| 摘要: | 移動機會網絡在具有無線連接和移動能力的網路設備中建立了一種新的數據傳播方法，其中，移動節點在短時通訊的機會接觸期間發現位置接近的鄰居並交換消息。此外，頻繁的鏈路故障以及缺乏最新的網絡拓撲信息，儲存－攜帶－轉送的數據傳遞模型被用來以延遲容忍的方式中繼消息，由於節點間通訊僅在未經排程的期間發生，在移動機會網絡中，數據傳遞效能將很大程度的受到節點移動的特性以及接觸的概率分佈影響。
在移動機會網絡中，基於複製的路由技術通常用於發送重複的訊息，以增加傳遞到目的地的機會，然而，這將造成巨大的傳輸、存儲與功耗成本，為此，相關的研究包括有效的中繼選擇、訊息轉發、訊息傳輸調度和緩衝區管理被用來降低重複複製訊息的相關成本。本論文的研究也將鎖定在移動機會網絡中的數據傳播方式，提出一種結合緩衝區管理和傳輸調度的方案以及兩種新穎的路由方法：(1) 在移動機會網絡中採用啟發式策略的基於訊息限額的路由和緩衝區管理;(2) 在移動機會網絡中利用移動聯繫模式進行訊息轉發; (3) 利用群體移動性在移動機會網絡中進行訊息傳播。
本論文的研究首先討論了關於節點群內選擇高效率中繼節點和訊息轉發的聯繫模式與其驅動知識的理解。在第二章中， 我們展示了中繼節點之間接觸頻率不同的發現，如一些節點可能永遠不會與另一個節點建立聯繫，而一些節點能夠聯繫多個節點，此外，節點具有聯繫週期性或重複模式，以及群體內的節點聯繫本質上是短暫的。
儘管之前的研究提出了各種緩衝區管理和訊息調度方案，但這些研究主要基於幾個前提，例如全球網絡知識的可用性、無限頻寬容量和同質聯繫模式，此外，先前的研究對結合緩衝區管理和中繼選擇的主題關注較少。在第三章中，我們提出了一種新方案，名為具有有限緩衝區管理與基於訊息限額的路由方法(QRBP)，我們透過選擇一部分移動性較高的節點，委託這些節點攜帶一定配額的訊息副本，並將其傳播給移動性較低的節點。這些節點可以根據訊息限額值、剩餘生存時間與目的地聯繫率的多參數啟發式演算法來調整訊息調度和丟棄以提高資料成功傳遞率，實驗結果顯示，QRBP的方法比傳統的Epidemic, SprayAndWait, Temporal Closeness and Centrality-Based (TCCB) 路由方法效能更加，此外在緩衝區管理方面，我們的方法也比 DropOldest (DO), DropNewest (DN), DropRandom (DR), and Space-Time-Drop (ST-Drop)表現更好。
在第三章中，我們考慮了節點在較大區域快速移動的實際情況，這將導致節點在地圖上不同的移動模式，當中，我們的研究注意到只要有足夠數量的高移動性節點，縱使只複製任何特定限額的訊息仍然可以維持傳遞的效率。因此在第四章中，我們進一步釋放了這種受限的考量，透過多次複製訊息以增加到達目的地的機會，特別是需要檢查聯繫模式的內在屬性並確定兩種聯繫類型，即是在移動機會網絡中的定期聯繫和零星聯繫，其中定期聯繫為具有週期性的接觸，零星聯繫則為不具有週期性的接觸。第四章提出了一種有效的訊息轉發方法，稱為基於移動節點之間的聯繫模式的定期和零星聯繫路由 (RSCR)，首先，我們根據移動機會網絡中任何節點對的接觸持續時間平均值推導出定期聯繫和零星聯繫模式，接著我們結合衍生的接觸模式與剩餘的訊息生存時間來準確確定兩個節點在移動過程中遇到時是否傳遞訊息。最後實驗結果顯示，我們提出的方法與Epidemic 和 PRoPHETv2 方案相比，以更低的成本獲得相同的訊息抵達率。
繼第三章與第四章中的有效緩衝區管理、傳輸調度和節點中繼選擇方案後，我們在第五章中研究節點組群內的聯繫模式。傳統方案通常分析節點之間的直接接觸與長時間觀察接觸資訊，並評估移動機會網絡中節點組群形成的靜態拓樸，然而，這樣的形成方法可能導致節點組群聯繫關係薄弱，並且隨機節點被錯誤地指定到群組中。相反地，我們的研究檢查了一些重要因素，例如特定時期的接觸一致性、接觸頻率和接觸持續時間，用以形成K個群組，並且著重在節點的接觸模式，我們利用接觸持續時間、接觸頻率和在特定時間區段內關於組內和組間接觸模式的聚合接觸屬性，定義了一個新的指標，時間聯繫強度。基於此，我們提出了一種新的路由方案，稱為基於時間聯繫強度的路由 (TS-GBR)，用以提高移動機會網絡中訊息轉發的成本效益性能，並且我們透過幾個真實數據集的模擬證明了此方法與Epidemic、TCCB 、Transient Community-based (TC)、與 Dynamic Transient Social Community (DTSC)相比，獲得更佳的效能。
最後，本文的貢獻可以提高移動機會網絡中訊息傳遞和中繼選擇方案的效率，我們堅信這些努力可集成到當前和新興的無線通信系統中，應用在例如人聯網 (IoP)、裝置到裝置通信 (D2D)、社群物聯網 (SIoT)、無人機網路 (Flying Ad-hoc Networks)、邊緣計算 (Edge Computing)與車聯網 (Vehicular Ad-hoc Networks)。
;The prevalence of networked devices capable of wireless connectivity and mobility propels a new data dissemination paradigm -- mobile opportunistic networking (MON). Wherein, mobile nodes discover neighbors in location proximity and exchange messages during their opportunistic contacts in short-time reciprocal communications. Because of frequent link failures and the lack of up-to-date network topology information in MONs, the \textit{store-carry-and-forward} data delivery model is employed to relay messages in a delay-tolerant manner. Since inter-node communications occur only during unscheduled meetings between nodes, the performance of data dissemination in MONs is heavily influenced by the dynamic nature of node mobility and contact opportunity in mobile environments.
In MONs, replication-based routing techniques are often used to distribute duplicate messages to increase the chances of delivering messages to a destination, which unfortunately leads to enormous costs of transmission, storage, and energy resources. To reduce the costs associated with replicating messages repeatedly, several essential techniques are required certainly: efficient relay selection, message forwarding, transmission scheduling, and buffer management. The study in this dissertation proposes one joint buffer management and message scheduling scheme and two novel routing schemes for data dissemination in MONs: (1) Quota-Based Routing and Buffer Management with Heuristic Strategies in Opportunistic Ad Hoc Networks; (2) Exploiting Mobile Contact Patterns for Message Forwarding in Mobile Opportunistic Networks; (3) Exploiting Group Mobility for Message Dissemination in Mobile Opportunistic Networks.
The study in this dissertation first discusses the understanding of contact patterns and contact-driven knowledge regarding the novel schemes of efficient relay node selection and message forwarding inside node communities. In Chapter \ref{chap2}, we show the findings that contact frequency is heterogeneous between relay nodes, i.e., some nodes may never establish contacts with the others, whereas some nodes may contact more than one node. Thus, nodes have contact periodicity or repeating patterns, and nodal contacts in communities are transient in nature.
Although previous studies proposed various buffer management and scheduling schemes, their efforts were mainly based on several premises, e.g., the availability of global network knowledge, unlimited bandwidth capacity, and homogeneous contact patterns. Prior research has paid less attention to the theme of joint buffer management and relay selection. In Chapter \ref{chap3}, we propose a novel scheme, named Quota-Based Routing Scheme with Finite Buffer Management (QRBP). This scheme selects only a portion of nodes that have higher mobility, and then delegates those nodes to carry a certain quota of message replicas and disseminate those messages to nodes with lower mobility. Nodes can manipulate message scheduling and dropping to improve the successful delivery rate according to heuristic strategies based on several measures of the quota value, remaining time-to-live (TTL), and contact rate with the destination. Simulation results manifest that the QRBP scheme obtains better message delivery performance than Epidemic, SprayAndWait, and Temporal Closeness and Centrality-Based (TCCB) routing and DropOldest (DO), DropNewest (DN), DropRandom (DR), and Space-Time-Drop (ST-Drop) buffer management schemes.
In Chapter \ref{chap3}, we consider the practical situation that some nodes move quickly over a larger area, which leads to different movement patterns in a network map. Then, our study notices that replicating only a quota of any particular message can still sustain the efficiency of message delivery provided with only a sufficient number of high-mobility nodes. Our study in Chapter \ref{chap4} further releases this restricted consideration and appeals to a general case, which is to replicate messages multiple times to increase the chances of reaching a destination. In particular, this general study should examine the intrinsic properties of contact patterns and identify two contact types, i.e., regular and sporadic contacts in MONs. Contacts occurring periodically are defined as $regular$, and occasional contacts are termed as $sporadic$. Chapter \ref{chap4} presents an efficient message forwarding scheme, named Regular and Sporadic Contact-Based Routing (RSCR), which is based on contact patterns among mobile nodes. First, we derive the regular and sporadic contact patterns based on the mean of inter-contact durations with respect to any node pair in MONs. Then, we jointly use the derived contact patterns together with a remaining TTL value to accurately determine whether or not to hand over messages when two nodes encounter during movement. Simulation results show that the proposed scheme attains the same delivery rate at a lower cost as compared with Epidemic and PRoPHETv2 schemes.
Following the efforts of efficient buffer management, scheduling, and relay selection schemes in Chapters \ref{chap3} and \ref{chap4}, we aim to investigate contact patterns inside groups/communities in Chapter \ref{chap5}. Conventional schemes often analyzed direct contacts between nodes, observed contact formation for an extended period, and evaluated a static graph for community formation in MONs. However, such formation methods can result in communities with weak contact relationships and incorrectly appointing random nodes to a community. Instead, our study examines essential factors such as contact consistency in a certain period, contact frequency, and contact duration to form $k$ groups of which nodes render strong contact patterns. Further, we specify a new metric, temporal-tie strength, by utilizing contact duration time, contact frequency, and aggregate contact properties with respect to intra- and inter-group contact patterns during a specific time window. We propose a novel routing scheme, named Temporal-Tie-Strength Group-Based Routing (TS-GBR), which is able to improve the cost-effective performance of message forwarding in MONs. Our simulation with several real-life data sets demonstrates the efficiency of the proposed scheme as compared with Epidemic, TCCB, Transient Community-based (TC), and Dynamic Transient Social Community (DTSC) routing schemes.
Therefore, the contribution of this dissertation can promote the efficiency of message delivery and relay selection schemes in MONs. We believe that these efforts to MONs can be integrated to the current and new emerging wireless communication systems such as Internet of People (IoP), Device-to-Device communication (D2D), Social Internet of Things (SIoT), unmanned aerial vehicles (UAVs), Edge computing, and Vehicle-to-Everything (V2X) in the coming future. |
