11/16/2022 8:45:58 AM
The work, Illinois CS professor Tarek Abdelzaher explained, provides services focused on automating decision processes.
Experts say that artificial intelligence is redefining the balance of power in military conflicts, motivating government investment in the use of different degrees of automation and coordination in the battlefield.
“There is near consensus today that a decisive military advantage can be gained in contemporary battles by use of AI,” says Illinois Computer Science professor Tarek Abdelzaher, who leads the Alliance for IoBT Research on Evolving Intelligent Goal-driven Networks (IoBT REIGN).
IoBT REIGN was established in 2017 by the U.S. Army Research Laboratory, and was just awarded a 5-year contract extension with up to $25.5 million in additional funding.
“IoBT,” which stands for “Internet of Battlefield Things,” was inspired by civilian IoT (Internet of Things) technologies. However, where IoT technologies might control things like home lighting and temperature, “in the battlefield, the services are more about automating decision processes,” says Abdelzaher. “So you have sensors; they measure things; and we have to react to them, to various threats. We need to understand what the threat is; we need to identify it, track it, locate it; we need to aggregate information about multiple threats to come up with a bigger picture.”
Abdelzaher, who is a Sohaib and Sara Abbasi Professor with Illinois CS and CSL, explains that “one of the reasons for IoBT was a heightened recognition that future battlefields will be increasingly utilizing AI as part of the battlefield infrastructure.”
To date, IoBT REIGN has focused on three main research directions. One is to run calculations efficiently using diverse and distributed battlefield resources—generally small, embedded devices with limited compute capacity—“because we’re not going to have a data center sitting on the battlefield!” Another is resilience, because systems must operate in a volatile, hostile environment. The third is to enable AI to run effectively in the battlefield at the “point of need,” meaning in the field, without cloud support.
Advances were made in all three areas during the first five years of the effort. For example, the team improved the efficiency of various intelligent computations by orders of magnitude compared to five years ago; built analytical foundations for understanding the limits of resiliency; and improved AI algorithms that optimize for the resource-constrained environment “at the edge,” meaning on the battlefield.
In the next five years, IoBT REIGN will build on those advances and pursue new areas, particularly scaling, coordination, and massively increased heterogeneity of devices. “In the future, there’s going to be hundreds, if not thousands or tens of thousands, of devices on the field that are collaborating. How do we ensure that everything we build at smaller scale can still work at larger scale... and with devices of vastly different types?” he said.
Another key area will be “multimodal” sensing, in which one uses diverse sensors to overcome adversaries’ deception tactics. For example, if a visual sensor is tricked into not recognizing a tank’s approach, a vibration sensor might still detect it.
Finally, the work will advance adaptation to rapidly changing, dynamic environments. “When things change very quickly, you need to constantly anticipate the next move and not be stuck reacting to the past. Decision cycles have become a lot shorter, and whoever has the shorter decision cycle wins,” says Abdelzaher.
IoBT REIGN is a collaboration led by UIUC and includes UCLA, the University of Massachusetts Amherst, the University of Southern California, and SRI International.
Read the original story from the Coordinated Science Laboratory.