How U.S. Navy Plans To Foil Massive ‘Super Swarm’ Drone Attacks

0
7
How U.S. Navy Plans To Foil Massive ‘Super Swarm’ Drone Attacks
Click here to view original web page at www.forbes.com

The Navy is developing offensive and defensive tactics for “Super Swarms” of up to a million drones.

While million-drone swarms may be many years away, attacks with hundreds or thousands of dronea are much closer and “ large-scale adversarial swarms” are already an “imminent threat ” according to Prof Isaac […]


The Navy is developing offensive and defensive tactics for “Super Swarms” of up to a million drones.

While million-drone swarms may be many years away, attacks with hundreds or thousands of dronea are much closer and “large-scale adversarial swarms” are already an “imminent threat” according to Prof Isaac Kaminer of the US Naval Postgraduate School, an expert in the subject of swarming and counter swarming tactics. Kaminer's work suggests that stopping a swarm is not simply a matter of throwing enough missiles or bullets at it; instead, the swarm has to be outsmarted. And his team are setting out to discover how a swarm’s intelligence can be used against it.

Counter swarm strategy diagram
Countering Super Swarms of thousands of drone requires a reliable strategy, not just more guns

In 2016 Kaminer was working on developing tactics to protect a High Value Naval Asset (an aircraft carrier) from a swarm of small unmanned boats. It’s s topical issue, as Iran’s Revolutionary Guard has long worked on swarming tactics for small boats against large naval vessels. More recently, Houthi rebels with Iranian assistance have deployed remote-controlled boats packed with explosives against tankers and other targets, making kamikaze boat swarms a more immediate threat. It soon became apparent such an attack would not happen in isolation: the robot boats could be supplemented with swarms of small unmanned aircraft and unmanned submarines.

In Kaminer's defintion, a ‘Super Swarm’ is one which has overwhelming numbers and can include multiple modes – air, surface and subsurface threats. Future unmanned craft may even switch between modes, like the U.S. Navy’s own Flimmer and Flying Sea Glider prototypes which could fly to a target area and dive into the water to become submarines, or approach underwater before executing a pop-up attack.

Super Swarms represent both an opportunity and a threat for the U.S. Navy, and Kaminer’s work covers both defensive and offensive operations. The Navy is already a leader in offensive swarm operations, with its LOCUST drone swarm developed by Raytheon. The key for both is how the swarm is controlled. Current drones are piloted remotely by humans; this becomes impossible with more than a few drones, both due to the demand for personnel and bandwidth restrictions. Instead, the swarm will need to control itself.

“A swarm with 10,000 or more drones must have extremely high levels of autonomy,” says consultant Zak Kallenborn. “No human being could handle the amount of information necessary to make decisions.“

Most solutions to the challenge involve swarming algorithms which work in ways similar to flocks or birds of swarms of insects seen in nature. If every member of the swarm follows the same few simple rules, the swarm can maintain cohesion without units colliding with each other. Similar rules allows insects colonies, from mound-building termites to foraging bees, to work effectively as a team without any central controlling intelligence.

A flock of starlings
A mumurmation or flock of starlings; thousands of birds take off, fly together in changing ... [+]

The swarm can be defeated by taking advantage of its internal rules – if these can be figured out. For example, an entire swarm whose members all have a collision-avoidance rule can be ‘herded’ by a few outsider drones, or may be fooled into running into each other. If the members of the swarm are all programmed to attack what they see as the highest-value target in range, then they can all be decoyed into attacking the same dummy.

“A targeting error could generate cascading errors throughout the entire swarm,” says Kallenborn.

However the challenge is figuring out the algorithms that govern an attacking swarm about which nothing is known. Kaminer’s previous work on such "xSwarms" has explored how to figure out the internals of a swarm’s operation by observing its movement and how it reacts to intruders. he believes sending in agent provocateur drones provokes reactions which can be analyzed and exploited.

The Navy has also laid the groundwork for swarm vs swarm tactics, in a study on Air Superiority via Decentralized Swarming Tactics and Autonomous Pursuit employing a fleet of small drones to develop tactics against multiple attackers. This built on the earlier Service Academies Swarm Challenge hosted by DARPA in 2017 in which three teams competed in an aerial version of Capture the Flag with swarms of 25 drones on each side. Perhaps unsurprisingly given their history in this area, the U.S. Naval Academy won that one.

The biggest challenge internationally appears to come from China, which is developing swarming capability as a means of asymmetric warfare, in particular to neutralize the U.S advantage in aircraft carriers. A lineup of Chinse combat drone systems spotted by satellite last year included not only various large drones such as the Sharp Sword stealth drone and the Wing Loong Reaper-alike, but also at least two groups of smaller swarming drones. No details are known about these swarms.

A future Super Swarm attack on a naval force is likely to take place at speeds that no human can follow, with attacking and defending forces attempting to work out each other’s algorithms, exploit and outmaneuver them in real time, over, on and under the water simultaneously. The question is what are the best tactics in such combats, and what matters more out of speed, agility, firepower, brains, or sheer numbers? That’s the sort of crucial question that the latest phase of Navy research will address.

This type of robot-wars future is not inevitable. Kallenborn has previous argued that some types of drone swarm under development have sufficient destructive power to count as Weapons of Mass Destruction and could be limited by international law. And this would include the type of Super Swarms the Navy is studying.

"A swarm with 10,000 or more armed drones absolutely should be classified as a weapon of mass destruction," says Kallenborn.

While existing laws might go some way to controlling such weapons, the legal framework needs strengthening. In particular, Kallenborn believes the U.S. needs to formally take the position that large swarms of autonomous lethal drones should be considered WMD, and current international discussions on lethal robots need to move forward.

"The opportunity to develop global norms and treaties around drone swarms and other autonomous weapons is now, “ says Kallenborn. “Collective limits on the number of armed drones in a swarm would reduce the risk to civilians and national security.”

Commentators are still chewing over the significance of last week’s human-versus-AI AlphaDogfight result. However, if Kaminer and Kallenborn are right about the potential of swarms, such dogfights will soon look as antiquated as duels in biplanes.