【作者】 唐丽均；

【导师】 柴毅；

【作者基本信息】 重庆大学 ， 控制理论与控制工程， 2013， 博士

【摘要】 处理器速度、移动通信和电池寿命的的快速发展促使计算机完成了从完全连接到完全移动的演变，在最极端的情况下，所有节点都是移动的，只有利用可获得的机会：传统的通信基础设施和中间节点的移动性，才能实现节点间的数据通信，这就是机会网络。同时，大量低成本、具备短距离无线通信能力的智能设备的出现，让机会网络在一些极端通信环境中有着巨大的应用潜力。机会网络中的数据通信主要有以下特点：①由于节点的移动导致网络拓扑动态变化，节点对之间不能建立端到端的通信路径，从而引起数据通信的低速率和高时延；②数据成功传输主要依靠节点移动带来的相遇机会，因而根据节点相遇机会制定的数据转发策略成为影响数据传输性能的重要技术；③网络资源有限，对节点缓存空间的需求矛盾较为突出，由于大量数据没有得到及时转发而长时间滞留在网络中，过度消耗节点的缓存资源，因此对缓存容量的利用率要求更为严格。在机会网络中转发数据面临的挑战性问题为：节点移动的不可预知性和网络资源的有限。因此在网络资源有限的情况下，有效的数据转发策略、缓存调度管理是保证数据成功传输的两个关键问题。围绕这两个关键问题，本文主要开展了以下几个方面的研究工作：首先，本文提出了一种基于效用的代表转发策略，该策略根据节点中心度、节点间的关联度和相似度计算出综合效用值，用来表征节点间未来的相遇概率，携带消息的当前节点依据效用值选择最佳的相遇节点作为代表来转发消息，仿真结果表明该算法能更准确地为消息选择下一跳携带节点，提高了网络的消息投递成功率。其次，针对节点静止的机会网络中存在的网络分割现象，提出了静态多摆渡路径设计方案。该方案引入了局部摆渡(Local Ferry)和全局摆渡(Global Ferry)两种移动实体负责不同地区范围内数据通信，为不能直接通信的静止节点之间提供消息转发。并且通过控制局部摆渡和全局摆渡的移动轨迹，实现在线同步协作转发消息，以提高分裂机会网络的消息投递成功率，降低消息投递时延。然后，针对移动机会网络中存在的网络分割现象，提出了局部摆渡节点的动态路径设计算法。该算法在满足节点丢弃消息最小化的条件下，采用流量感知启发式算法为局部摆渡节点计算出最优移动路径，引入固定的中继节点完成局部摆渡和全局摆渡节点之间的异步协作通信，仿真结果表明该算法提高了分裂机会网络的数据传输性能。最后，针对机会网络的间断连接性导致节点间通信时间有限，节点不能将缓存中的消息在一次相遇机会中全部转发，提出了基于接触频率的缓存调度算法。该算法通过计算相遇节点与消息的目标节点之间的历史接触频率，以此确定当前节点缓存中消息的调度优先级，同时结合基于消息副本的丢弃策略，提高了节点缓存空间和链路通信带宽的利用率，消息的转发效率也因此提高。综上所述，本文在机会网络数据转发策略和缓存调度管理方面做了大量的分析研究，针对现有策略的不足，提出了以上几种数据转发和缓存调度管理策略。仿真验证表明，本文提出的相关算法在消息投递成功率、传输时延和网络资源开销等方面能明显改善机会网络的传输性能。更多还原

【Abstract】 With the rapid development of the processor speed, mobile communication andbattery life, Computers evolve from completely wired to completey mobile.In the mostextreme case, even all nodes are mobile, only relying on available opportunities:traditional communication infrastructure and the mobility of intermediate nodes, thedata communication can realize between pair nodes. These are Opportunistic Networks(ONs). Moreover, the appearance of plenty of low-price intelligent devices equippedwith short-distance wireless communication interface boosts the fast rise ofopportunistic networks application in many extreme communication circumstances.Several typical characteristics in terms of data communication in ONs are shown asfollows.①Network topology dynamically change due to the mobility of nodes.End-to-end communication paths can not been established between pairs of nodes,which cause the low rate and high delay take place in data communication;②Successful rate of data forwarding mainly relies on the contact opportunitiesarising from the movement of nodes. Thus data forwarding policy made based onmeeting probability between pair nodes becomes the essential technique, whichinfluences the performance of data transmission.③Contention about the buffer space is very prominent in ONs when the networkresources are constrained. Due to a plenty of data that has not been forwarded timelystays in networks for long time，the storage space of nodes is consumed fast. Thus thenodal buffer is requested to obtain more efficient utilization.Data forwarding in opportunistic networks has two key challeges: theunpredictable mobility of nodes the network resouces constraints.With the limitedresources, Effective forwarding strategies and high-performance schedulingmanagement schemes become two key factors in the process of data transmission.Around such two key factors, study works in this thesis are shown as follows.Firstly, A delegation forwarding policy based on utility is proposed for ONs.Theforwarding policy computes a hybrid utility based on the information that containsnodal centrality, contact degree and similarity between pair nodes. The utility representsthe meeting probability between two nodes in the near future. Current node carryingmessages selects the best encountered node as delegation to forward message based on the utility. Simulation results show that the forwarding algorithm can select correctlythe appropriate carrying node for messages and improve the successful rate of deliverymessage for networks.Secondly, focus on the existence of partition in stationary opportunistic networks,stationary ferry route design for multiple ferries is proposed for ONs. The route designfor multiple ferries introduces two types of mobile entities which are called as LocalFerry (short as LF) and Global Ferry (short as GF). They are responsible for datacommunication within different areas respectively and provide forwarding message forstationary nodes that can not communicate directly. The route design makes LF and GFrealize online collaboration to delivery messages in order to improve the delivery rateand reduce latency for messages in patitioned ONs.In addition, aiming at the existence of partition in mobile opportunistic networks,dynamic ferry route design for local ferries is proposed for partitioned ONs. The ferryroute design utilizes traffic-aware heuristic to find the optimal movement path for localferries with the goal of minimizing message drops of nodes. At the same time,stationary relay nodes are used to complete asynchronous cooperative communicationbetween LF and GF. Finally simulation results demonstrate the multi-ferry routescheme can improve the performance of data transmission in partitioned ONsFinally, to solve the problem that nodes can not forward all buffered messagesduring limited contact duration because of intermittent connectivity of ONs, a bufferscheduling algorithm is proposed based on historical contact frequencies between pairnodes. The scheduling algorithm decides the forwarded priority for buffered messagesin current node based on historical contact frequencies between the encountered nodeand messages’ destinations. Furthermore, combining with dropping policy based onmessage copies, the utilization ratio of the nodal storage and the communication linkbandwidth get improvement，and the delivery efficiency for messages has beenheightened accordingly.In summary, this dissertation widely analyses data forwarding polices and bufferscheduling management policies in ONs. In order to overcome the limitations of theexisting strategies，a few data forwarding scheme and buffer scheduling policies abovementioned are proposed. Simulation experiments demonstrate that the above proposedmethods can significantly improve the performance of opportunistic networks in termsof the message delivery rate, message transmission delay and the network overheads.更多还原