Feature Based Localization and Navigation for Miniature Underwater Vehicles

The goal of this proposal is to produce and demonstrate a navigation system combining a small, low power, mechanically-scanned sonar and a MEMS-based inertial measurement unit that will enable a miniature undersea vehicle to travel tens of kilometers and then return to its home location within 20 meters without surfacing. The system will be designed for an overall weight less than 10 pounds and average power consumption less than 6 Watts.The navigation goals will be achieved by incorporating the sensor data into a real-time Simultaneous Localization and Mapping (SLAM) framework.This framework will exploit recently developed sonar correlation techniques such as optical flow, feature matching and acoustic shadow prediction, adapting them to the sensor suite limitations, for example properly accounting for the data smearing associated with mechanically-scanned sonars.The phase I effort will focus on demonstrating the viability of the navigation solution, in particular the ability to use features extracted from scanning sonar data for dead-reckoning and localization.The navigation sensors will be integrated and tested on a tow body platform in the Chesapeake Bay during the Phase I option.