Date of Completion
Bias estimation, space tracking, observability, statistical efficiency, CRLB, composite measurements, maximum likelihood.
Prof. Yaakov Bar-Shalom
Prof. Peter Willett
Prof. Krishna Pattipati
Field of Study
Doctor of Philosophy
Most of the literature pertaining to target tracking assumes that the sensor data are corrupted by measurement noises that are zero mean (i.e., unbiased) and with known variances (accuracies). However in real tracking systems, measurements from sensors exhibit, typically, biases. This thesis aims to solve several key problems in passive sensors residual bias estimation, when only targets of opportunity are available. For angle-only sensors, imperfect registration leads to systematic Line of Sight (LOS) angle measurement errors in azimuth and elevation. If uncorrected, registration errors can lead to large tracking errors and potentially to the formation of multiple tracks (ghosts) on the same target. The first step is to formulate a general bias model for synchronized optical sensors; then we use a Maximum Likelihood (ML) approach that leads to a nonlinear least-squares estimation problem for simultaneous estimation of the 3D Cartesian state of the target of opportunity and the angle measurement biases of the sensors. The evaluation of the Cramer-Rao Lower Bound (CRLB) on the covariance of the bias estimates, and the statistical tests on the results of simulations show that this method is statistically efficient.
Belfadel, Djedjiga, "Passive Sensor Bias Estimation Using Targets of Opportunity" (2015). Doctoral Dissertations. 827.