Our objective is to develop, demonstrate, and evaluate a prototype feature-based navigation and localization system for miniature Unmanned Underwater Vehicles (UUVs) that meets the performance requirements for these systems yet provides a drastic reduction in size, weight, power, and cost requirements over current solutions. The proposed method utilizes a micro mechanically scanned imaging sonar (uMSIS) as the perception sensor to aid an Inertial Navigation System (INS) based on a low-cost MEMS IMU. Our approach implements a deconstructed Inertial-SLAM method optimized for light-weight embedded systems to reduce the order of the state-vector and model the errors for the inertial pose and feature pose estimates as well as make map corrections. The focus of our approach is to reduce size, weight, and power requirements for the entire navigation system by developing computational and image processing methods that will make use of commercially available micro and miniature sensors possible. This effort will extend the current body of knowledge towards providing an accurate, robust, and commercially viable navigation and localization solution using a light-weight inertial feature-based method. We will evaluate a prototype of our proposed system on a miniature ROV in open-water structured and unstructured environments to assess the feasibility of this approach.
Keywords: inertial navigation, feature-based navigation, SLAM, MEMS, MSIS, miniature UUV, feature extraction, image processing
