Collaborative Adaptive Sampling & Mapping
Abstract
We propose to develop and demonstrate the capability of a squadron of unmanned underwater vehicles (UUVs) to collaboratively and adaptively sample and map their environment. The simultaneous deployment of multiple UUVs is motivated by the desire to characterize dynamic structures in the ocean such as fronts, plumes, and currents (Shcherbina 2008a,b). Specifically, we will address the areas of UUV navigation and adaptive map building with an emphasis on scalability to larger squadrons. Our efforts will focus on software and hardware integration into two small, man-portable UUVs. We will conduct local tests off of a small boat using simulated environmental data or a natural opportunity such as estuarial tidal outflow after an intense rainfall event. Subsurface navigation in the absence of GPS typically utilizes expensive inertial sensors that are not practically scalable to squadrons of small UUVs. One-way travel time inverted ultra-short baseline (OWTT-iUSBL) acoustic localization solves this scalability problem using known signals that are periodically broadcast from a source platform and heard by any number of receiver platforms. The array on each receiver platform is able to resolve the range, bearing, and inclination to the transmitter, providing an instantaneous estimate of vehicle location relative to the source platform. This has previously been demonstrated using broadcasts from static and autonomous surface platforms (Rypkema 2017). A specific focus of our research effort will be to demonstrate this technique using broadcasts from a member of the UUV squadron. Constructing a map both adaptively and collaboratively requires a framework that can be readily augmented with new data while remaining lightweight enough to share between platforms over the bandwidth-limited acoustic channel. Online summaries offer asolution to this problem by clustering incoming data streams and representing those data as a subset of exemplary data points. They are particularly well suited to real-time robotics applications since they provide a current summary at any point in the mission and do not require the entire dataset to be collected first. We have previously demonstrated the ability of online summaries to select characteristic imagery from aUUV optical camera survey (Kaeli 2015). A second focus of our research will be to extend this technique to summarizing both in the feature space of the sensor as well as the physical space of the environment for the purposes of constructing a map.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 26, 2018
- Source ID
- N000141812543
Entities
People
- Jeff Kaeli
Organizations
- Office of Naval Research
- United States Navy
- Woods Hole Oceanographic Institution