 使用者的應用。

中飯過後，就開始 technical session 了，我的論文被安排在下午的第二場

workshop(WCN02: Hybrid Delay Tolerant Networks)，這個主題完全符合我發表的

論文”A Location-based Content Search Approach in Hybrid Delay Tolerant

Networks”。同一個 session 的第一、二篇為來自北京交大，第三篇是我的，第四篇

為來自南京郵電。雖然剛好都是來自中國的，不過聽眾有幾位是從美國來的，也有

澳洲來的。

報告完後，來自南京郵電大學的王堃教授主動來跟我對談，因為他對我的報告內容

很有興趣，他在這個領域也初有涉略，對我的報告讓他有嶄新的收穫，希望能繼續

跟我交流，我們就互換的名片，留下聯絡方式。後來回國後，也收到他的來信，希

望我們將來有機會能進一步交流。

二、與會心得

參加國際會議當然最重要的就是交流與吸取別人研究經驗與成果分享，這個會議舉

3

辦在桂林。根據多次在中國參與國際會議的經驗，中國在舉辦上非常的積極與盡心

盡力，這點我們國內倒應該要好好努力才成。而且整個中國 ICT 在政府相關的大型

計劃都投入相當大的資源與人力，相信再過不久，他們的成就是可預期的。這不禁

讓我們感到有點憂心，台灣再不努力的話，很多技術可能都會被大陸趕過去，學生

的用功程度與企圖心也一代不如一代，應該多鼓勵補助研究生多來參加會議，看看

別人想想自己，自己來看絕對勝過老師苦口婆心。

另外，因為桂林山水甲天下，來到桂林當然要去看一下山水了。桂林這個城市其實

很大的收入來源就是觀光了。然而讓我印象深刻的是，它們的一些景點門票的收入

動則人民幣百元以上，但是本地人卻只要幾塊錢就行了。我覺得這樣是好的，在台

灣很多觀光景點的門票都沒有分本地人外國人，其實可以考慮分開收費。畢竟有些

景點確實也是台灣之寶，對外國人來說付這樣的門票應該是值得願意付的，這樣可

以大大提升我國的觀光經濟。

三、考察參觀活動(無是項活動者略)

無

四、建議

無

五、攜回資料名稱及內容

大會論文集 USB 隨身碟一個

六、其他

無

A Location-based Content Search Approach in Hybrid Delay

Tolerant Networks

Tzu-Chieh Tsai

National Chengchi University Taipei, Taiwan ttsai@cs.nccu.edu.tw

Hsin-Ti Lee

Department of Computer Science National Chengchi University

Taipei, Taiwan g9910@cs.nccu.edu.tw

Abstract—In Delay Tolerant Networks (DTNs), the offline users

can, through the encountering nodes, use the specific peer-to-peer message routing approach to deliver messages to the destination. Thus, it solves the problem that users have the demands to deliver messages while they are temporarily not able to connect to Internet. Therefore, by the characteristics of DTNs, people who are not online can still query some location based information, with the help of users using the same service in the nearby area. In this paper, we proposed a Location-based content search approach. Based on the concept of three-tier area and hybrid node types, we presented four strategies to solve the query problem, namely, Data Replication, Query Replication, Data Reply, and Data synchronization strategies. Especially we proposed a Message Queue Selection algorithm for message transferring. The priority concept is set associated with every message such that the most “important” one could be sent first. In this way, it can increase the query success ratio and reduce the query delay time. Finally, we evaluated our approach, and compared with other routing schemes. The simulation results showed that our proposed approach had better query efficiency and shorter delay.

Keywords: Delay Tolerant Networks, Location-based, Content, Query, Routing protocol

I. INTRODUCTION

The mobile network technology is approaching to maturity recently, e.g. 3G(UMTS, CDMA2000), 3.5G(HSDPA) or 3.9G(WiMax), even 4G, 5G. And they already have a bunch of commercial products. People can use their mobile devices to play with Facebook or search information about their location. Although the usability of the mobile network is well in most of the cases, it may still not be available or stable in some scenarios: (1) Temporarily no network coverage: some location may have no network signal or temporarily not stable signal; (2) Too many users: if there are too many users use the same service in the same area simultaneously, it will cause serious network traffic and may congest the network; (3) Infrastructure broken: if the base station of the cellular networks or the Wi-Fi access point doesn’t work, users can not connect to the Internet; (4) User did not subscribe to the mobile network service. Even though for some reasons, the user can’t connect to the Internet, we still can adopt the technique of “Delay Tolerant Networks” (DTNs) to fulfill the need of sending message or searching information.

Delay Tolerant Networks (DTNs) are intermittently connected networks [1]; all nodes are lacking of continued

connectivity [2]. It can be used in some cases, like Battlefield [7], Outer space [8], disaster or emergency environment [9] [10], etc. Node-to-Node transfer message is the main feature. However, the node sends message to the destination by the encountering nodes, which leads to take much time on delivery. If the application can tolerate long delay, then DTN technique can be applied to solve the challenging instable networking problem.

As we are located in an area where no stable Internet connection, but we want to search some information about that area. What can we do? The DTN approach can give the solution. First, ask the encountering node for the information. If it doesn’t have, we can ask it for help to spread the searching message and so far so on.

Our objective is to let the query user get the information in a shorter time efficiently. In order to achieve this purpose, we assume all nodes in the network are “collaborative” and some nodes can connect to the Internet. The “collaborative” means all nodes are willing to store other nodes’ messages into their storage and help each other to deliver messages. The issue of the forwarding incentive is beyond the scope. To this end, we presented four strategies to solve the query problem. They are Data Replication, Query Replication, Data Reply and Data synchronization strategies (in Section 3).

The rest of this paper is organized as follows. Section 2 introduces related works in common DTN routing protocols and the Query-based DTN routing protocols. In Section 3, we present our searching rule, location-based content search approach. Section 4 presents the simulation results. Finally, Section 5 concludes this work and discusses the future work.

II. RELATEDWORK

In the past research, most literatures [3-6, 15-17] focused only on how to send data from source node to destination node, but little considered the message response problem. Strictly speaking, it is involved with one way transmission protocol. If a node sends a request to destination node, it wouldn’t know the message being delivered successfully or not. Therefore, one way transmission is not enough. We have to consider how to respond the message to let users can not only send query but also get reply information. Several works focused on query in DTNs [11-14, 18]. In these papers, nodes send requests to the message owners, and the owners have to send the response message back. The challenge is all nodes are

moving and we don’t know where they are g is difficult to get the answer.

N. Thompson et al. [11] proposed a q which used three-tier location area concep look up the relevant information according Every data object will be given a utility val will vary with the distance between the creat data object and the position of the node whic This value is used to forward data.

P. Yang et al. [12] proposed a complete search steps. It includes Data copy, Quer reply. Its main contribution is the use of Fr (FM, the average number of nodes it enco observation window) to decide on the data de C. L. Jie et al. [13] introduced a Hyb scheme in file sharing. Every data will be d fragments. And every fragment has a description. Then these metadata will be network. When the user wants to query a da all the metadata to download all the fragmen In [14], they focused on query - how to many queries are replicated. An estimation r avoid too many query replicates. It will transmission overhead and look up data e approach can estimate the query rep accurately,

III. LOCATION-BASED CONTENT SEAR

As mentioned in previous sections, our two-way query/reply message can be transf time efficiently. We proposed the Locatio Search Approach (LCS) to solve the query p disruptive Internet connection. In our rese nodes in the networks are categorized into tw (1) Online, and (2) Offline. Online-node c Internet to access the remote server, and O connect to that.

To sum up, our goal is to let Offli information they need by help of other efficiently. In order to accomplish this, w strategies: (A) Data replication, (B) Query rep reply strategic, (D) Data synchronization and

We assume the area range of the poin known, and we divide the area into three p Border Area and Outside Area (shown as in F

going, so the asker query mechanism pt. Users want to to their positions. lue, and this value tion position of the ch carries this data. e approach of data

ry copy and Data riendliness Metric ounters during an elivery.

brid DTN routing divided into many

metadata for its e spread out the atum, it has to find nts.

forward and how rule is proposed to greatly decrease efficiently, if this plication number

RCH APPROACH

goal is to let the ferred in a shorter on-Based Content problem in case of earch scenario, all wo types of statuses: an connect to the ffline-node cannot ine-nodes get the

nodes soon and we proposed four plication, (C) Data

update strategies. nts of interests is parts: Inside Area, Figure 1).

Figure 1: Three

Inside Area is within a radi outside the Inside Area mete area is Outside Area. This inte area use different ways to sel order to identify the node posi formula[19] (Equation (1), see the earth’s ellipsoid, and use calculate the shortest distance accurate within 0.5 millimeter. and the Area Central Point (CP

, it indicates the node + , it indicates if + , it means the no

= (

When two nodes encounter, first which includes node’s pos or Offline node), node past mo nodes, node collected Data me index list and Reply message in

A. Data replication stategy

When two nodes A and B e shown in Equation (2). If the d Area Central Point (CP) is le indicates Node A is in the encountered Node B, it will ad Data dataset D_set to prepare t purpose is to let the Inside A messages. If any node wonder get the Data messages quickly However, the encounter tim transmission are limited, two metadata before transport. T identical Data messages, and ad distance d between Node A a and less than γ + , it indicates According to Node B’s past p Node B will likely enter to the the Data messages into the D_

Node A and CP is greater than the Outside Area, then, Node A

We use ∆ . to predict w the Inside Area or not. In Equ between Node B position Pt-1 a

is the distance between Node B 2. Then, subtract . from than 0, it means Node B is like the result is greater than 0, it m away from the Inside Area.

( , _ )

=

, .

, . .

,

parts area figure

ius of γ meters from area central, rs is Border Area and the other ention is to let nodes in different ect messages for delivering. In ition, we refer to the Vincenty’s e [19] for details). It considers es Longitude and Latitude to e s, and the deviation can be We use the node position point P) to calculate the distance d. If

is in the Inside Area; if s the node is in the Border Area;

de is in the Outside Area.

∆ ) (1)[19]

, they will change their metadata sition, node’s type (Online-node oved path point, node encounter ssage index list, Query message ndex list.

encountered, the selection rule is distance d between Node A and ss than or equal to radius γ, it Inside Area. When Node A dd all the Data messages into the

o send to the Node B. The main Area be filled with related Data red to query in this area, it can and has high query success rate. e of the two nodes and the o nodes have to change itself Therefore, Node A skips the

dds the others into D_set. If the and CP is greater than radius γ s Node A is in the Border Area. path, if ∆ . 0, it indicates e Inside Area, then, Node A add

_set. If the distance d between n γ + , it indicates Node A is in A should not do anything.

whether Node B’s will enter to uation 3, . is the distance

and CP at time t-1, and . B position Pt-2 and CP at time

t-m . , if the result is less ely to enter to the Inside Area; if means Node B is likely to move

+

2 ∆ . 0