Objectives:The objective of this proposal is to pursue the in water validation and performance analysis of a new algorithm for feature-based navigation of autonomous underwater vehicles (AUVs) using sonar. The goal of concurrent mapping and localization (CML) is to enable AUVs to build and maintain feature-based maps of the ocean environment from sonar data and to use these maps to navigate accurate for long duration missions over large areas of the ocean. The proposed research will develop a feature-based navigation algorithm that is suitable for operation in real-time onboard an Odyssey III AUV. The new method will be tested via extensive experimentation in the Charles River.Methodology:The theoretical approach to feature-based navigation is based on Decoupled Stochastic Mapping (DSM), a new technique for CML that has been developed in our laboratory under prior Sea Grant and NSF funding support. The philosophy behind DSM is to compute multiple, computationally efficient, partial solutions to the full CML problem, rather than to compute a single computationally intractable solution, enabling real-time operation. The proposed research will extend the DSM acquired by a moving AUV. Initial experiments will use a single sonar that scans in the horizontal plane. Distinctive geometric objects, such as the supports of the Mass. Ave. Bridge or PVC pipes, will be used as features. Subsequent experiments will add a sonor that scans in the vertical plane, and will integrate mapping of bottom terrain features. Performance metrics will be developed to quantify CML algorithm performance as functions of mission duration, operating area size, feature density and proprioceptive sensor performance.Rationale:Navigation has been identified as one the key hurdles in the deployment of AUVs for a wide range of ocean applications. The capability of concurrent mapping and localization would provide a tremendous benefit to commercial and scientific users of AUVs, by enabling critical new applications of AUVs that are currently not feasible. Advances in navigation and sensing for AUVs can facilitate the solution of critical problems faced by the scientific community, such as underwater search, mapping, inspection and repair, climate change assessment, and marine habitat monitoring. CML has been one of the central research challenges in the mobile robotics research community over the past several decades. There has been considerable progress in recent years, specifically in the development of CML algorithms for land robots. To date, however, there have been no real-time implementations of concurrent mapping and localization onboard an AUV. A real-time implementation of CML onboard an AUV would be a significant research milestone.