Military Considers the Human Factor in Independent Robots
Three projects endeavor to improve how troops benefit from autonomous vehicles.
An intelligence analyst from the U.S. Army’s 3rd Infantry Division assembles the Raven unmanned aerial vehicle (UAV). Command and Control of Small Robotics Assets Display (C2RAD) project personnel have studied the Raven to understand better how the C2RAD could benefit UAVs. C2RAD is designed to support the field-deployed human operator using a robotic vehicle as a forward observer.
Researchers are conducting cutting-edge investigations in the area of unmanned systems. The efforts aim to change how humans operate the vehicles by reducing the number of personnel hours and dedicated resources necessary to execute the systems. The projects also could both improve how systems interact with one another and increase their autonomy.
The growing use of unmanned systems across all military sectors is the direction of the future. Optimizing human-robot interaction is critical because the farther removed the operators are from the system, the more important their ability to intervene becomes. Along the same lines, the fewer the number of people involved in a system’s operation, the more important each individual becomes. Because humans tend to be the most flexible part of “unmanned” systems, considering the role of the human early in the development process creates the best functional devices later.
According to Michael Paley, senior vice president for business development at Aptima Incorporated,
To that end, Aptima is working on one project for the U.S. Army and two with the U.S. Navy to improve unmanned systems. The Command and Control of Small Robotics Assets Display (C2RAD) is an Office of the Secretary of Defense-funded program administered by the Army and intended to support the human operator in the field using a robotic vehicle as a forward observer. C2RAD developers envision it as a handheld device that can plan the route of a robotic asset and link its intelligence to other members of the system, such as soldiers in a platoon. In addition to the Army and Aptima staff, Lockheed Martin personnel are working on C2RAD.
The first Navy program, contracted out of the Office of Naval Research (ONR), is the Mixed Initiative Machine Instructed Computing (MIMIC) project. MIMIC aims to help unmanned systems work more independently, capturing knowledge from human operators and embedding into the control devices the decision-making skills of unmanned aerial vehicle (UAV) commanders. The second Navy program, Collaborative Optimization System for Mixed-Initiative Control, or COSMIC, is designed for large-scale military operations. COSMIC, also an ONR program, should result in a collaborative environment that enables operators to coordinate multiple unmanned vehicles. It builds on the Mission Displays for Autonomous Systems (MiDAS) program and will integrate with the Lockheed Martin Intelligent Control and Autonomous Replanning of Unmanned Systems. A goal of COSMIC is to reduce operator workload while improving the shared situational awareness across an entire mission team.
Looking to the future, the Navy wants to move from having a team of operators for one unmanned system to having a one-to-one or one-to-many operator systems ratio. With COSMIC, the Navy is examining more closely how humans should interact with unmanned systems and how humans would interact with different types of systems.
“MIMIC is looking at something a little different, which is the question of how a human could actually participate in trying to have an unmanned systems vehicle learn things during operation,” explains Marc Steinberg, program manager for COSMIC and MIMIC at the ONR. He adds that users want more intelligent systems because humans need more automation to achieve their goals.
While the work being conducted in both Navy programs focuses mainly on UAVs, COSMIC also looks at how to manage heterogeneous systems. Operating varied unmanned systems poses a unique set of problems because these robots move at different speeds, have different capabilities and encounter different hazards. For example, undersea vehicles may not have contact for hours while UAVs may have constant communications. Researchers are striving to determine how humans can understand all the data from the various platforms.
Safety is another concern, Steinberg shares. Developers are exploring a safe zone where groups of vehicles could go and where operators would know the unmanned systems are safe. With different types of systems, such a location may not prove useful. A UAV could reach a potential safety zone quickly, but an unmanned surface vehicle (USV) might require much more time.
Both COSMIC and MIMIC provide new capabilities to troops. With COSMIC, the Navy is looking at different approaches not only for how humans interact with unmanned systems but also for how the systems work with each other while interacting with humans. With MIMIC, the sea service is determining how unmanned systems could perform tasks without preprogramming. Instead, humans would teach the systems. Steinberg says such a capability would be a huge leap in terms of being able to use unmanned systems in unpredictable and unplanned environments.
The skills taught to unmanned systems in MIMIC would have general applications. For example, humans could teach the systems what factors to look for to optimize a new route. However, the system has to be constrained to keep it from performing unwanted activities. “That’s really the hard technical challenge with this type of approach,” Steinberg explains. Researchers need to ensure that the system can generalize and not simply develop only a point conclusion. MIMIC has a high technical risk and of all the programs, it has the greatest chance of producing no discernible outcome. The benefit to warfighters is to have greater capabilities with unmanned systems and to use the systems in a broader range of applications.
COSMIC’s advantage for the military is manpower consolidation. In addition to having multiple people operating one system, these personnel would undergo extensive training to work with each system. COSMIC aims to create more commonality among systems, as well as simpler devices, and to control more—or the same number of—systems with fewer operators.
|A PackBot remote-control robot is tested in preparation for an improvised explosive device drill. Developers of the C2RAD are studying PackBots to determine how C2RAD could implement with them and other small, unmanned ground vehicles.|
In the Navy, COSMIC is being examined for use on the littoral combat ship (LCS), which has a combination of UAVs, USVs and other unmanned systems. The sea service is working to mitigate the problem of having a small number of people to control a large number of unmanned systems on the LCS. As a research project, program personnel are looking ahead to future platform developments. COSMIC could have applications for future unmanned combat systems.
Steinberg echoes Paley that in the past, developers tended to ignore the human factor when creating autonomous systems. The thinking then was that programs needed to create a good system. Someone developed an interface, and a project was ready to roll. “I think you need to really understand the role of the human,” Steinberg states. “Programs like MiDAS and COSMIC are giving a lot of insight into that area.”
In the Army’s C2RAD program, researchers also are examining how humans can interact with autonomous systems and how to integrate with various unmanned vehicles. According to Thomas Aten, C2RAD project manager for Aptima, “C2RAD is envisioned to not be specific to a given robotic system platform. In essence, the concepts we have come up with to date would apply to any small, unmanned robotic system used by small units as a force multiplier for achieving mission objectives.”
Despite C2RAD’s broad platform, program officials are exploring opportunities to transition its application for use with PackBots and other small, unmanned robotic systems that iRobot Corporation is developing. C2RAD personnel have focused on coming to an understanding of PackBots’ capabilities and limitations. At the same time, officials have been identifying how such variables of small, unmanned ground vehicles (UGVs) are related to decisions that operators and patrol leaders make while determining how to deploy these systems to maximize their force-multiplying capabilities to achieve given mission objectives. “This, in turn, helped us in defining our information architecture as well as our user interface concepts,” Aten explains. C2RAD officials have gone through the same process with the Raven UAV platform to understand how the capabilities and limitations of small unmanned aircraft systems are related to leadership decisions.
C2RAD is trying to solve the problem that future small units may have trying to coordinate information from several small, unmanned vehicles. The program aims to assist operators and patrol leaders during mission execution by deploying the unmanned assets and maximizing force-multiplying capabilities through visualizations and other tools. C2RAD will provide an integrated display that maps and displays the locations of red, green and blue entities such as snipers, obstacles and friendly forces. The information is shared with troops and commanders.
Program personnel are creating technology that will provide the robot-asset controller with appropriate information without including extraneous, distracting data. C2RAD could be used during replanning of deployment to ensure troops can effectively collaborate and supply necessary information to the patrol leader. The asset controllers need to know where they are deploying the robots, where to be looking and what they need to tell the patrol leader. The systems allow for this to occur effectively.
All information is pushed back graphically, and all the information will be related in the displays. With aerial footage, height and distance will affect visibility, enemy detectability and sensor footprint quality. C2RAD will be able to relay consequences to the patrol leader because all the information will be linked.
As with COSMIC and MIMIC, the major benefits of C2RAD relate to the human factor in the systems. Aten believes troops will benefit from the system’s ability to manage data when small field units work with multiple assets in the future, because humans will be at risk for information overload. C2RAD will allow the effective integration of the robotic assets into units.
C2RAD developers plan to integrate their technology with Lockheed Martin’s Distributed Operations (DISOPS), a software toolset for soldiers at platoon level and below. C2RAD will run on DISOPS, which will be integrated into platforms. As an example, humans controlling Raven UAVs would look at DISOPS and understand what information they need to collect and supply back to leadership, but the operator would continue to control the asset to gather necessary information. DISOPS would be used to show the C2RAD information to the leadership.
While none of the C2RAD, COSMIC or MIMIC projects is starting from scratch, the work involves more than simple modifications to prior technologies and ideas. Development requires a mix of changes in terms of sensing technologies, visible elements, more accurate radars and the Global Positioning System. According to Paley, the functions unmanned vehicles can carry out continue to multiply. The human’s role becomes the variable, and researchers have to determine where the automation fits in.
Paley says debate still rages about what systems should do autonomously and which actions users should perform on their own. “Ultimately what’s nice about these projects is that there really are some interesting challenges that let you shape the future of the way people might interact with technical systems,” he shares.
Researchers are developing theories as well as prototype displays that could be built into actual systems. The projects will hold off on building the real control for the robots until the next round of funding. The end goal is to have real, fielded prototypes—not just data—at the end of the projects.