非线性加权观测融合滤波算法及其渐近最优性研究

项目名称： 非线性加权观测融合滤波算法及其渐近最优性研究

项目编号： No.61503127

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 郝钢

作者单位： 黑龙江大学

项目金额： 22万元

中文摘要： 非线性多传感器系统广泛存在于各个领域，传统处理非线性多传感器系统的方法是线性拟合、线性滤波器和线性融合算法相结合，可有效压缩数据量，减少计算负担，但该方法由于舍去了大量信息，因而存在较大误差，甚至导致滤波发散。本项目拟对具有加性噪声的非线性多传感器系统，通过函数逼近方法，提出一种具有普适性的加权观测融合算法。结合非线性状态估计算法（UKF、PF等），根据噪声统计分布，提出非线性加权观测融合UKF滤波算法以及PF滤波算法，并与集中式观测融合算法在精度和计算量方面进行比较分析，在理论上严格证明两种算法的渐近最优性。该算法将有效压缩多传感器系统冗余信息，提高系统实时性，并可根据实际精度要求调整融合精度。该项研究具有重要的理论意义，将为非线性多传感器信息融合提供一种有效途径，并在组合导航、GPS定位、目标跟踪、通信和海量数据信息处理等领域具有广泛潜在的应用价值。

中文关键词： 非线性系统；状态估计；加权观测融合；渐近最优

英文摘要： Nonlinear multisensor systems are ubiquitous in various fields. Traditional method to handle these nonlinear multisensor systems is the combination of linear approximation, linear filter, and linear fusion algorithm. This approach can compress data effectively, and reduce computational burden, but it will lead to a big error and even filtering divergence, due to round a lot of information. In proposed project, a universal nonlinear weighted measurement fusion algorithm is presented via function approximation for nonlinear multisensor system with additive noise. Then nonlinear weighted measurement fusion UKF and PF are presented based on nonlinear state estimation (UKF, PF, etc.) according to noise statistics. The two algorithms will be compared with centralized measurement fusion algorithm in accuracy and computation, and their asymptotically optimality will be proofed mathematically. The algorithms will compress redundant information of the multisensor systems effectively, and improve performance in real-time, and can be adjusted according to actual requirements of accuracy. This study has important theoretical significance, and provides an effective way for nonlinear multisensor information fusion, and has a widely use of potential applications in navigation, GPS, target tracking, communications and massive data processing.

英文关键词： nonlinear systems;state estimation;weighted measurement fusion;asymptotic optimality