网络式海洋声学层析动力学状态—空间建模与处理

发布时间：2018-08-20 12:12

【摘要】：海洋内部存在着复杂的运动与变化。了解海水运动的变化规律对开发和利用海洋具有重要意义。这需要对海洋进行大面积、长时间的观测。动力环境(如温度场、流场的分布)作为海洋环境的重要组成部分,其有效测量技术一直受到关注。声波是水下遥测的重要手段，，利用海洋声学技术可以实现大范围、长时序、动态、实时的监测。由于海洋参数存在一定的动力学特征,采用序贯滤波思想跟踪参数的变化是一种自然的选择，而如何建立准确、自适应的参数演化模型成为问题的关键。因此，论文以二维流速场的序贯跟踪为例,在分布式水声网络架构下,探讨海洋环境测量的状态一空间动力学建模和信号处理。海洋声学层析(Ocean Acoustic Tomography,简称OAT)技术从精确测量的声传播时间或者其他声传播属性推断声场所穿过的海洋状态,一般采用分布于观测区域周边的若干发射／接收节点,适合于观测区域内部中小尺度过程。沿岸声学层析(Coastal Acoustic Tomography,简称CAT)作为OAT的一个特例，用于监测半封闭海域的声速或流速的空间分布。由于流速在一定尺度的空间与时间上存在相关性,论文提出了基于空时演化模型的CAT跟踪方法,将空间网格流速状态的演化方程建模为空时AR模型,且AR系数根据状态估计值自适应更新。研究结果表明,基于空时演化的CAT跟踪方法能够有效抑制测量的抖动误差。海洋环境监测面临着分辨力与工作距离、空时尺度与资源之间的矛盾,水声传感网络(Underwater Acoustic Sensor Network,简称UASN)为解决上述矛盾提供了新的思路.UASN由大量分布于观测区域内部的低功耗小型传感节点组成,通过节点密度获取分辨力,通过节点分布获取覆盖范围,通过节点之间的通信实现协同工作。UASN参数场估计可以看作是OAT技术在分布式网络框架下的一种实现,我们将应用于声层析问题的UASN称为分布式声学层析网络(Distributed Acoustic Tomography Network,简称DATN).论文提出一种基于三角形网格离散化的流速场空时跟踪方法,依据节点拓扑对区域进行子三角形划分,以子三角形的流速作为研究对象,根据距离对子三角形进行分类,距离相近的子三角形属于同一阶,AR系数相同,将子三角形的流速变化建模为空时AR过程。研究结果表明,所提出的空时跟踪方法可提高估计性能,并且在应对诸如链路失效、突发误差等DATN的实际问题时具备较强的宽容性。水下通信资源的限制使得基于DATN的流速场估计的数据融合过程颇具挑战性。分布式处理在相邻节点间进行信息交换,处理过程在每个节点内部进行，使估计过程不依赖于任何一个中心,具备较强的环境宽容性和可扩展性。结合基于DATN的流速场估计问题的特点,论文提出基于一致新息的分布式流速场估计方法。该方法从Kalman滤波的信息表示形式入手,对平均伪测量向量进行一致新息处理,局部测量值提供新息,节点之间的信息交换达成估计一致化。为了提高分布式估计的收敛速度,论文还提出一种基于路径选择的一致化加权矩阵构造方法,并给出了所提出的分布式估计方法的无偏性和收敛性证明。研究结果表明,所提出的分布式流速场估计方法能达到局部流场估计的一致,且通信量小,收敛速度快,适用于水下高延迟通信环境。论文的理论与方法通过仿真或实验进行了验证。论文的研究工作进一步丰富了OAT技术,通过利用流速场的空时相关性提高估计精度;发展了基于DATN的参数估计方法,为实现大范围、高分辨流场测量提供新的途径;同时创新发展了基于声学层析网络的分布式估计方法，为水声传感网络在海洋环境监测中的应用奠定理论基础。
[Abstract]:Understanding the law of sea water movement is of great significance to the development and utilization of the ocean, which requires large-scale and long-term observation of the ocean. Acoustic wave is an important means of underwater telemetry, which can be monitored in a wide range, long time series, dynamic and real-time by using ocean acoustic technology. Because of the dynamic characteristics of ocean parameters, it is a natural choice to track the changes of ocean parameters by using sequential filtering. How to establish an accurate and adaptive parameter evolution model becomes a problem. Therefore, the state-space dynamic modeling and signal processing of ocean environment measurement are discussed in this paper. Ocean Acoustic Tomography (OAT) technique is used to deduce the acoustic propagation time or other acoustic propagation properties from precise measurements. Coastal Acoustic Tomography (CAT), as a special case of OAT, is used to monitor the spatial distribution of sound velocity or current velocity in semi-enclosed sea area. In this paper, a method of CAT tracking based on space-time evolution model is proposed. The evolution equation of velocity state in space grid is modeled as space-time AR model, and the AR coefficient is updated adaptively according to the state estimation. Underwater Acoustic Sensor Network (UASN) provides a new way to solve the above contradiction. UASN consists of a large number of small sensor nodes with low power consumption distributed in the observation area. UASN parameter field estimation can be seen as an implementation of OAT technology in a distributed network framework. We will apply UASN to acoustic tomography problems called Distributed Acoustic Network (DAT). A space-time tracking method of velocity field based on triangular mesh discretization is proposed in this paper. The region is divided into sub-triangles according to node topology. The sub-triangles are classified according to the velocity of sub-triangles. The sub-triangles with similar distances belong to the same order and AR coefficients are phase. Similarly, the velocity variation of the sub-triangle is modeled as a space-time AR process. The results show that the proposed space-time tracking method can improve the estimation performance and is more tolerant in dealing with practical DATN problems such as link failure and burst error. The combination process is challenging. Distributed processing exchanges information between adjacent nodes and processes within each node, so that the estimation process does not depend on any one center and has strong environmental tolerance and scalability. In order to improve the convergence speed of distributed estimation, this paper also proposes a uniform method based on path selection. The results show that the proposed distributed velocity field estimation method can reach the agreement of the local flow field estimation, and has the advantages of small traffic and fast convergence speed. It is suitable for the underwater high-delay communication environment. The research work of this paper further enriches the OAT technology and improves the estimation accuracy by utilizing the spatial-temporal correlation of velocity field; develops the parameter estimation method based on DATN, which provides a new way to realize large-scale and high-resolution flow measurement; and innovates and develops the distributed estimation based on acoustic tomography network. The method lays a theoretical foundation for the application of underwater acoustic sensor network in marine environmental monitoring.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：P733.2

【相似文献】