Automated Acoustic Arrival Matching for Long Range Subsurface Positioning of Autonomous Seagliders

Cristian Eric Graupe, University of Rhode Island

Abstract

An automated method was developed to align underwater acoustic receptions at various depths and ranges to a single reference prediction of long range acoustic arrival structure as it evolves with range in order to determine source-receiver range. Acoustic receptions collected by four autonomous underwater vehicles deployed in the Philippine Sea as part of an ocean acoustic propagation experiment were used to demonstrate the method. The natural sound speed duct in temperate waters known as the Sound Fixing and Ranging (SOFAR) channel enabled stable acoustic transmissions measured at ranges up to 700 km. The arrivals were measured in the upper 1000 m of the ocean at ranges up to 700 km from five moored, low frequency broadband acoustic tomography sources. Acoustic arrival time structure for pulse compressed signals at long ranges is relatively stable, yet real ocean variability presents challenges in acoustic arrival matching. The automated method takes advantage of simple projections of the measured structure onto the model space that represent all possible pairings of measured peaks to predicted eigenrays and minimizes the average travel-time offset across selected pairings. Compared to ranging results obtained by manual acoustic arrival matching, 93\% of the automatically-obtained range estimates were within 75 m of the manually-obtained range estimates. Least squares residuals from positioning estimates using the automatically-obtained ranges with a fault detection scheme were 55 m rms.

A machine learning model was developed to automatically align underwater acoustic measurements taken at various depths and ranges from a transmitting source in the Philippine Sea to a reference model of long range acoustic arrival structure, simultaneously determining source-receiver range and travel-time offsets associated with multipath arrivals. Ocean sound-speed variability complicates the task as the measured arrivals may exhibit scattering not present in range-independent predictions. Monte Carlo style broadband parabolic equation simulations through random internal wave fields consistent with the Garrett-Munk internal wave energy spectrum were used to generate a large data set of simulated acoustic receptions including scattered multipath arrivals, with known source-receiver ranges and imposed travel time offsets. These simulated receptions were used to train and evaluate the machine learning model for arrival pattern matching to the reference model. The inclusion of various data dimensions, such as peak amplitude, width, and contextual information such as range and depth were also explored as input to the model. Ranging results for the machine learning model were compared to a programmatic solution engineered for the same task.