| The problem of three-dimensional tracking of a vehicle maneuvering in the David Taylor Model Basin is addressed.; Two filters are developed: (1) Using the pingers independently and (2) Using dual pingers in the filter.; In both cases a linear model is used for the vehicle dynamics. Measurements are fundamentally non-linear, and additional non-linearities are introduced when both pingers are filtered together. Major issues involve proper rejection of many bad measurements caused by multi-path and attenuation, and being able to maintain proper linearization of the measurements. A technique to improve robustness by sorting measurement residuals according to their magnitudes is introduced. The philosophy is that the smaller measurement residuals are likely to be related to more accurate measurements.; First, a simulation is generated to ensure that the algorithms are fundamentally correct. Then, real data are applied to the algorithms for both filtering cases. The quality of results is evaluated by investigating the recovered path of the vehicle, calculating measurement residuals, comparing estimates to known quantities (such as range and range rate between pingers), and other techniques which will provide insight into accuracy, consistency and reliability.; Results show that EKF tracks the vehicle with high accuracy, and can sense and recover from potentially unstable situations caused by severe data conditions. In the case of using pingers independently in the algorithm, the filter is able to screen out the bad measurements and produce accurate estimates. However, when the measurements are all bad (situation when the pinger broke the water surface), the filter would reject all measurements, resulting in losing track of the vehicle. However, the algorithm demonstrated that it could appropriately restart when valid measurements were available again. We were able to demonstrate that RMS accuracy in horizontal position is lower than .04 ft.; Furthermore, results from using dual pingers in the algorithm show that the coupled pinger filter can improve the accuracy of the tracking especially when the system encounters many bad measurements due to one of the pingers breaking the surface of the water. Results show more smoothness of the recovered path indicating that the vehicle was tracking properly. In addition, lower RMS magnitudes of the measurement residuals indicated the excellent performance of the coupled pinger algorithm. (Abstract shortened by UMI.)... |
