Parameter estimation and data association for multitarget tracking

This dissertation work deals with parameter estimation and data association for tracking multiple targets. ^ In parameter estimation part, first we consider the problem of finding an aircraft\u27s trajectory, speed and altitude in three dimensional (3-D) space using range and azimuth (bearing) measurements from a two dimensional (2-D) radar that does not measure elevation. A Probabilistic Data Association based Maximum Likelihood (ML-PDA) estimator is proposed for estimating the target motion parameters in the presence of false measurements and missed detections. The estimation accuracies are quantified using the Cramer-Rao lower bound (CRLB) for different sensor-target configurations. Of particular interest is the achievable accuracy for estimating the target altitude, which can be only inferred from the 2-D observations. ^ Then we consider another estimation problem of identifying the parameters of a proportional navigation guidance missile (pursuer) pursuing an airborne target (evader) using angle-only measurements from the latter. The purpose is for the evader to take appropriate countermeasures in a timely manner. The time-to-go estimate can also be obtained from the proposed parameter estimate. ^ In the data association part, first we present a new assignment-based algorithm for data association in tracking ground targets employing evasive move-stop-move maneuvers, using Ground Moving Target Indicator (GMTI) reports. We develop a novel “two-dummy” assignment approach that can distinguish the non-detection event due to PD \u3c 1 and the non-detection event due to a target\u27s stop. Using this new algorithm, we can obtain reductions in both RMS estimation errors as well as the total number of track breakages. ^ Finally, a data association (DA) method using assignment, combined with the probability hypothesis density (PHD) approach based on particle filtering implementation is proposed for multitarget tracking. The combined algorithm, consisting data association and PHD filter is illustrated using the result of the PHD filter to get the estimated number and locations of the targets for data association and using the result of data association for PHD peak extraction. A novel PHD representation defined in a resolution cell (PHDRC), is also presented. It is shown that the final proposed algorithm, DA-PHDRC, shows better performance than the PHDRC filter and the traditional MHT/assignment algorithms.