Adaptive multiview planning for maximum information value in autonomous sonar imaging
Abstract
Project SummaryApproved for Public Release by Silvia Ferrari.Autonomous underwater vehicles (AUVs) are now commonly employed to p"erform imaging ofunderwater objects and environments for many Naval applications ranging from targetclassification to mapping and inspection tasks. Because the image characteristics depend notonly on the target class, but also on the environmental conditions and imaging sensor modalities,it can be particularly challenging for an AUV to decide how to improve upon the image qualitybased on existing information, without human intervention. The proposed research will develop anew class of autonomous sensor path planning algorithms for imaging AUVs that are able toplan and react to the sonar images collected in real time. The theory and algorithms developed inthis project will be applicable to a broad class of autonomous imaging problems and sensingobjectives. For example, the proposed theoretical framework will provide solutions forminimum-time or minimum-distance problems in which the sonar-equipped AUV seeks tomaximize detection and/or classification performance or cover a set of targets or area of interest,while minimizing the energy required based on bathymetry and ocean currents information. Byexploiting recent results by the PI in the area of deep-learning information value combined withmethods for in-situ environmental characterization, this project will develop an approach forplanning both the mode and path of underwater mobile sensors based on target information andknowledge of environmental conditions obtained before and during the mission. The proposedresearch will also develop collaborative fusion and planning algorithms for multiple imagingAUVs applicable to teams of heterogeneous sensors with low-bandwidth communication anddifferent imaging resolution and aperture. The image processing and planning algorithmsdeveloped in this project will be demonstrated on a high-fidelity simulation of AUVs equippedwith synthetic aperture sonar (SAS) developed and implemented by Jason Isaacs at the NavalSurface Warfare Center (NSWC), Panama City (FL). The PI will continue to work closely withNSWC, as well as Tom Wettergren at the Naval Undersea Warfare Center (NUWC), on theintegration and interface of the sensor planning theory developed in this project with futureNaval target recognit"ion and scheduling algorithms for autonomous undersea sensingapplications.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 24, 2019
- Source ID
- N000141912144
Entities
People
- Silvia Ferrari
Organizations
- Cornell University
- Office of Naval Research
- United States Navy