【作者】 张佳琳；

【导师】 赵琳；

【作者基本信息】 哈尔滨工程大学 ， 导航、制导与控制， 2013， 博士

【摘要】 为使分布式导航系统在GPS故障或信号中断时能够继续完成导航任务并完成作战计划，有必要提高分布式导航系统的容灾容错性能，即研究分布式容错导航仿真平台，导航信息存储数据结构、导航节点间通信协议、分布式导航系统的容错算法四个方面的容错关键技术。由于容错机制的本质是在系统出现故障时各功能单元继续执行所要求功能的能力，在分布式导航系统中以通过GPS得到的定位数据为基础进行导航，因此对系统中GPS定位数据进行有效管理和合理利用，能够完成在GPS信号中断状况下快速恢复分布式导航系统各节点的通信功能和导航功能，论文都是围绕如何运用分布式导航系统中各节点既有的GPS“残迹”交互信息和寻路导航展开论述。本文主要从分布式导航系统容错机制中容错系统仿真平台和容错算法两个方面进行深入研究和改进，以满足分布式导航系统的容错要求。论文主要研究内容如下：1.运用兵棋推演系统进行分布式导航系统容错算法仿真。由于传统的兵棋推演系统采用人工驱动方式，即对象间的信息交互完全由人工交互实现，需要完成的工作量巨大，导致系统中的若干干扰因素都被忽略，在对容错导航算法进行仿真验证时将严重失真，因此将消息驱动机制引入兵棋推演系统的系统结构与功能模块中，解决兵棋系统驱动的自动化问题和对象间的通信问题，通过设计人机接口模块、消息驱动模块、棋盘和棋子的结构，在单位棋子节点中引入导航数据结构和属性，可以加载本文提出的基于GPS“残迹”的容错导航算法；同时，添加大量的仿真测试接口，以便测试不同的导航算法。经与传统兵棋系统仿真测试进行比较，完成相似任务的时耗降低50%左右，人工干预次数仅是传统系统的25%，CPU负载率降低了15%。2.研究移动自组织网络中的对等节点容错导航算法（GVDN）。在故障时系统中各节点将脱离GPS信号，因无法继续完成导航，造成迷航等问题，而移动自组织网络能够在不依托固定通讯设施的基础上，支持节点动态的通信，并能较好控制流量，因此提出一种基于移动自组织网络的对等节点容错导航算法，以解决分布式导航系统中的容错问题。该算法以故障前的GPS“残迹”数据为基础，设计“残迹”存储的数据结构、系容错通信协议、电子地图接口、人机接口和对等节点交互协议和算法的流程。在“残迹”采集与处理模块应用朴素贝叶斯分类法对已获得“残迹”和关键点进行分类、在道路预测与评估模块中采用智能曲线识别算法对道路或可行路径的相似性进行判断。经兵棋推演系统仿真试验，协议具有较强的抗死锁能力，平均通信信道占用率均未达到峰值，平均内存占用量控制在10M以内，容错过程中时耗和CPU占用率均有所降低。3.研究群组分布式容错导航算法（GFTN）。由于对等节点容错导航算法在节点数量多的大规模分布式导航系统中，虽然算法精度较高，但容易耗尽系统资源，因此，采用群体智能理论来解决系统的全局优化问题，以适应广域分布的群组节点容错导航。针对算法的体系结构、功能模块与数据结构、交互协议、数据包结构等方面进行了设计，本文提出了按地理位置进行残迹存储，改进滑动窗口进行数据转发和利用的方案，解决了单个领航节点能力与群内节点分布广、通信量大的问题；提出了分组蚁群算法，使同组节点尽可能选择相同或相近的同行道路或区域，以解决移动节点数量多与可通行路径少的矛盾；提出了群组内共享信息的方案，解决了导航过程中重复计算的问题。4.在GVDN和GFTN算法的验证和仿真试验中，采用反应式Agent体系结构进行仿真，避免了复杂的逻辑推理验证，规定了Agent模型的属性、处理和行动规则，经过测试，Agent模型仿真粒度小，可用性高，可模型大规模虚拟场景，容纳更多的导航对象，在系统性能方面具有稳定和低耗的优势。更多还原

【Abstract】 It is necessary to improve the disaster recovery and fault tolerance of the distributednavigation system in order to continue to complete the navigation task and the operations planof the distributed navigation system when the CPS failure or signal interruption happens. Thatis to study the key aspects of fault tolerant which has four aspects: the simulation platform ofdistributed fault tolerant navigation, the data structure of navigation information storage, thecommunication protocol of navigation node, and the fault-tolerant algorithms of distributednavigation system. Because the nature of the fault tolerance mechanisms is the ability tocontinue to perform the required function when there is failure in system and the ability tonavigate based on positioning data acquired by GPS in distributed navigation system,effective management and rational utilization of the GPS positioning data can immediatelyrestore the communication and navigation function of each node in the distributed navigationsystem when the GPS signal is interrupted.This article is center on how to use the existing GPS ‘vestiges’ in the distributednavigation system to communicate information and path finding navigation. The essay mainlygives a deep research and improvement of the simulation platform of distributed fault tolerantnavigation and the fault-tolerant algorithms of the distributed navigation system, in order tomeet the requirements of the fault-tolerant distributed navigation system. The main researchcontent of the paper is as follows:1.Using the war-game system to simulate the distributed navigation system fault-tolerantalgorithm. As the traditional war-game system used the artificial driving mode, that is thecommunication interaction of the objects is all realized by the artificial driving mode. Thetraditional war-game system needs to complete a large amount of work which leads to theignorance of the interference factors in the system and a serious distortion when thesimulation verification of the fault-tolerant navigation algorithm is done. So we introduce thismessage-driven mechanism into the war-game system structure and function modules whichcan solve the problem of driving automated and communication between the objects. Also thepaper designs the human-machine interface module, message-driven module and the structureof board and pieces. The unit pawn nodes are introduced in the structure and properties of thenavigation data to adapt to the fault-tolerant algorithm based on the GPS ‘remnants’; At the same time large amount of the simulation test interfaces are added in order to testdifferent navigation algorithm. Using the war-game system simulation tests the time of taskcomplement reduced about50%, artificial intervention times is25%compared with thetraditional one and the CPU load is reduced by15%.2.Study the node fault-tolerant navigation algorithm in the mobile ad network(GVDN).Eachnode in the system will separate from GPS signal in the event of fault. Trek and otherproblems will appear without the continuing the navigation. And also the mobile and hocnetwork can support the communication of node dynamic without depending on the fixedcommunication facilities. It also can control the flow. So present a fault-tolerant navigationalgorithm based mobile ad hoc networks, peer-to-peer node, in order to solve the problem offault-tolerant distributed navigation system. The algorithm based on the GPS ‘vestiges’ databefore the fault this algorithm designs the data structure of ‘remnants’ storing, systemfault-tolerant communication protocol, the electric map interface, man-machine interfaces,and peer node interaction protocols and give the algorithm procrsses. In’vestiges’ collectionand the processing section this algorithm uses na ve Bayesian classification to classify the‘vestiges’ and the key point and in the road predict ang assess section the algorithm uses theintelligent curve recognition algorithm to judge the similarity of the road or the proper road.Though the war game system simulation test the agreement has a strong anti-deadlock ability,the average communication channel occupancy rate doesn’t reach the peak and the averageamount of memory usage is controlled within10M.And also consumption and CPUutilization rate are decreased.3.Study the group distributed fault-tolerant navigation algorithm(GFTN). Because thenode fault-tolerant navigation algorithm is in the large-scale distributed navigation systemwhich has a large number of nodes, the swarm intelligence theory is used to sole the globaloptimization problems of the system in order to adapt group node fault-tolerant navigation ofdistribution group, though the algorithm is higher precision it is easy to run out of systemresources. Based on the design for the architecture, the functional modules and the datastructures, the interaction protocols and the packet structure of the algorithm, this article putsforward the storage of the remnants by the location, improves the sliding window to forwardand use the datas, using the datas and solving the problem of single pilot node capacity,widely distributed notes in the group and the large communication capacity. It puts forwardthe packet ant colony algorithm which makes the nodes of the same set choose the same orsimilar peer-road or regional as much as possible so that the contradiction of more mobilenodes and fewer passable path. It also comes up with the program of sharing information within the group and solves the problems of double counting in the navigation process.4.In the simulation of the simulation of the GVDN and GFTN algorithm, the reactiveAgent system structure is used to simulate, which avoids the complex logic verification andsets the attributes of agent model, processing and action rules. The test shows that the agentmodel simulation efforts are small, the availability is high, a large scale virtual scenes can besimulated, more navigation objects can be contained more and in the system performance ithas the advantage of stability and low consumption.更多还原