Interesting article here from 2016 but probably still...

Interesting article here from 2016 but probably still applicable...

"U.S. Navy ships are equipped with either the Ship Self-Defense System (SSDS) or Aegis Combat System to defend against airborne missile attacks. Neither system is designed to defeat small, maneuverable drones. That leaves the close-in weapon system (CIWS) as the lone defender of the ship. CIWS identifies, tracks, and neutralizes close airborne contacts that have penetrated the ship’s long-range defenses. If the swarm of attacking drones were numerous enough, it is likely that a few could slip past the CIWS defense bubble. And a few is all it would take to disable a ship or seriously interrupt operations. For instance, drones could be “trained” to hunt and kill — with shrapnel or explosively formed penetrators — the radar, flight deck, bridge, and communications antennae. Even if only one of these targets were hit, the ship would be taken out of the game.

The Navy is in the process of developing lasers to protect against drone attacks. However, these lasers are not likely to enter the fleet until 2020 and have only proven their effectiveness against slow moving, single drones, rather than maneuverable swarms. Current methods are not viable for the immediate threat of non-state weaponized drones. If non-state actors are able to present a clear danger to U.S. ships operating in littoral waters, then the U.S. Navy will have to change its decision calculus. The Navy already faces a shortage of ships and a tight budget. On top of this, a viable drone risk to ships operating close enough to a coast could make presence missions unpalatable."

https://warontherocks.com/2016/10/the-navy-litorally-has-a-drone-problem/

Also article proposing drones as countermeasures with interesting references to AI towards the end of article:

"However, in the past year, the Defense Advanced Research Projects Agency has moved toward an interesting middle-ground: SyNAPSE (Systems of Neuromorphic Adaptive Plastic Scalable Electronics). The project seeks the middle ground between biological neurological processes and computers through biomimicry: creating a neuromorphic device—the functioning structure of a mammalian brain—out of artificial materials.

Processing is no good without recognition. Google software engineer Quoc V. Le announced to the San Francisco Machine Learning Conference that Google’s Deep Learning clusters had learned to recognize the basic aspects of certain objects. “He realized the deep learning clusters had made a breakthrough when they were able to recognize discrete workplace objects. . . . Le didn’t train the machines this way—the software had figured that out on its own.”6 Though Google’s deep-learning processes are orders of magnitude larger than what is available to a drone, those “lessons” (on, say, the shape of friendly/enemy warships and aircraft) might be imparted to a cadre of drones designed to penetrate a countermeasure swarm.

Brigham Young University has also developed a genetic “smart object recognition algorithm” able to recognize images from photos and video without human calibration. With the computer choosing the important aspects of selected identifying images, BYU engineer Dah-Jye Lee and his students found 100 percent accurate recognition in four of CalTech datasets used (motorbikes, faces, airplanes, and automobiles). The other published well-performing object recognition systems scored in the 95–98 percent range.

7Unmanned and manned offensive platforms can likely both work together, furthering the complementary combination of arms. Autonomous strike drones could specialize in easier-to-recognize and harder-to-reach conventional targets at sea while serving as escorts for manned strike missions inland or more complicated conflict operations."

https://www.usni.org/magazines/proceedings/2014/february/bring-countermeasure-drones