3. 最後，根據相遇節點和基線之間的歐幾里得距離的大小，這個方案會排出訊息傳輸的順序。 因此，modified 的地理路由方案將根據路由策略給予不同的複製 payoffs 來增加成功傳送的可能性。此外，本論文設計了一個最佳化管理機制來最大化平均訊息傳遞率。我們聚焦在一個關鍵問題上：當 buffer size 是滿的時候，訊息應該首先被丟棄。我們開發一個使用 Local 的網路資訊的 utility function 進行最佳化每份訊息的 平均抵達率。訊息將根據 utility value 來被丟棄或保存。 最後，我們使用 ONE 模擬器與其他方案相比來檢視我們方案的成果。結果表示我們提出的方案在不同的模擬參數設定下，抵達率大概有 70％以上和較高的平均延遲時間。;In delay tolerant networks (DTNs), persistent end-to-end route paths from the source to the destination are not truely existed. The DTN technology aims to solve the challenges and obstacles of data transmission caused by environmental factors, such as long distance transmission, uncertain network topology, sparse node density and high mobility. Unlike conventional routing protocols in DTNs, the geographic routing strategy is the only one that does not need the network addresses of nodes to obtain the information of the route paths in a network. Nodes use the satellite positioning system or other positioning systems to decide a routing mode and the next relay node, which is chosen to transfer messages based on the store-carry-and-froward scheme. However the geographic routing model is difficult to predict the follow-up movement of nodes based on historical records of node movement in high dynamic network environments. This situation affects the policy of choosing the next relay node and decreases the opportunity to deliver messages to the destination. In addition, the constrained buffer space of nodes makes the congestion problem more possible. This thesis proposes a modified geographic routing scheme which is suitable in high dynamic DTN environment. 1. This scheme uses the encounter node’s moving direction and base line which is established between source and destination to get the encounter node’s moving angle. 2. After getting the moving angle, this scheme will follow specific conditions to choose the next relay node. 3. At last, according to the size of the Euclidean distance between an encounter node and base line, this scheme will row the queue of the message transmission. Accordingly, the modified geographic routing scheme will give different copy payoffs to increase the probability of successful delivery according to the routing strategy. Furthermore, this thesis designs an optimal buffer management to maximize the average message delivery rate. We focus on one key issue: messages should be dropped first when the buffer is full. We develop a utility function using local network information to optimize per-message of average delivery rate. Messages will be dropped or kept according to their utility values. Finally, we use the ONE simulator to examine the performance of our scheme in comparison with other schemes. The results indicate that the delivery rate of our proposed scheme approximately has 70% and higher average delay in different value sets of parameters in simulation.
