Advanced data fusion methods with application to the multitarget multisensor tracking problem

This dissertation presents a solution to the multitarget multisensor aircraft tracking problem, with a focus on the design and implementation of advanced methods for combining data from a distributed network of sensors. Practical solutions to sensor registration, data association and data filtering problems are provided. The advantage of using multiple sensors is demonstrated through a radar jamming example.; The tracking system designed in this dissertation uses a clustering approach to the data association problem combined with a 2D assignment algorithm. It is tested on elaborate multitarget tracking scenarios, using real aircraft trajectories and including false measurements. Methods for evaluating the performance of the multitarget multisensor system are developed. The subtractive clustering algorithm is shown to provide improved tracking performance over an equivalence relation clustering algorithm. Track initiation and track termination are part of system design, allowing the handling of unknown and changing number of targets. The system is shown to effectively track seven crossing aircraft trajectories of different durations.; Track maintenance is performed by centralized extended Kalman filters, designed to simultaneously solve a sensor registration problem involving sensor position and alignment errors. Decentralized filtering algorithms are developed and compared to the centralized solutions but are not implemented because of their data transmission requirements and suboptimality. Monte-Carlo simulation results show that the relative alignments and positions of 3D track sensors can be successfully estimated to a high degree of accuracy along with the track variables. An analytic estimation of achievable position tracking accuracy confirms the performance obtained in the Monte-Carlo runs and is used to develop an optimal sensor placement strategy.; The simulation results are complemented by a covariance analysis showing the influence of the error sources on the tracking accuracies. The major source of error is the measurement noise if relative sensor errors are estimated. When absolute sensor uncertainties are estimated, the registration errors become weakly observable and bias the aircraft position estimates.; The non linear coordinated turn aircraft maneuver model is implemented in the tracking filters and validated by comparison to real aircraft trajectories and other popular maneuver models.