An Information Theoretic Approach to Interacting Multiple Model Estimation for Autonomous Underwater Vehicles
Abstract
Accurate and robust autonomous underwater navigation (AUV) requires the fundamental task of position estimation in a variety of conditions. Additionally, the U.S. Navy would prefer to have systems that are not dependent on external beacon systems such as global positioning system (GPS), since they are subject to jamming and spoofing and can reduce operational effectiveness. Current methodologies such as Terrain-Aided Navigation (TAN) use exteroceptive imaging sensors for building a local reference position estimate and will not be useful when those sensors are out of range. What is needed are multiple navigation filters where each can be more effective depending on the mission conditions. This thesis investigates how to combine multiple navigation filters to provide a more robust AUV position estimate. The solution presented is to blend two different filtering methodologies utilizing an interacting multiple model (IMM) estimation approach based on an information theoretic framework. The first filter is a model-based Extended Kalman Filter (EKF) that is effective under dead reckoning (DR) conditions. The second is a Particle Filter approach for Active Terrain Aided Navigation (ATAN) that is appropriate when in sensor range. Using data collected at Lake Crescent, Washington, each of the navigation filters are developed with results and then we demonstrate how an IMM information theoretic approach can be used to blend approaches to improve position and orientation estimation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2022
- Accession Number
- AD1184921
Entities
People
- Timothy J. Ii Howarth
Organizations
- Naval Postgraduate School