Surface Autonomy Is Heading for the Fleet

By Dr. Robert, Brizzolara, Office of Naval Research

The technical challenge of operating an unmanned vehicle with little or no human supervision increases rapidly as a function of the complexity of a mission and its environment. Today, the Navy’s unmanned surface vehicle (USV) prototypes require substantial (remote) manpower to perform even relatively straightforward missions—and are therefore dependent on a radio communications link to a human controller. For example, collision avoidance is performed remotely by a human operator; this requires a large amount of bandwidth to transmit video from the USV and places a high cognitive workload on the human operators. This limits USVs to operating close to manned platforms or a ground control station. In addition, a USV’s operational capability deteriorates quickly as environmental conditions worsen, in large part because of the degraded situational awareness of the remote human operator.

Science and technology (S&T) will overcome these issues and allow USVs to realize their full potential to make them capable of accomplishing complex tasks in all environments, as directed by the Secretary of Defense in a 2011 memo that made autonomy one of seven priorities for the Department of Defense. As a result of the efforts of the Office of Naval Research (ONR) and other naval S&T organizations over more than 10 years, the Navy’s first procurement of USVs began in 2013. USVs are of military interest because they possess outstanding platform performance characteristics, such as range and speed, that result from air-breathing propulsion, access to radio communications, potential for stealth design features, and low cost per quantity of payload.

In 2002, as the Navy’s interest in USVs began to increase, ONR’s Sea Platforms and Weapons Division engaged with the Navy’s USV program office regarding the S&T that would be needed to build what the Navy envisioned. One of the needs identified was autonomous control. In 2004, I initiated a program the objective of which was autonomous control of USVs over long, complex missions in unpredictable or harsh environments. Since then, several sponsoring organizations have contributed to this effort, including ONR, the Unmanned Maritime Systems program within Program Executive Office Littoral Combat Ships, and the Naval Sea Systems Command Engineering Directorate’s Technology Office. The autonomous control system developed to date has been installed on eight different USV types, has achieved more than 3,500 nautical miles of testing on the water, and has participated in numerous fleet experiments. One USV, called the “Unmanned Sea Surface Vehicle—High Tow Force,” was developed by ONR’s Sea Platforms and Weapons Division. It is about 12 meters long and is therefore a “fleet class” USV as defined by the USV Master Plan. An adaptation of this design in conjunction with a mine influence sweep payload developed by ONR’s Ocean Battlespace Sensing Department was used for the two “Unmanned Influence Sweep System” prototypes operated by the Unmanned Maritime Systems program office.

Autonomous control of USVs in this context means operation of the craft either with no human operator (after specification of the mission goals and constraints at the start of a sortie) or a limited human supervisory role, perhaps in degraded conditions, in the vicinity of other maritime traffic and performing an operationally useful task. This is a substantial technical challenge because of the unique and often harsh dynamics of the sea surface and their impacts on situational awareness, which degrades the range at which hazards can be detected. There may be only a short warning time of a hazard in a USV’s path. Therefore, the USV’s autonomous control system must have fast, accurate perception and decisionmaking capabilities.

A team composed of the California Institute of Technology’s Jet Propulsion Laboratory, Spatial Integrated Systems Inc., Daniel Wagner Associates, and Naval Surface Warfare Center Carderock Division has made substantial progress toward this goal by developing a technology called Control Architecture for Robotic Agent Command and Sensing (CARACaS). This system consists of two components: a perception component that provides situational awareness, and a decision-making component that determines boat course and speed based on the output of the perception component.

The decision-making component of an autonomous USV must make route planning determinations over a wide range of time scales.

The perception component employs multiple sensing modalities, principally electro-optical/infrared sensors and radar, to increase the probability of detection and accuracy over what a single modality can provide. Commercially available sensors, such as radar, are used in CARACaS when available; if these do not exist, then the sensor is developed. For example, in the ONR program, the Jet Propulsion Laboratory developed a stereo electro-optical system that provides sufficient range and near-real-time processing speed to support high-speed USV operations.

Stereo electro-optical was chosen as one of the sensing modalities because of its ability to achieve the range to provide the necessary reaction time for the desired USV speeds, and to support future incorporation of vessel classification algorithms necessary for implementation of certain collision regulation rules. A companion stereo infrared capability for night operations is currently being developed. The perception component provides the range, speed and bearing of all contacts to both the reactive and deliberative decision-making components.

The decision-making component of an autonomous USV must make route planning determinations over a wide range of time scales. Since a particular autonomous planning algorithm works best only within a limited range of time scales, the Jet Propulsion Laboratory employed a hybrid of a short time scale, or reactive, component and a longer time scale, or deliberative, component. For example, since the highly dynamic environment means that some contacts will not be detected until they are at relatively short distance from the USV, commonly available graph theoretic path planners are not suitable for reactive decision making for USVs because they are not fast enough. Instead, the Jet Propulsion Laboratory used its Robust Real-time Reconfigurable Robotics Software Architecture” (R4SA), which employs a behavior-based approach using a very fast path planning algorithm based on “velocity-obstacles.” Velocity obstacles is a route planning approach, similar in principle to the maneuvering boards used by human navigators, that can accomplish very fast, reliable computation of routes that accomplish hazard avoidance and compliance with the maritime rules of the road.

In addition to reactive decision making, USVs must plan their routes over periods of hours or days. The velocity obstacles approach is not capable of planning at these longer time scales. So, an existing capability called CASPER (Continuous Activity Scheduling Planning Execution and Replanning) is used within the hybrid framework as the deliberative decision-making component of CARACaS. CASPER employs a graph-theoretic approach that plans a route based on mission goals and constraints. The series of waypoints determined by CASPER is provided to R4SA once every several seconds. R4SA then executes each waypoint in order while avoiding collisions and obeying rules of the road.

A few years ago, when CARACaS was still in early development, it was clear that it would be feasible to accomplish autonomous control of USVs at low speeds and in benign conditions. The notion, however, of being able to achieve the reactive autonomy necessary for a high-speed craft, especially in degraded environments, involved much higher technical risk. Although CARACaS had been tested extensively on water for several years, a key milestone in the ONR program was in 2011 with the first on-water demonstration of the CARACaS reactive autonomy performing a complex action. The CARACaS system combined three behaviors: “go to waypoint;” “comply with collision regulations;” and “avoid collision.” This is referred to as parallel behavior composition. The test showed that autonomous control using perception and decision making was fast enough for a fleet-class USV in a relatively complex situation.

There is a significant challenge in fostering greater acceptance of autonomously controlled USVs by the Navy. Before turning control over to an autonomous system, a commander must have the confidence that it will perform the appropriate action in a given situation.

There is much that remains to be done in the autonomous control of USVs, such as attaining: accurate and fast situational awareness in higher sea states; efficient and effective algorithms for handling multiple competing objectives; cooperative decision making across multiple unmanned platforms; accurate, fast, distributed fusion of sensor data across multiple unmanned platforms; reliable detection of submerged hazards; activity recognition of other maritime vessels; and compliance with additional collision regulations (e.g., recognition of day shapes and lights). In addition to these technical challenges, there is a significant challenge in fostering greater acceptance of autonomously controlled USVs by the Navy. Before turning control over to an autonomous system, a commander must have the confidence that it will perform the appropriate action in a given situation. A “human oversight mode” will be a useful and important initial approach that will enable the Navy to gain trust in the technology. As confidence grows, the degree of human oversight can be reduced. Furthermore, trust in the system can be gained by carefully selecting the situations in which the autonomous capability is initially employed (areas with low-contact density and good environmental conditions) and using the USV to perform relatively simple tasks. As trust grows, autonomous USVs will be used for more complex tasks in more complex environments.

As autonomous control of fleet-class USVs continues to be developed and demonstrated at sea, attention is now turning to the possibility of autonomous control of larger unmanned surface vessels. There are compelling reasons to explore development of larger unmanned craft—they have larger payload capacities, the ability to operate in higher sea states, and longer ranges (permitting the possibility of self-deployment). Vessel sizes necessary to achieve self-deploying ranges are still quite small compared to today’s Navy surface ships,  including the littoral combat ship, meaning that their cost still can be modest. A self-deploying unmanned vessel requires a high degree of autonomy since it will operate at long distances from other ships and ground control stations. Achieving autonomous control of a larger vessel involves additional technical challenges since the number and complexity of shipboard systems is much larger. The Defense Advanced Research Projects Agency is addressing these challenges by developing the Anti-Submarine Warfare Continuous Trail Unmanned Vessel, a 130-foot autonomous trimaran scheduled for launching in 2015.

Given current budget realities, the Navy must find innovative ways to perform its missions in more cost-effective ways. Fleet-class USV prototypes have already demonstrated capabilities that previously required much larger and costly manned platforms. Development of autonomous control for USVs will enhance this cost savings by further increasing the capability of the platform. Capability increases will accrue by freeing USVs from the tether of the high-bandwidth communications, thereby allowing them to venture farther from humans. In addition, the USV can be designed such that no one ever has to set foot on board. This benefit is much greater than the space and weight savings derived from the absence of human support systems. There also are benefits from relaxing the structural and safety requirements associated with manned vessels. All of this translates into additional space and weight for payload and fuel, further increasing the warfighting capability that can be delivered using small, unmanned combatant craft.

About the author:

Dr. Brizzolara is a program officer with the Office of Naval Research’s Small Combatant Craft science and technology program.