变结构多模型估计单Kalman滤波跟踪机动目标算法  

A Single Kalman Filtering Algorithm with Variable Structure Multiple-Model Estimation for Maneuvering Target Tracking

作　　者：刘士建[1] 郭立[1] 李士民[1] 

机构地区：[1]中国科学技术大学电子科学与技术系,安徽合肥230026

出　　处：《电路与系统学报》2003年第5期28-31,共4页Journal of Circuits and Systems

基　　金：安徽省自然科学基金资助项目(00043404)

摘　　要：在机动目标跟踪中,如果我们估计出目标的机动量,就可以使用单Kalman滤波来跟踪目标且跟踪精度和IMM滤波方法接近。文献[1]中提出用N个离散加速度iU覆盖目标机动量,然后用它们加权的方法来得到机动量的估值。这些离散加速度量iU的选择是个难点。本文提出使用变结构多模型方法来估计目标的机动量,该方法可以实时改变参与估计机动量的模型数目和参数,排除一些不必要的模型来减少模型数目和竞争。仿真实验结果表明,在大大缩短计算时间的前提下,本算法依然能够获得与原算法一样的跟踪精度。For maneuvering target tracking, after estimating maneuver of the target, single Kalman filtering algorithm can be used to obtain tracking accuracy similar to Interacting Multi-Model (IMM) algorithm. Some researchers chose to use N discrete acceleration values Ui to cover the possible range of maneuver and estimate the maneuver by weighting[1]. In such process, It is difficult to choose Ui properly. In this paper, We propose to use variable structure multiple-model to solve this problem. By this approach, the number of models and parameters can be adjusted in real-time. Unnecessary models can be excluded from the action of estimating as to reduce excessive competition. Simulation results show that the time for estimation can be shorten, while the accuracy can be maintained as good as the original algorithm..

关 键 词：卡尔曼 滤波 变结构 多模型 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]