Feature Based Localization and Navigation for Miniature Underwater Vehicles

This proposal outlines an approach for reliable autonomous Feature Based Localization and Navigation (FBLN) in the absence of self-localizing reference signals such as GPS or long baseline acoustic positioning systems. Our proposed solution will solve the FBLN problem using several complementary approaches that minimize navigation errors. These approaches include developmental hardware, SLAM, and a path planner. We propose using a 9 Degree-of-Freedom (DoF) Inertial Navigation System (INS), a Doppler Velocity Log (DVL), and side-scan sonar imagery as inputs to our FBLN system. The 9-DoF INS is a sensitive motion-measurement device that contains Micro-Electro-Mechanical accelerometers, gyros, and magnetometers. An Extended Kalman Filter is used to produce continuous yaw, pitch, and roll estimates in real-time from the 9-DoF INS data. Path planning software will attempt to generate path plans that maximize the information gained about the environment. By following a path that maximizes the feature information, the location estimates of the features will have minimal variance and therefore the FBLN system will have minimal navigation error. Part of the Phase I study will be to determine the feasibility of achieving a solution of 10 pounds and less than 15W. The proposed hardware has been equipped on platforms with a 6-inch diameter.