无序信息的网络融合及在舰船组合导航中的应用

【作者】 葛泉波；

【导师】 文成林； 汤天浩；

【作者基本信息】 上海海事大学 ， 电力电子与电力传动， 2008， 博士

【摘要】 无线传感器网络(Wireless Sensor Networks,WSN)环境下的融合系统在进行信息搜集和综合处理时,将不可避免地受限于诸多与网络系统有关的约束条件(简称约束),如有限的通信带宽和节点能量、存储和计算约束、动态网络拓扑以及路由协议等,从而导致传统信息融合理论和技术无法有效地满足日益增长的实际工程需求,并使得基于传感器网络的信息融合(简称网络融合)研究面临着诸多新的难题和挑战。各种网络约束的共同存在使得信息经传感器网络传输后,其定常性、因果性和序惯性易受破坏,若将丢包视为无限的延迟,则上述约束条件对融合系统的影响将集中体现于局部传感器信息传输具有随机延迟。而延迟的随机性必然导致有序采样的局部信息到达融合中心时呈现无序的现象,即后发的信息可能先到,先发的信息可能后到,简称为无序信息。从而,传统基于信息有序到达的融合方法将无法直接有效地应用于无序信息系统,并使得无序信息的融合算法设计已成为网络融合研究的重要内容之一。此外,Y.Bar-Shalom等提倡的无序量测(“Out-of-Sequence”Measurements,OOSM)更新方法只能解决部分无序信息系统的估计融合问题。因此,由于OOSM研究刚处于起步阶段,加之诸多自身无法克服的不足,使得大量关键性的理论和技术问题亟待进一步研究和解决。近年来,海上舰船航行安全问题受到了人们的高度重视,其主要内容之一就是对舰船航行进行动态监控并为舰船的自主避碰和其它海事险情的发生提供早期预警。实现上述功能的首要前提就是舰船精确和可靠的导航定位。传统舰船组合导航定位方法大都研究如何最优地融合来自于舰船各内部导航子系统提供的信息,并没有考虑信息的异构性和不确定性、系统的网络特性以及对GPS过度的依赖性,这样就大大降低了传统方法在实际应用中的稳定性和可靠性。实际上,海上舰船网络系统是一个典型的无线Ad hoc传感器网络,网络中的各个节点以多跳和广播的方式在网内进行信息通信和交换,从而使得各舰船节点均能接收来自他船的信息。因此,舰船节点可通过有效综合其他节点提供的信息来改善自身舰船组合导航定位方法的精度和可靠性。由于海上各舰船的导航系统往往异步工作且各船之间的信息交流均是通过无线网络进行,加之船台以短信方式转发信息和船载AIS设备在规定时间内发送不完数据均可能将造成整个数据队发送列的推迟等,都有可能导致网络节点间的信息传送具有不确定的延迟。那么,实现上述功能的关键就是在海上网络环境中目标舰船如何有效利用其他节点的信息来更新和改善自身组合导航定位估计的性能。因此,开展无序信息的网络融合及在舰船组合导航中的应用研究具有重要的理论意义和广泛的应用前景。鉴于此,本文重点开展了无序信息的网络融合算法设计以及网络环境下舰船组合导航定位的方法研究,其主要研究贡献和创新点如下:1)建立带有任意随机延迟的多传感器系统通用OOSM融合框架。为了客服现有无序量测方法研究存在的局限,利用“伪量测”技术和传统噪声相关Kalman滤波技术,系统地开展单传感器、多传感器同步系统以及一大类异步采样系统具有任意随机延迟的OOSM估计融合算法设计,最终建立一个通用的无序量测估计融合算法框架。2)提出基于分布式融合框架的无序估计(“Out-of-Sequence”Estimates,OOSE)方法来解决无序信息的网络融合问题。鉴于无序量测方法的集中式估计特性所引发的诸多不足,建立分布式融合框架下的OOSE方法来解决OOSM方法存在的问题,并针对任意随机延迟的多传感器系统建立一个通用OOSE融合方案。同时,针对一类特殊异步采样的树形网络无序信息系统,开展OOSM和OOSE联合方法的研究,并建立了相应的混合无序信息融合方案。3)结合网络信息融合技术建立舰船相对导航定位方法。考虑舰船网络中信息丰富特性以及实际舰船组合导航子系统信息的不确性,提出利用相邻舰船的信息来改善目标舰船自身组合导航定位估计的精度、稳定性和可靠性。同时,将前述建立的OOSM和OOSE方法应用到带有延迟的舰船网络节点的相对定位中。更多还原

【Abstract】 Multisensor fusion system based on Wireless Sensor Networks (WSN) will suffer many network constraints when it is collecting and processing the information from local sensors, such as limited communication bandwidth and energy, limited storage and computation of network nodes, dynamic network topology and routing protocol and so on. Accordingly, the traditional information fusion theory and technologies can not be satisfied with the increasing practical engineering requirement. Synchronously, multisensor information fusion based on WSN, for short network fusion, must be faced with many new difficult problems and challenges.The corporate effect of the network constraints is that the properties of cause and effect, constant and sequence are easily destroyed. If the package drop is considered to unlimited delay, the effect to the fusion system from the network constraints jointly results in the random delay of local sensor information. And the randomicity of transmission delay must make that the local information sampled orderly arrive in the fusion center out-of-sequence, namely the latter information may arrive earlier and the earlier information arrive latter, and it is called to Out-Of-Sequence Information (OOSI). Consequently, the final result is that the traditional information fusion based on ordered information can not be used effectively for the OOSI case and the OOSI estimate fusion algorithm design has been one of the most important researches in the network fusion. At present, the "Out-of- Sequence" Measurements update method named as OOSM, which is advocated by Y. Bar-Shalom etc., only solved the partial OOSI estimate fusion cases. Thereby, because it is primary phase for the research of OOSM method and possesses many insuperable shortcomings induced by its centralized fusion character, it makes that plentiful key academic and technical issues should be further researched.Currently, the research for the shipping safety is paid more attention and one of the main contents is to monitor the shipping and provide the early warning for active collision avoidance and other marine dangers. Then the principal key issue to realize above functions is the more accurate and reliable navigation and positioning for ships. The traditional ship integrated navigation positioning methods are mostly to fuse optimally the information from the navigation subsystems on the ships. Accordingly, many problems, for example the information heterogeneous and uncertainty characteristics, the system networks property and excessive dependency for GPS, make the implementation and reliability of the traditional integrated navigation positioning methods in the practical systems reduced. In fact, marine shipping network system is a scale wireless Ad hoc sensor network and each network node communicates each other by multihop and broadcasting matter and the result is that each ship in the network can receive the other ship’s information. So, the accuracy and reliability of the traditional integrated navigation positioning methods for ship node can be improved by effectively fusing the positioning information from other ships. Because the navigation system of each marine ship works asynchronously and the communication between two ships is often finished by the wireless network, additionally the slipway’s information transmission by short message and the delayed total data alignment for ship AIS equipment can not finish some data transmission during the stated time, they can result in the uncertain delay for the information communication between random two ships. Therefore, the key to realize above function is to effectively use other ship’s information to improve the navigation positioning estimate of target ship node in the marine shipping network.So, it has important theoretic significance and extensive application foreground to develop the research of network fusion based on OOSI and its application in the ship integrated navigation. In the view of this, the dissertation lays stress on the research for the design of OOSI network fusion algorithms and the ship integrated navigation method under the marine shipping network, and its main contributions and innovations are listed as follows:1) Establish the relatively perfect OOSM estimate fusion frame for multisensor asynchronous sampling system. Aiming at the limitation for the current OOSM update achievements, by use of the pseudo-measurement and the traditional Kalman filtering based on noise correlation, it roundly develops the design of out-of-sequence measurements estimate fusion algorithms for the single sensor system, multisensor synchronous system and a kind of asynchronous sampling system with random delay. Sequentially, one universal frame of OOSM fusion algorithms is established.2) The "Out-Of-Sequence" Estimates (OOSE) method is proposed to treat with OOSI estimate fusion. In view of many insurmountable shortcomings existed in the OOSM produced by its centralized fusion property, the OOSE method is presented to solve these OOSM shortcomings under the distributed fusion frame. Ultimately, a unified OOSE fusion frame for general asynchronous sampling system with random delay is finished. Synchronously, the combination of OOSM and OOSE is discussed deeply for a kind of special asynchronous sampling system with tree form. This research accords with the idea that the traditional centralized and distributed fusion methods can be jointly used.3) The novel ship relative navigation positioning method is presented by introducing the network fusion theory. For the characters of information abundance in the marine shipping network and the uncertainty of ship integrated navigation positioning subsystem, we presents to use other ship’s information to improve the accuracy, stability and reliably of the active integrated navigation positioning estimate of target ship. Synchronously, the established OOSM and OOSE methods are both applied to deal with the relative navigation positioning of ship network node with transmission delay.更多还原