摘要
Kalman滤波中不可避免地存在模型误差,模型误差的存在会引起滤波结果失真,严重时还会导致滤波发散。在对Kalman滤波模型进行检验时,有三种常用的残差——观测残差,预测残差和状态预报值残差。以往不同学者选用不同的残差对模型进行检验。当模型含有误差时,分别推导这三种残差变化后的表达式并对它们检验模型误差的能力进行比较,结果表明预测残差在模型检验中反应最灵敏,并且计算量最小。最后给出仿真算例,结果也证实预测残差能够更有效、更快捷地探测模型误差。
It is unavoidable that model errors exist in the Kalman filter in practice. Existence of model errors can not only induce the distortion of filtering result, but also result in filtering diffusion sometimes. When it goes to judge whether there are errors in Kalman filter model or not, three residuals --measure- ment residual, predicted residual, predicted state residual---- are in common use. Different residuals are chosen to test the model by different people in the past. When errors exist in the model, changes of three residuals are respectively deduced and the ability to test model errors are compared only to find that predicted residual is the most efficient one to test the model. Finally, a simulative experiment is given to show that predicted residual can detect model errors more effectively.
出处
《黑龙江工程学院学报》
CAS
2008年第2期22-25,共4页
Journal of Heilongjiang Institute of Technology
基金
国家自然科学基金资助项目(40474007)
国家杰出青年科学基金资助项目(40125013)
