USBL pose estimation using multiple responders
Leif Christensen, Martin Fritsche, Jan Albiez, Frank Kirchner · OCEANS'10 IEEE SYDNEY · 2010
Ultra Short Baseline Systems (USBLs) to track a submerged target normally consist of the transceiver unit mounted to a vessel and usually only one responder per underwater vehicle to track. Since the system measures vertical and horizontal angles of the returning signal in conjunction with the range, the system is only capable of providing a 3D position and not the entire pose (position and orientation) of the submerged vehicle or its dynamic state, which sometimes is needed. In this paper we propose the use of multiple USBL responders mounted onto a sub sea crawler to improve the accuracy and robustness of the tracked vehicle's position and velocity and, in the case of restricted or non-available orientation sensor data from the vehicle itself (e.g. no expensive inertial navigation system, no real-time communication with the vehicle), to make an external sub sea orientation estimation possible at all. In the paper, the decisions that lead to the chosen probabilistic filter to accomplish the state estimation task mentioned above and the concrete filter adaption is explained in detail with special focus on state composition, the corresponding measurement model with three responders and outlier rejection using the Mahalanobis Distance. The tracking system application consisting of the Bayesian filter and the data acquisition interfaces and the actual data flow is presented as well as the design of the related simulation software. Finally, the performance of the filter with varying parameters and its robustness against outliers and signal loss is shown.