Research On Routing Protocol In Delay Tolerant Network

In the last decade, wireless ad hoc network were widely applied in many areas dueto the emergence of plenty of intelligent wireless devices. However, in some specificnetworks, intermittent connectivity might arise due to low density, high mobility,limited transmission range and so on. In such kind of networks, it's hard to find acomplete route from a source to a given destination, and the data delivery may sufferfrom long delay. Accordingly, these networks are defined as Delay Tolerant Networks(DTNs).Due to the incomplete connectivity, routing is an important issue in DTN, and thetraditional routing mechanisms in Mobile Ad Hoc Network (MANET) cannot beadopted directly, thus the key challenge in DTN is to design the efficient routing schemewhich can provide good delivery performance in front of the intermittently connectednetwork topology. This dissertation focuses on the routing strategies of DTN in twoscenarios: the mobile sensor network (MSN) and the DTN with properties of socialnetwork. Based on the comprehensive analysis of the current research, the author makesa thorough study about the routing strategies in those two scenarios. The majorcontributions of this dissertation are elaborated below.1. How to delivery message to the sink node efficiently in the changeableenvironment is the objective of MSN routing. This dissertation proposes a locationprediction based routing protocol (LPR). LPR uses the relative distance between thesensor node and the sink node to evaluate the possibility of delivering message to thesink node. LPR is different from most existing routing protocols which make routingdecisions only according to nodes historical locations, while nodes moving tendency inthe near future are not considered. Comparing with these protocols, LPR makes locationpredictions for nodes based on k-order Markov chain model, and the predicted resultsare further used to provide guidance for route selection. Meanwhile, LPR assigns tasksof message replication to nodes according to their abilities of delivering message to thesink successfully. Moreover, LPR provides an effective buffer management mechanismto decrease the redundant copies as well as improving the delivery performance. Simulation results show that compares to existing DTN routing protocols, LPR achievesbetter tradeoff between the message delivery ratio and the delivery cost/delay.2. According to encounters between nodes, in this dissertation, a deliveryprobability and activity based routing protocol (DPAR) for MSN is proposed. DPARuses the delivery probability to represent the possibility of delivering message to thesink node, which is achieved according to the encounter duration between nodes. Inorder to deliver messages to the sink node successfully, nodes with higher deliveryprobabilities will be selected as next hops. However, in a typical MSN with large scaleand sparsely nodes distribution, when the message custodian with low deliveryprobability moves along the regularly route or in a limited range in most of the time, thedata delivery may be subject to high latency since the custodian will often have to wait along time until it encounters the node with higher delivery probability. In order to solvethis problem, DPAR leverages active nodes as relays. DPAR also employs an effectivebuffer management mechanism to decrease the redundant message copies as well aspromoting the delivery performance. Simulation results show that DPAR achieveshigher delivery ratio with lower delivery delay comparing with existing routingschemes.3. In DTN with properties of social network, portable devices are carried byhuman, this dissertation proposes a preference location based routing scheme (PLBR)based the observation that human have location visiting preferences in their dailymobility traces. PLBR acquires one's preference locations according to its long-termand periodically location visiting records, and then the degree of closeness of any pairof nodes is measured by the similarity of their preference locations, moreover, thecloseness metric can be used to provide guidance for data forwarding. Simulations arecarried out based on human mobility traces from MIT Reality project, and the resultsshow that PLBR achieves better tradeoff between the data delivery ratio and thedelivery overhead/delay comparing with existing schemes.4. Based on PLBR, an expected shortest path based routing protocol (ESPR) isproposed. PLBR only considers the direct connection between the current node and thedestination, and messages are thereby forwarded to the node which is the one-hopneighbor of the destination. Different with PLBR, ESPR utilizes the indirect multi-hopconnection between the current node and the destination. Specifically, ESPR gets the direct distance between nodes according to their closeness, and then it employs theDijkstra algorithm to calculate the expected shortest path (ESP) from the source to thedestination, furthermore, messages are forwarded to nodes which have shorter ESPsfrom themselves to the destination than the current node. In brief, ESPR makes a betterutilization of the indirect connection chance between nodes than PLBR. Simulations arealso carried out based on human mobility traces from MIT Reality project, and theresults show that ESPR achieves better performance comparing with existing routingprotocols and PLBR in terms of delivery ratio and delay.