分布式协同组网中多特性动态系统偏差的快速配准方法研究

In multi-sensor collaborative network, system model mismatch cause by many uncertain disturbances under complex environment. Furthermore, it is the reason why systematic biases show some characters including dynamic, uncertain, multiple sources, delay, nonlinear, and so on. The existing methods with known priori information are no longer suitable for registering above systematic biases. At the same time, it is a brand new research task that utilizing the rapidness of distributed construction with the optimization of concentration construction to rapidly and precisely track maneuvering target. Therefore, taking the collaborative detection network based on multi-sensor as the research object, and taking some theories including the identification modeling, filtering estimation and collaborative optimization as technical means, we carry out many researches into some rapid registration methods of feature-rich dynamic systematic biases. Firstly, we propose a dynamic modeling method for generalized systematic biases considering coexisted feature-rich, and this model is of benefit in both dynamic evolution and uncertain disturbance input. Secondly, system models in state space are projected to the equivalent system models in systematic biases space by space projection method. We analyze model errors before and after projection, and research the decoupling conditions of their dynamic correlation. Subsequently, we design an on-line and unbiased estimator of systematic biases under constrain of aforementioned decoupling conditions. Thirdly, we research approximate methods for optimal solution, and design the rapid estimation strategies with distributed based on layered feedback mechanism, these strategies consider both high precision and rapidity. Finally, some works of principle verification for new methods will be carried out based on simulation tests/data analysis.

多种来源的不确定扰动会引发多传感器协同组网系统模型失配，系统偏差呈现出动态、不确定、多源、延迟及非线性等多特性共存特点，系统先验信息已知的方法不再适用配准此类系统偏差；同时在系统偏差背景下，如何利用分布式结构的快速性优势并结合集中式结构的最优性特点，对机动目标进行快速、高精度跟踪是目前协同组网研究的新问题。本项目拟以多传感器协同探测网为研究对象，以建模辨识、滤波估计和协同估计为技术手段，开展多特性动态系统偏差的快速配准方法研究。首先，对多特性共存的广义系统偏差动态建模，使其兼具描述动态演化和不确定扰动输入的功能；其次，通过状态空间投影方法，建立等效系统偏差模型，进而对投影前后系统模型误差的动态相关性进行解耦，并在此基础上设计广义系统偏差在线无偏估计器；再次，研究网络调度最优解近似方法，设计基于分层反馈的分布式快速协同估计策略；最后，开展基于仿真测试/数据分析的方法原理性验证。

本课题在多传感器融合系统和多动态系统偏差标定的基础上，探索了多传感器系统偏差配准的相关理论和技术，包括系统偏差建模与估计、多传感器协同管理、网络优化和定位。主要研究内容和成果涉及以下4方面：.1.分布式无线传感器网络（WSN）系统偏差估计与配准：.针对WSN中无测距设备节点range-free最小二乘定位方法精度较低问题，引入节点距离估计配准项，构建节点定位新目标函数。并针对该函数非凸且不可导引起的定位算法精度和计算复杂度难以协调问题，利用矩阵变换重构和逐次凸逼近迭代求解方法，研究成果有望推广至水下传感器系统偏差配准及定位等应用领域。针对多传感器协同探测系统中系统偏差呈现出的随机性和突变性，以及估计方法的误差增加、估计鲁棒性较难保证等问题，提出基于贝叶斯估计框架的系统偏差估计等方法。.2.系统偏差估计相关的目标状态估计、跟踪与信息融合：.针对WSN中视频跟踪器中固定系数空时正则化约束项对目标外观变化不敏感问题，提出基于在线空间分析思想的滤波器训练方法-基于时间正则化模型的实时目标外观配准滤波器。针对分布式WSN背景下目标跟踪系统中时变过程噪声和随机异常量测噪声,以及传感器量测缺失对目标状态估计精度的影响，将变分贝叶斯与传统目标跟踪算法结合，提出基于变分贝叶斯的分布式融合目标跟踪算法。.3.系统偏差估计相关分布式WSN物理层安全：. 针对采用同步无线信息与电能传输技术的多用户无线传感器网络中存在高能耗和信息安全问题，提出用户辅助的非正交多址物理层安全传输方案和中继前向功率、接收机信干噪比和窃听者信干噪比约束下保密速率优化模型，并针对该模型非凸属性，提出求解算法。.4.系统偏差估计相关的分布式WSN路由设计：.针对簇头随机选择、工作节点冗余、簇头传输路径等影响传感器网络能耗的问题，提出了基于路径树的多跳路由算法。针对WSN中工作节点过早死亡、网络能耗大、节点能量分布不均匀的问题。在传统低能耗自适应聚类层次（LEACH）算法基础上，考虑被监控目标位置对网络性能的影响，提出改进LEACH目标定位约束等算法。