摘要
论文分析了当目标距离或雷达极角量测误差超过一定限度时 ,扩展卡尔曼滤波器跟踪精度迅速下降的原因 ,并在精确计算线性化误差及实际量测误差的基础上提出一种补偿线性化误差的跟踪滤波算法 (PTLKF) .计算机仿真结果表明 ,论文提出的算法有效地减小了线性化误差的影响 。
In many areas, such as active sonar and radar syste ms, a very important topic is how to estimate as accurately as possible the stat e of a dynamic system from its noisy measurements. In solving problems of thi s kind, such as the problem of tracking moving objects, since the measurements m ade in radar centered polar coordinate are expressed as nonlinear equations in the Cartesian coordinate which is well suited to describing the target dynamics, it is serious to eliminate the nonnegligible nonlinear effects. Although the well known extended Kalman filter (EKF) and converted Kalman filter (CMKF) have been widely used to solve the problem, it has been found that in case the range of the target or the measurement errors of bearing is large enough, the approx imation errors resulting from linearization will affect the performance of the EKF filter considerably. Lerro [2] has developed a debiased consistent converted approach (CMKF D ), which improves the performance of the CMKF filter. But the drawback of that i ts formulation is rather complex, and applicable only in two dimensional cases. In this paper a more precisely filtering algorithm (PTLKF) comparing with the EK F is presented on the basis of correctly evaluating approximation and measure ment errors. To calculate approximation errors nonlinear measurement equations a re divided into a linear part and a nonlinear part. It can be seen t hat the nonlinear part is just the approximation error resulting from linearizat ion, known as linearization error. The expression of the linearization error is given, and from the expression it can be seen that the linearization error is in direct ratio to the range of the target and the measurement errors of bearin g. To compensate the influence of the lineartiztion error, the lineariza tion error is taken as a part of the measurement error. The resulting filter str ucture is similar to the structure of the EKF, just a little more complicated in calculat ing the statistics of measurement error. Computer simulation results show that t he proposed algorithm reduces the influence of the approximation errors effica ciously and offers superior performance in comparison to the EKF filtering algo rithm.
