非线性系统可观测度理论及其应用研究

发布时间：2018-09-17 10:11

【摘要】：空中任务传感器网是指通过多传感器对跟踪目标进行协同定位与跟踪的传感器网络。信息融合和卡尔曼滤波研究的兴起和发展使得空中合作传感器网络的众多任务功能(如系统目标跟踪定位,决策和评估等)成为可能。传感器网络管理是空中任务网络的主要任务之一,也很早就出现在信息融合的研究范畴中。可观测性研究的是能否通过观测对状态值进行估计。可观测度是信息融合领域研究由输出或观测获得估计状态的准确程度。将可观测度运用到空中传感器网络分配中,可针对不同的状态分量进行传感器资源的分配管理,使得传感器网络能够得到优化。现在已有很多可观测度的分析方法,但主要都针对线性系统。针对非线性系统可观测度分析仅有基于李导数的方法,计算量大。目前并没有完善的基于非线性可观测度的空中传感器分配算法。卡尔曼滤波的估计误差协方差被用于设计传感器分配管理的代价函数,但基于估计误差协方差的传感器分配方法,在考虑非线性问题时,需要在得到融合滤波结果后才能对传感器分配进行调整,不能在滤波前有一个初步的判断。针对上述问题我们做了以下工作:(1)结合UIF和线性SVD可观测度分析方法,对非线性系统可观测度分析方法进行了研究,并与线性系统SVD可观测度分析方法进行了对比。(2)研究了不依赖滤波结果的基于SVD分解的非线性可观测分析方法。通过加入尺度调节因子,使其具有了尺度变换不变性(Scale Transform Invariant)。(3)研究了基于非线性系统可观测度的传感器网络分配方法。该传感器网络分配方法不依赖于滤波结果,比基于估计误差协方差的传感器网络分配方法有更广泛的适用性。
[Abstract]:Aerial mission sensor network is a kind of sensor network which can locate and track the target by multi-sensor. The rise and development of information fusion and Kalman filtering make it possible for many mission functions (such as target tracking, decision making and evaluation) of air cooperative sensor networks. Sensor network management is one of the main tasks of air mission network, and it has appeared in the field of information fusion for a long time. Observability studies whether state values can be estimated by observation. Observability measure is the degree of accuracy of estimating state by output or observation in the field of information fusion. The observability measure is applied to the allocation of sensor network in the air, and the sensor resources can be allocated and managed according to different state components, so that the sensor network can be optimized. There are many methods for analyzing observable measures, but they are mainly for linear systems. There is only a method based on the lie derivative for the analysis of observable measure of nonlinear systems. At present, there is no perfect aerial sensor assignment algorithm based on nonlinear observable measure. The estimation error covariance of Kalman filter is used to design the cost function of sensor allocation management. The sensor allocation can only be adjusted after the fusion filtering results are obtained, and there can be no preliminary judgment before filtering. In view of the above problems, we have done the following work: (1) combining UIF and linear SVD observable measure analysis method, the observable measure analysis method for nonlinear systems is studied. It is compared with the SVD observability analysis method for linear systems. (2) A nonlinear observable analysis method based on SVD decomposition is studied, which does not depend on the filtering results. The sensor network assignment method based on the observability of nonlinear systems is studied by adding a scale adjustment factor to make it scale-invariant (Scale Transform Invariant). (_ 3). The proposed assignment method is independent of the filtering results and is more widely applicable than the sensor network assignment method based on estimation error covariance.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212

【相似文献】