S&AS: INT: Taskable and Adaptable Autonomy for Heterogeneous Marine Vehicles

Oregon State University, Corvallis OR

Investigators

Abstract

Underwater exploration using unmanned robotic vehicles has opened up vast new areas for understanding the world's oceans. However, in the current state of practice, human operators must provide specific waypoints for the vehicles to follow, which is both time-consuming and inflexible. This project will develop autonomy capabilities that facilitate on-vehicle intelligence, leading to longer duration deployments of unmanned underwater and surface (on-water) vehicles. Specifically, the project will enable autonomous information gathering missions, such as searching for an area of interest and tracking targets of interest (e.g., biological hot spots in the ocean). As the autonomous capabilities improve over the course of this project, the amount of time that humans must dedicate to interacting with, or correcting, the vehicle's actions will be reduced, drastically improving the efficiency and reducing the cost of ocean data collection. The ultimate objective is to develop intelligent autonomy capabilities that expand the functionality of unmanned marine systems and reduce the need for human intervention and control. This goal will be achieved by focusing on three key areas of marine vehicle autonomy: (1) the ability for humans to provide high-level, taskable specifications to vehicles in a way that is understandable to the autonomy system; (2) cognitive capabilities of autonomous vehicles to provide intelligent in-situ adaptation with changing conditions in long-duration missions; and (3) reflective and introspective failure recovery allowing the vehicles to self-assess faults and respond to them appropriately while still meeting the mission specifications. The proposed techniques will be validated through two ocean deployments, consisting of multiple underwater gliders and autonomous surface vehicles, as well as extensive simulated testing using satellite data and data from field deployments.