YIP Modeling Natural Dynamic Scenes for Autonomous Littoral Operations
Abstract
YIP Proposal Submitted to Marc Steinberg & Tory Cobb Recent interest in the robotics community in metric semantic mapping is making first steps towards enabling autonomous systems to utilize high-level knowledge in the decision-making process. These techniques build a metric semantic scene graph that encodes a relationship between detected objects or areas in a scene (such as the fact that aperson is in a room or that a laptop is on a table in the office). These modeling techniques have made great leaps forward in progress and have the potential to enable high-level reasoning in tasks such as autonomous littoral operations by unmanned underwater vehicles (UUVs) and unmanned surface vessels (USVs).However, coastal environments vary significantly in time, space, and topology (due to coastal processes, seasons, tide, weather, and human-actions) resulting in the need for additional research to enable autonomous operation in littoral zones. In this project, we propose to investigate themodeling of natural dynamic scenes that vary in time, space, and topology and how these techniques can be applied to the tasks of localization and autonomous operation planning in dynamic coastal regions. We will explicitly focus this research on the use case of unmanned underwater vehicle (UUV) and unmanned surface vessel (USV) operations in dynamic coastal environments and conduct infield testing and data collection to validate the proposed ideasin this context. Specific research to be carried out through this grant include 1) The introduction of a novel scene graph representation called a Natural Dynamic Scene Graph (NDSG) that supports modeling of littoral environments and their variations over time as induced by tidal processes, weather, seasons, and human operations; 2) The development of novel techniques for robust GPS-denied localization in dynamic coastal environments by merging semantic mapping and graph-theoretic consistency-based techniques; and 3) The development of generating verifiable temporal-logic plans from natural language for autonomous operation in dynamic littoral zones. Expected outcomes include research dissemination in tier-one robotics journals and conferences, multiple datasets capturing the dynamics and challenges associated with autonomous littoral zone operation across short and long-term time-scales, an open-source simulator that supports littoral zone dynamics, and training of multiple graduate and undergraduate students in fields relevant to the US Navy research enterprise. Approved for Public Release
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 15, 2024
- Source ID
- N000142412301
Entities
People
- Joshua G. Mangelson
Organizations
- Brigham Young University
- Office of Naval Research
- United States Navy