Sensor fusion for AUV localization

A system for the effective integration of multisensory information into the piloting operation of underwater vehicles (AUVs) and for its interpretation is presented. The operators are assisted in the localization and piloting tasks by synthetic visualization and data fusion tools. It is possible to estimate an optimal location by fusing data from log and gyroscope, or a nonoptimal location, by a combination of compass, depthmeter, and log data. This relative reckoning is achieved, taking into account an hypothesis on the undercurrent value. A second absolute reckoning module is proposed for improving the AUV location, independent from the undercurrent. The module takes into account proximity to an offshore jacket, accumulates and fuses data given by all echo-sounder belt together with AUV attitude, and takes into account an a priori map of the surroundings. The methodological tools for the accumulation-fusion process and how to compute the optimal knowledge model steps are demonstrated. Results on simulated and real data obtained by the experimental sensor frame (provided with a compass and all echo-sounder belt) are presented.<<ETX>>