基于拉依达准则与线性拟合的改进型无迹卡尔曼滤波粗大误差补偿算法

Improved Unscented Kalman Filter Based on the Gross Error Compensation Algorithm with Pauta Criterion and Linear Fitting

摘要:

无迹卡尔曼滤波是卡尔曼滤波技术的重要组成部分,它有效地克服了扩展卡尔曼滤波的估计精度低、稳定性差等缺陷。然而无迹卡尔曼滤波未考虑粗大误差(如离群值、静差和漂移)的影响。目标跟踪经常受到不同种类粗大误差的影响,研究无迹卡尔曼滤波器对粗大误差的检测和补偿,对目标跟踪准确性的提高有重大意义。本文针对观测值中各种粗大误差影响目标跟踪精度的问题,采用拉依达准则对观测值进行检测。为了对误差进行补偿,本文提出了一种观测数据残差线性拟合的方法,使用拟合产生的预测残差补偿粗大误差,使补偿后的目标运动轨迹能够减小粗大误差的干扰。经过目标跟踪仿真实验和对比,本文提出的改进型无迹卡尔曼滤波算法能有效地减小粗大误差观测值对状态预测过程的影响,能实现对目标的准确跟踪,提高了滤波的稳定性和准确性。

Abstract:

Unscented Kalman filter (UKF) is an important part of Kalman filter. UKF can effectively overcome the defects of the extended Kalman filter such as low estimation accuracy and poor stability. However, UKF seldom considers the influence of the gross errors, such as outlier, bias and drift. The measurements of target tracking are often affected by different kinds of gross errors, so it is of great significance to study the detection and compensation of gross errors based UKF for improving the accuracy of target tracking. In this paper, in order to decrease the influence of gross errors in the measurements on the results of target tracking, the measurement information is checked by using the Pauta criterion to detect gross errors. In order to compensate the gross errors, the results of linear fitting of residual errors are used. Through target tracking simulation and comparison, the improved UKF can effectively decrease the influence of gross errors on the state prediction procedures, and can achieve accurate target tracking, improving the stability and accuracy of filtering.