Tele-Supervised Adaptive Ocean Sensor Fleet

A software architecture and system deploys robotic boats to study ocean surface and subsurface phenomena such as coastal pollutants, oil spills, and hurricanes.

The Tele-supervised Adaptive Ocean Sensor Fleet (TAOSF) is a multi-robot science exploration architecture and system that uses a group of robotic boats (the Ocean-Atmosphere Sensor Integration System, or OASIS) to enable in-situ study of ocean surface and subsurface characteristics and the dynamics of such ocean phenomena as coastal pollutants, oil spills, hurricanes, or harmful algal blooms (HABs). The OASIS boats are extended-deployment, autonomous ocean surface vehicles. The TAOSF architecture provides an integrated approach to multi-vehicle coordination and sliding human-vehicle autonomy.

One feature of TAOSF is the adaptive re-planning of the activities of the OASIS vessels based on sensor input (“smart” sensing) and sensorial coordination among multiple assets. The architecture also incorporates Web-based communications that permit control of the assets over long distances and the sharing of data with remote experts. Autonomous hazard and assistance detection allows the automatic identification of hazards that require human intervention to ensure the safety and integrity of the robotic vehicles, or of science data that require human interpretation and response. Also, the architecture is designed for science analysis of acquired data in order to perform an initial onboard assessment of the presence of specific science signatures of immediate interest.

TAOSF integrates and extends five subsystems developed by the participating institutions: Emergent Space Technologies, Wallops Flight Facility, NASA’s Goddard Space Flight Center (GSFC), Carnegie Mellon University, and Jet Propulsion Laboratory (JPL). The OASIS Autonomous Surface Vehicle (ASV) system, which includes the vessels as well as the land-based control and communications infrastructure developed for them, controls the hardware of each platform (sensors, actuators, etc.), and also provides a low-level waypoint navigation capability. The Multi-Platform Simulation Environment from GSFC is a surrogate for the OASIS ASV system and allows for independent development and testing of higher-level software components. The Platform Communicator acts as a proxy for both actual and simulated platforms. It translates platform-independent messages from the higher control systems to the device-dependent communication protocols. This enables the higher-level control systems to interact identically with heterogeneous actual or simulated platforms.

The Adaptive Sensor Fleet (ASF) provides autonomous platform assignment and path planning for area coverage, as well as monitoring of mission progress. The System Supervision Architecture (SSA) provides high-level planning, monitoring, tele-supervision, and science data analysis. The latter is done using the Inference Grid (IG) framework to represent multiple spatially- and temporally-varying properties. The Inference Grid is a probabilistic multi-property spatial lattice model, where sensor information is stored in spatially and temporally registered form, and which is used for both scientific inferences and for vehicle mission planning. The information in each Inference Grid cell is represented as a stochastic vector, and metrics such as entropy are used to measure the uncertainty in the IG. The IG is used for analysis of science data from both the OASIS platforms and external sources such as satellite imagery and fixed sensors. These data are used by the SSA in planning vessel navigational trajectories for data gathering. The SSA also provides an operator interface for those occasions when a scientist desires to exert direct monitoring and control of individual platforms and their instruments.

Using this architecture, multiple mobile sensing assets can function in a cooperative fashion with the operating mode able to range from totally autonomous control to tele-operated control. This increases the data-gathering effectiveness and science return while reducing the demands on scientists for tasking, control, and monitoring. This system is applicable also to areas where multiple sensing assets are needed like ecological forecasting, water management, carbon management, disaster management, coastal management, homeland security, and planetary exploration.

This work was done by Gregg W. Podnar and John M. Dolan of Carnegie Mellon Univeristy, Alberto Elfes of Caltech, and Jeffrey C. Hosler and Troy J. Ames of Goddard Space Flight Center for NASA's Jet Propulsion Laboratory.
NPO-45478