Development and testing of a real-time terrain navigation method for AUVs

Underwater localization is still a significant challenge for an Autonomous Underwater Vehicle (AUV) during long endurance missions. Typically the vehicle must rely on external infrastructure such as Global Positioning System (GPS) satellites or underwater acoustic positioning system transponders to provide navigation corrections. These options will not always be feasible during some missions. This paper describes applied research into addressing this challenge through the development of a Monte Carlo Localization (MCL), or particle filter framework. The goal is to supplement an AUV navigation system with independent position fixes based on environmental observations. A particle filter framework for bathymetric terrain navigation is developed and tested using real in-water vehicle and sensor data collected with NUWCs 12¾" Mid-Size Autonomous Reconfigurable Vehicle (MARV) AUV. The results illustrate how this technique can be successfully applied to allow an AUV to recover from significant position uncertainty and errors during a mission. Additionally it is concluded that extending the particle filter to three dimensions to account for depth measurement bias is preferable to other techniques for handling tidal bias.