No abstract

No abstract

No abstract

No abstract

A Cartesian coordinate linear regression filter is utilized for tracking maneuvering aircraft targets. Measurements of target position are made in a line-of-sight coordinate frame, but filtering is performed in Cartesian coordinates. Numerical results are given for optimizing the truncation time constant such that a good balance is obtained between the dynamic errors and the standard deviations. Lower bounds on the dynamic errors are established for the Cartesian coordinate linear regression filter and compared with a line-of-sight coordinate Kalman filter.Accurate tracking of maneuvering aircraft targets often requires the utilization ofdigital filters. Some ofthe better known filters include the linear regression filter [1]- [4] and the Kalman filter [5], [6]. Measurements of target position are made in a line-of-sight coordinate frame; however, filtering may be performed in either line-of-sight or Cartesian coordinates. A previous paper by the authors [7] has shown that when linear regression filtering is utilized for tracking nonmaneuvering targets, the filtering should be performed in Cartesian coordinates. This paper extends those results by considering filtering and prediction for maneuvering targets. As in the previous cases, a simple two-dimensional target geometry is investigated. Fig. 1 shows the target geometry. The target initiates a 3 g maneuver at time t = ti. The flight path for a nonmaneuvering target is also shown. The true range, velocity, type of maneuver, and heading of the target are unknown to the tracker. The numerical results are utilized to determine the optimum truncation time constant for a good balance between the dynamic errors and the standard deviations.Finally, the lower bounds on the dynamic errors are established for the Cartesian coordinate linear regression filter. These bounds are then compared with the lower bounds on the dynamic errors for a line-of-sight coordinate Kalman filter. All results are obtained for a data sampling rate of 30 samples per second. Linear Regression in Cartesian CoordinatesSmoothing the data in line-of-sight coordinates has been shown to be undesirable mainly because the linear model leads to large dynamic errors. However, in Cartesian coordinates, the constant speed target does obey a linear law and smoothing in this reference frame does not lead to dynamic errors. It is therefore recommended that the raw data be transformed first into inertial coordinates and then smoothed. The smoothed estimates of the Cartesian position are then transformed back into line-ofsight coordinates. The target position in inertial coordinates is given as x(t) = x = R cos q = xo + ,Yt y(t) -y = R sini = y + i't.The variance of the target position in inertial coordinates when the present range and angle data is used to calculate those coordinates is (TX2= [Cos2 t]