Comparison of fix computation and filtering for autonomous acoustic navigation

Oceanographic data acquired by an autonomous underwater vehicle (AUV) must he correlated with accurate position information in order to he of value to scientists. Accurate navigation is therefore an essential requirement which can he fulfilled by use of acoustic long-baseline navigation systems and dead-reckoning sensors. Two approaches are presented in this paper. The fix computation approach consists of dead reckoning between fixes computed from a set of acoustic travel times. The filtering approach consists of correcting the vehicle dead-reckoned motion by taking into account the influence of the measured travel times. Fix computation is widely used for the positioning of manned submersibles and there seems to be some reluctance to switch to a filler-based approach, mainly because of the fear for divergence. After a detailed description of these two approaches, their advantages and drawbacks are compared by applying both algorithms to real data sets collected by the Odyssey II AUV developed at Massachusetts Institute of Technology Sea Grant. The filter presented in this paper is not subject to divergence. In the worst case, it would reject all the acoustic travel times measurements and proceed by dead reckoning only, so that re-initialization would be needed. This situation would also happen in the fix computation approach if the algorithm locked on an erroneous fix.