摘要
为降低随机选取Kalman滤波器中系统噪声方差和测量噪声方差所带来的估计误差,提出利用遗传算法对Kalman滤波器中系统噪声方差和测量噪声方差进行优化,降低其对初始值的敏感度。以线性二自由度车辆仿真模型算例分析不同方差组合对Kalman滤波器的影响,确定出两个方差值对滤波器效能的影响程度,将影响程度的大小作为遗传算法的适应度函数,在此基础上利用遗传算法对系统噪声方差和测量噪声方差进行了优化,通过优化确定出最佳的方差组合,再将优化后的方差组合输入Kalman滤波器,从而提高Kalman模型的广泛适用性。利用优化后的Kalman滤波器对车辆仿真数据进行滤波处理,对比分析结果表明,优化后的Kalman模型对方差的随机性输入敏感度明显降低,能够准确逼近仿真模型的真实值。
To reduce the estimation error caused by random selection of the system noise variance and measurement noise variance in Kalman filter, optimizing the system noise covariance and measurement noise variance by using the genetic algorithm is proposed so as to reduce its sensitivity to the initial value. Firstly, the linear two degrees freedom vehicle simulation model is taken as example to analyse the different variance combinations influence on Kalman filter. In addition, the influence on the filter efficien- cy is determined and the influence is as the fitness function of genetic algorithm. On the basis of that, genetic algorithm is adop- ted to optimize system noise and measurement noise variance. Through the optimization, the best variance combination is achieved and the optimized combination of variance is as the input of Kalman filter, which improves the applicability of Kalman. Finally, the optimized Kalman filter is adopted to filter the vehicle simulation data. By comparative analysis, the optimized Kal- man can significantly reduce the sensitivity of the random input of variance and the simulation model can accuratly approximate the real value.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第2期641-645,共5页
Computer Engineering and Design
基金
教育部长江学者与创新团队支持计划基金项目(IRT1286)
国家自然科学基金项目(51178053)
中央高校基本科研业务费专项基金项目(2013G1221024)
关键词
卡尔曼滤波
系统噪声
测量噪声
遗传算法
优化
Kalman filtering
minimum variance estimation
noise variance
genetic algorithm
optimization
