Research On Routing Mechanism And Efficient Data Delivery Technology In Mobile Delay/Disruption Tolerant Networks

Delay/Disruption tolerant network (Delay Tolerant Network, DTN) is a special network architecture receiving wide attraction after Mobile Ad hoc Network (MANET), In case of the hostile communication environment, unstable channel or rapid link changes, it becomes an efficient support to maintain basic communication for critical data transmission. The underministic nodal mobility, long delay, constrained nodal energy or interruptible channel, etc, potentially lead to link interuption and network splitting, thus cannot guarantee an end-to-end connected path for data transmission. Therefore, endpoints (or network nodes) have to store the data in case of network disconnection and forward data during the next communication opportunity. Data routing is the primary issue in DTN. This dissertation targets the UnderWater Sensor Network (UWSN) and Mobile Opportunistic Network and explores how to establish efficient routes based on current network configuration from network layer.First of all, for the network separation problem due to channel variation, we propose link prediction and route selection based on link state detection. When the communcation channel suffers from frequent changes, signal transmission is often disruppted from link failure, so that the resulting massive propagation latency, memory overwhelming, etc, will finally lead to data delivery failure. Being aware of the potential link disconnection and the possible disconnection duration in advance, nodes can take effective measures to avoid using this link during route selection. Based on Kalman filter, this paper proposes the link prediction model based on channel detection, which estimates the chance of link disconnection and the link connection duration to help nodes achieve more efficient route selection. Once the link disconnection is confirmed, a route decomposition and reconstruction approach is presented to realize route reselection. The simulation verifies the feasibility and efficiency of the proposed approach.Furthermore, the communication interruption resulting from nodal mobility is exploited, thus efficient data communication is achieved by making use of the contact events between nodes. The routing strategy is derived from an optimization model. Taking the expected successful delivery ratio as the constraint, we try to obtain an optimal transmission strategy to minimize the overall communication overhead based on the predefined time budget. The successful delivery ratio and communication overhead are calculated by the delay distribution between two nodes. Plugging the two parameters into the optimization model, the optimal transmission strategy can be attained. In addition, to reduce the computation complexity and accelerate the algorithm scalability, a distributed approach is proposed to approximate the centralized one, which can be applied to larger scale networks. The proposed algorithm is implemented on Dell Streak tablets in an experiment with 25 nodes for a period of two weeks. Moreover, the algorithm codes are extracted from the prototype to run simulations based on Haggle trace and evaluate its performance trends under various network settings. Both of them verify that the distributed algorithm can approximate the optimal one even with reduced communication and computation complexity while satisfying the optimization model.At last, in mobile opportunistic networks, we investigate the data dissemination scheme based on constrained resource and delay while considering the competition and cooperation between nodes to achieve the optimal utility. First, we analyze the problem from a centralized perspective and introduce a dynamic programming algorithm to solve it. The centralized algorithm offers useful insights but is impractical to implement in real network settings due to high computation and communication complexity. To address this, two distributed approaches are proposed. One is based on cells in a divide-and-conquer manner by partitioning the network into opportunistic Voronoi cells and then running optimization algorithm in each cell. The other is a task splitting scheme by recursively delegating the recruiting task to newly joined nodes. The former is more applicable in scenarios with more seed nodes while the latter is preferred when fewer seed nodes present. To empirically demonstrate the feasibility and evaluate the efficiency of the proposed algorithms, we have implemented a prototype on Android and carried out testbed experiments using 25 Dell Streak tablets for 24 days. Besides the prototype, we have run extensive simulations based on Haggle trace by varying network settings. Both of them validate that the proposed approaches can achieve the goal of our optimization model and the resulting performance varies from the four different approaches adopted based on different network settings and nodal resource constraints.