Next-gen 'Internet of Battlefield Things' on the way
In the not-too-distant future, humans and smart technology may merge to function as a cohesive network, providing warfighters with “extrasensory” perception in terms of battlespace situational awareness, prediction powers, and risk assessment.
What would the ability to network computers, sensors, cameras, data, weapons, soldiers, smart wearables, and media analytics together mean for the military? A $25 million initiative, funded by the Army Research Lab and led by researchers at the University of Illinois at Urbana-Champaign, is currently underway to find out and to lay the scientific foundations for what it calls the Internet of Battlefield Things (IoBT).
Today, there are already many examples of the military connecting “things,” such as ships, planes, soldiers, and operating bases. But researchers believe the network can be greatly expanded to leverage advancements in unmanned systems and machine intelligence to achieve superior defense capabilities.
The idea of “interconnecting military ‘things’ at a global scale dates back at least to the 1990s, with concepts such as the Global Information Grid and Network Centric Warfare,” says Tarek Abdelzaher, a professor of computer science at Illinois and the academic lead of the Army Research Lab’s Alliance for IoBT Research on Evolving Intelligent Goal-driven Networks (IoBT REIGN). “What’s new is that we’re no longer talking about merely interconnecting these assets. The IoBT is more about building intelligent battlefield services that benefit from this connectivity and keep people out of harm’s way.”
One goal is “to execute missions with fewer soldiers and more machines, because the cost of life cannot be compared to the cost of hardware,” he says. “This leads to fundamental research questions in machine intelligence and autonomy.”
In the future, you should expect every battlefield “thing” within the IoBT realm to have a little bit of a brain or some smarts, not just a radio. As you might imagine, this ushers in “new research questions about understanding and managing the aggregate behavior of heterogeneous intelligent things when their ‘brains’ start interacting,” says Abdelzaher. Among the questions: “How can we make them collaborate to achieve mission goals? How do we do this without micromanaging them – because we simply won’t have time to do so? How do we specify human-commander intent to them in ways they can understand so they can adapt to unexpected situations in ways that meet the spirit of the mission? These are the issues surrounding IoBTs, even more so than the interconnection fabric. Within 10 years, we want to solve these hard challenges to reduce the loss of life in military operations.”
What are the biggest challenges ahead for the IoBT? Security and robustness are definitely right up there: “We’re talking about withstanding attacks by determined, highly intelligent adversaries,” Abdelzaher elaborates. “Tomorrow’s adversaries will be technologically sophisticated, so this means warfare between two advanced technologies. Any distributed capabilities we build out of IoBT assets will need to know how to deal with disruptions and reconfigure themselves after failures. They will need real-time reflexes that allow them to adapt, react, and regroup – without human help. The system must continue to function safely and successfully even after some parts have been damaged, disconnected, captured, or compromised by the enemy.”
Connecting the IoBT with artificial intelligence (AI) is one of the “key centerpieces of the problem,” he says. “If we simply add a lot of new gadgets to the battlefield and connect them, it increases the cognitive load on the warfighter. Someone will need to configure all of those ‘things’ and troubleshoot them when some device fails or disconnects, which will be normal in an adverse environment such as a battlefield.”
Ever connect a laptop to a projector or use video conferencing technology? As Abdelzaher points out, it’s fairly rare to have these experiences go smoothly. “Someone always has a setup problem,” he notes.
But failures on the battlefield are highly undesirable while you’re being shot at, so “machines will need to figure it out, and AI is the core. The more ‘things’ you interconnect, the more intelligence you need,” Abdelzaher adds.
The researchers’ goal is to create a cyber network of “things” that are capable of adapting as a mission changes or evolves. This means that the system will need to be able to analyze its available resources and reassemble itself in response to changing requirements. Systems must also be “self-aware” and able to reason about their goals, state, vulnerabilities, and other characteristics to meet commanders’ intent.
So what types of technologies are being considered for connecting AI, which involves a lot of compute power, with IoBT at the edge of the network? Is software-defined networking (SDN) among them? “I think SDN is a promising technology,” Abdelzaher says. “We’re doing basic research, so many technology solutions are on the table and we want to understand their strengths, weaknesses, and domains of applicability, where each can really shine.”
AI’s role once it’s connected to the IoBT: The IoBT system will feature cognitive abilities and be able to fuse data from sensors and technology with data provided by humans. And the system will need to function in a continuous state of learning at multiple time scales, such as learning from previous actions while acting in the present and anticipating future moves. The goal is to have systems that can provide commanders with the most relevant information at any given moment.
“When people imagine autonomous interconnected machines, dark sci-fi movies often come to mind. But the prospect really has a lot of promise for humanity,” Abdelzaher says. “One of my friends, a civilian, was once driving over a bridge during an air raid in a war zone when an enemy jet appeared. The jet circled the bridge a couple of times until my friend finished crossing, then detonated the bridge. The human pilot in that jet clearly made a determination to spare the civilian’s life. Can we build smarter weapons that can adapt similarly to avoid loss of civilian life? In fact, can we use smarter techniques to attain the ultimate goal of zero collateral damage sometime in the future? With sufficient AI and networking of sensors and weapons, I think it’s possible.”
This effort to understand and exploit the unique capabilities of networked battlefield systems is an interdisciplinary problem that brings together researchers from cyber-physical computing, information theory, security, formal methods, machine learning, networking, control, and cognitive science.
IoT devices tend to be vulnerable to cyberattacks, but according to the IoBT REIGN researchers, their work for battlefield applications may to lead to wider advances in making future civilian IoT technologies more secure and robust when it comes to cyberattacks.
“While commercial IoT provides some of this capability, they aren’t challenged in the same manner as on the battlefield,” says Stephen Russell, the Army Research Lab’s Battlefield Information Processing Branch chief and the government lead for the Alliance. “The ‘B’ in the IoBT is our key focus.”
The Army Research Lab’s Alliance for IoBT REIGN also includes collaborators from the Army Research Lab; Carnegie Mellon University; University of California, Berkeley; University of California, Los Angeles; University of Massachusetts; University of Southern California; and scientific research institute SRI International. The project’s $25 million funding covers the first five years of a potential 10-year effort.
Editor’s note: IoT security is a huge concern, mainly because many sensors or devices come with default passwords and known vulnerabilities that can be easily found on the internet, as well as without a way to patch or push security updates to them. But a group of U.S. senators recently introduced a bill, The Internet of Things Cybersecurity Improvement Act of 2017, which aims to change this. It essentially sets a fairly low bar establishing basic security best practices so that IoT devices sold to the federal government – including the military – will be required to eliminate default passwords and known vulnerabilities, and also include a way to push security patches or updates to them.