Bringing unmanned systems into service for the U.S. military increasingly requires integrating complex hardware, software, algorithms and other elements from different vendors and developers. Five experts on the challenges of safe, effective integration offered their insights Wednesday in a panel discussion called, “Advancing Autonomy Through Platform Integration.”

Moderator Kevin Hagan, unmanned systems account executive for Peraton, opened the discussion with a description of its scope.

 “I think that most would agree that autonomous systems are the future of military systems,” he said. “How do we ensure integrated systems work as designed and in a safe manner when faced with all the variables of a real world environment? How do we test those systems, and how do we build trust in them?”

Among other topics, panelists discussed the value of standardization and modularity in elements of unmanned systems, a rigorous certification process, and failure tracing through explainable AI.

“A big buzzword with the Department [of Defense] is modular open systems,” noted Luis Molina, deputy director of the department’s Office of Strategic Capabilities. He advised developers to adopt common architectures, as opposed to creating new ones. “The DoD wants to avoid paying for a new controller every time you get a new toy.”

Panelist Robert Simmons, former assistant program manager for underwater systems in the Navy’s expeditionary  missions program office and currently an analyst for Grey Hawk Strategic, spoke about the importance of a rigorous certification process to ensure that new system components and algorithms are secure and that they can hold up in sometimes unexpected environments.

“It’s very important that if the UAV or the autonomous system that is in operation falls into the wrong hands, or goes to an area in which it is unintended, that there is self-protection in the system,” Simmons said.

Panelist Daniel Cook, program office in the Office of Naval Research, pointed out that while much attention is focused on the front-end work to ensure a component of an integrated system will succeed, there is also great value in building in ways to ensure failures can be examined. An area of technology devoted to that is explainable AI, or XAI, which incorporate means for breaking down for humans how a decision by a machine was made.

“It’s important for us to be able to trace that failure,” Cook said, “and then use that knowledge to base future certifications on.”

Panelist Daniel Kucik tackled a question about development techniques to incorporate to give autonomous systems and supporting elements the best possible chance of actually being used in the field. Kucik serves as distinguished engineer for mine warfore, unmanned systems, at the Naval Service Warfare Center, Panama City Division.

He advised trying to deploy new capabilities incrementally, allowing time for end users to gain confidence in each, and building training and opportunities for feedback into implementation plans.

“If you jam too much down their throats at once … and they’ve never seen an unmanned system before, they’re probably not going to trust it,” Kucik said. “So what you want to do is incrementally field capabilities as they become available from a technical standpoint, and ensure there’s a feedback loop that includes the end user, the developers and other stakeholders because you want to be able to build on experiences as you accumulate them.”