Real-time, model-aided ranging for small scale operations in the Beaufort Sea, an ice-covered, double ducted environment

Abstract

This talk will cover operational methods and results for vehicle navigation from the Ice Exercise in March 2020 (ICEX20), in the Beaufort Sea. For short range acoustic transmissions, the total ice cover and the double ducted environment co-create complex multipath uncertainty. We assumed that the horizontal group velocity between source and receiver was smoothly varying, and embedded a “Virtual Ocean” with a real-time ray tracing engine to predict the horizontal group velocity for range estimation. This model-aided psuedorange approach, which enabled a successful vehicle recovery, favored the least multipath possible such that psuedoranges tended to overestimate the GPS-derived range by roughly 10 m. In post-processing, we modify the group velocity estimation method to consider varying degrees of multipath and achieve a mean absolute error of roughly 4 plus or minus 4 meters, rivaling GPS performance. To our knowledge, these results are the first field experiment to demonstrate a real-time, model-based data processing to constrain ranging error for underwater navigation. [Work supported by ONR & the NDSEG fellowship.]