The U.S. Defense Department has launched a research initiative aimed at potentially building autonomous robots in a few years -- machines increasingly able to perform a greater range of tasks without human intervention.
“The projects that were selected cut across at the fundamental cutting edge of the science of autonomy,” Alan Shaffer, acting assistant secretary of defense for research and engineering, said in an interview with Military.com.
The pilot, called the Autonomous Research Pilot Initiative, is designed to inspire young scientists to innovate, research and develop next-generation levels of autonomy for ground, air and underwater robotic systems, said Shaffer.
The Pentagon has awarded about $45 million to seven teams of government researchers, each tasked with solving a problem in the areas of robotics and autonomy. It encourages collaboration among scientists from military research facilities such as the Army Research Lab, Naval Research Lab and Air Force Research Lab, among others.
“We have challenged our in-house government researchers to come up with topics that will help us better understand how to do autonomous systems," Shaffer said. "We’re giving them a really hard problem, funding it and then letting them go do actual work in a strategically important area.
The research seeks to develop new algorithms to increase a system's level of autonomy so, for example, robots could interact with one another and share or “network” information in real time.
Researchers also want to improve sensors such that autonomous systems can move and react to obstacles more quickly. The goal is to link autonomous systems to one another such that they can respond to other moving objects in a fast-paced dynamic environment.
For instance, the research explores the potential of deploying small “swarms” of miniature unmanned aerial vehicles able to work in tandem with other manned and unmanned systems.
Each of the seven research focus areas has a particular title describing the area of inquiry. One is titled, "Autonomy for Air Combat Missions: Mixed Human/UAV Teams."
This approach may help to massively increase the intelligence, surveillance and reconnaissance capabilities over a certain area by blanketing it with miniature drones able to talk to one another, as well as manned and unmanned ground assets.
“With a flock of smaller, less expensive UAVs, you have a more resilient ISR capability because even if an adversary shoots down one of them, the rest of the swarm can continue on with their mission,” Shaffer said.
Research into autonomy is on the brink of such breakthroughs, one analyst said.
“Autonomy gives you many things such as safety and increased reliability," Daniel Goure, vice president of the Lexington Institute, a Virginia-based think tank, said. "Autonomy also allows you to expand mission roles for various systems. For instance, can you program it to find certain things through recognition software and then continuously track something?”
Developing systems to autonomously share sensor data with one another in real time is a big part of the next generation research. For instance, what if UAVs could send ISR data to bomb-hunting ground robots thousands of feet below, alerting them to a potential hazard?
“When you think about autonomy, think of it as the fusing of sensors with algorithms and actions," Shaffer said. "When we deploy troops, we rely on things in different domains. How do you get a UAV to interact with perhaps a ground robot and have it go investigate something? Let’s say you have a UAV flying around and it notices some disturbed earth in Afghanistan. Can you then send the signal for a ground robot to go interrogate and find out if there is an IED there? That is a hard problem."
Another research area, “Exploiting Priming Effects in Autonomous Cognitive Systems," seeks to build algorithms such that a robot or machine can perceive the surrounding environment like a bird or an ant might perceive its surroundings, Shaffer said.
The multidisciplinary research draws upon the expertise of computer scientists, neurologists, even biologists. Birds, for instance, engage is “swarming” behavior, which can provide useful insights for scientists looking to engineer swarms of mini-UAVs, Shaffer said.
“If you bring in biomimetics, you are trying to model a bird’s brain or an ant’s brain – and look at how they perceive the world around them. The idea is to provide the algorithms that would allow an autonomous system to complete a mission,” Shaffer said.
Another research category, "Autonomous Squad Member" is an effort by the military's research labs, including the Navy Center for Applied Research in Artificial Intelligence, to improve the reasoning, understanding and perception of ground robotic vehicles. Armored trucks may be able to employ acoustic and optical sensors to receive cues, input and information about roadside bombs, known as improvised explosive devices, or IEDs.
“Different types of IEDs will have different signatures and be receptive to different types of sensors," Shaffer said. "If it is command wire, you can detect the wire through effectively detecting the electromagnetic field along the wire. If it is a buried metallic device, you may need ground penetrating radar. How about having these operate together?”
Yet another research category, "Autonomy for Adaptive Collaborative Sensing," is geared toward locating targets more quickly, though it won't focus on the autonomous use of weapons, given the U.S. doctrine requiring a "man-in-the-loop" when it comes to the use of force, Shaffer said.
Overall, the project is aimed using technology to lighten the workload of humans, in some cases decreasing manpower requirements or literally the amount of gear infantry soldiers must carry on patrols, Shaffer said.
“One thing you will find is that scientists don’t really work for money – scientists work to solve problems," h e said. "These are solid projects. At the end of the time period, we will have the intellectual property either to go to a prototype or provide the information to industry."