WASHINGTON — A computational neuroscientist is studying whether a dragonfly’s excellent hunting skills can be replicated in a missile’s ability to maneuver and destroy targets midair with better precision.
Dragonflies are vicious little creatures with a hit-to-kill track record of 95 percent, meaning only 5 percent of its prey escapes.
Sandia National Laboratories’ Frances Chance is building algorithms that simulate how a dragonfly processes information when intercepting prey, and she’s testing them in a virtual environment. So far, the results are promising.
The laboratories are federally funded and focus on national security missions through scientific and engineering research. The project is a yearlong, high-risk, high-gain effort that will wrap up in September, and it is funded by Sandia’s Autonomy for Hypersonics Mission Campaign, Chance said.
“I think what is really interesting about insects, in general, is they do something really fast and really well, but they are not particularly smart in the way you or I would think of ourselves as being smart,” Chance told Defense News in a recent interview.
While insects may not be the right fit for studying cognitive capabilities to develop complex artificial intelligence, they are ideal for developing efficient computations for intercept capability. A dragonfly can react to a particular prey’s maneuvers in 50 milliseconds, Chance explained. That amount of time accounts for information to cross three neurons in a dragonfly’s brain. This indicates the dragonfly doesn’t learn how to hunt, but rather the skill is inherent and part of its brain’s hard-wiring.
“The challenge then is: Is there anything that we can learn from how dragonflies do this that we can then bring to the next generation of missiles, or maybe even the next-next generation of missiles?” Chance said.
By developing an artificial neural network that mimics a dragonfly’s ability to hunt and then applying it to missile capabilities that rely on computation-heavy systems, one could reduce the size, weight and power needed for a missile’s onboard computers; improve intercept techniques for targets such as hypersonic weapons; and home in on targets using simpler sensors.
If the model of a dragonfly’s neural circuit developed through Chance’s research shows enough promise, she would then pass the information to scientists, who would try to directly apply it to weapons systems.
One of the greatest leaps involves adapting an algorithm to handle the speed at which a missile flies. While a dragonfly is fast, it’s not nearly as fast as a missile. Animal brains process information significantly slower than a computer, so it’s possible computations can be sped up to better align with the speed at which a missile approaches targets.
“The hope is that even if the algorithm isn’t wildly successful, you might be able to say something about what you can get away with in terms of what types of capabilities you give the next generation of weapons,” Chance said.
The model she’s building is several steps removed from implementation onto a weapon. “I would consider the project complete when we have a viable model — ‘viable’ meaning it does interception — and a bonus if it’s neurobiologically plausible. There is no reason to force that for this type of research, but only because it doesn’t necessarily matter; so something biologically inspired that works I would consider a success.”
Jen Judson is the land warfare reporter for Defense News. She has covered defense in the Washington area for 10 years. She was previously a reporter at Politico and Inside Defense. She won the National Press Club's best analytical reporting award in 2014 and was named the Defense Media Awards' best young defense journalist in 2018.