耐延遲網路環境下由於節點的高移動性及傳輸距離等限制，無法保證兩節點間持續存在有一條固定的資料路由路徑，嚴重影響封包路由機制的運作，因此為了改善資料傳輸效能，傳統的耐延遲路由設計大多是利用增加封包複製數，或透過與相遇節點建立機會性連線時，選擇有利的轉送節點等方式來增加封包抵達率或減低傳輸延遲。然而，過度增加封包複製數及選擇轉送節點所需額外記錄的節點資訊，將可能導致網路壅塞及網路資源的浪費，並且延伸出路由協定的低度層次性和高執行複雜度等問題。本論文提出一套基於節點密度感知的路由機制，利用節點分佈具有疏密度的差異，將封包轉送至前往高密度或處於密度較高區域的節點，使之有更大的機會相遇更多節點，從而提高相遇目的地端的機率及縮短封包遞送延遲的時間。另外，於區域密度估測的設計分別討論線狀及樹狀歷史記錄結構，並加入權重、相遇角度等做為考量，以加強區域密度估測的精準度。再者，本機制僅需感知環境中節點密度變化的趨勢，其所需記錄的節點資訊不受系統內節點數量多寡的影響，相較於其他路由方法需要記錄每一節點相遇頻率等歷史性資訊，所提出的設計將可有較好的可層次性與擴展性，同時本機制也僅在網路上複製固定少量的副本數，如此可以避免在有限的頻寬及儲存空間下造成的網路壅塞和資源浪費的情況。最後，論文研究已完成進行大量的模擬實驗，從模擬數據的結果顯示所提出的路由機制在封包抵達率和遞送延遲時間兩項效能指標上皆擁有良好的表現，研究更進一步分析歷史記錄追溯長度和環境疏密程度對於效能提升的效益，充分了解所提出的路由機制在不同環境和量測參數的設定之下所表現出的各式特性。 Delay-tolerant networks emphasize high node mobility and connecting opportunity in wireless and mobile ad hoc network environments. Since network topologies in such environments are partitioned extremely, it is difficult to ensure the existence and reliability of end-to-end paths between any pair of source and destination nodes. Delay-tolerant data delivery mechanisms perform in a store-carry-and-forward routing manner where nodes repeatedly replicate messages and forward message copies to encountered nodes during their movements. In order to improve message delivery ratio and reduce transfer delay, conventional delay-tolerant routing mechanisms mainly apply replication-based or history-based routing protocols to increase the delivery ratio or decrease the transfer deal. However, routing based on message replication can induce extra message traffic and communication overhead; on the other hand, routing based on encountering history information can complicate the routing decision and cause database overhead. The study of this paper proposes a density-aware routing scheme in delay tolerant networks. Considering non-uniform node distributions, it has higher probability to encounter target nodes with lower delay time. This paper formulates the tendency of inter-meeting time between nodes to determine the node density in proximity and then keep message replicas in dense areas as more as possible. This design derives the linear and the tree-based models for density estimation in proximity. The density estimation takes into account the temporal weighting and spatial angle difference to improve the estimation accuracy. In addition, the density estimation depends on only the variance of inter-meeting time, so its scalability is irrelevant to node population in the network. Furthermore, this design generates only a small and constant number of message copies to avoid traffic congestion and resource waste. Finally, this study conducts extensive experiments to evaluate the performance sensitivities to the metrics of message delivery ratio and message transfer delay under a variety of simulation parameters. Consequently, the proposed density-aware scheme enables mobile nodes to estimate the local density and forward messages towards the dense areas, significantly increasing the delivery ratio in delay-tolerant networks.
