Isegoria

In contrast to the DASH, which started out as a combat aircraft and ended as a target, David Hambling explains (in Swarm Troopers), the Teledyne Ryan Firebee started out as a target and ended as much more:

The Firebee was a sleek, jet-powered machine, twenty-three feet long and with a top speed of over 700 mph. It could fly at any height from the treetops to fifty thousand feet. It could be launched from an aircraft and remotely controlled from two hundred miles away. The Firebee would return to the ground on a parachute, an easy feat for a small plane with no human inside risking broken bones.

There was little interest from the Air Force’s mainstream, but the highly unconventional BIG SAFARI team liked the idea. BIG SAFARI was set up to circumvent the usual complexities of Air Force procurement, to provide quick solutions to urgent problems. They funded development of a version of the Firebee called Fire Fly or Model 147, and it went through their streamlined channels without the interference it might have otherwise endured.

[…]

In the first trials the F-102 Delta Dagger and F-106 Delta Dart pilots never even saw the drones they were trying to shoot down, and only caught brief glimpses of them on radar. Further tests followed. In one, a Delta Dagger fired a burst of cannon fire at the drone, but the rounds missed. Before the pilot could line up for another shot, his jet engine flamed out because of the high altitude. He dropped to lower altitude to reignite the engine, at which point other planes mistook his aircraft for the target. Fortunately, they did not shoot, but the Fire Fly had escaped. Later on two Delta Darts achieved a radar lock on the Fire Fly, but not for long enough to fire a missile.

[…]

The military was unhappy with the results. Many felt the test was intended to make them look bad. Robert Schwanhausser of Teledyne Ryan says the results were classified Top Secret, and he was ordered to burn every piece of information on them.

[…]

They were sent on virtual suicide missions, to test Vietnamese radar and missile defenses.

When losses mounted, the developers at BIG SAFARI started equipping their drones with electronic bags of tricks. One device, known as High Altitude Threat Reaction and Countermeasure (HAT-RAC) responded to being lit up by radar by throwing the drone into a series of sharp turns.

[…]

When the Chinese downed their first Fire Fly in 1964, it was only after some sixteen MiGs had made over thirty passes trying to hit the little drone.

[…]

A decoy version of the Fire Fly was produced. This was known as the 147N and was fitted with radar reflectors to make it look like a bigger aircraft. The 147Ns were originally purely intended to distract defenders away from the real Fire Flies equipped with cameras, but they survived and managed to return so frequently that they were later fitted with cameras of their own.

[…]

On one mission, the pictures from a Fire Fly captured the subject’s faces from close range: “You could see features on the guy’s face. If it would have been in color, you could have seen the color of his eyes.”

This was at a time when the U-2 spy planes were taking pictures from fifty thousand feet or higher, with resolution only good enough to recognize objects two feet across. The low-level Fire Fly pictures were a revelation in the art of the possible.

[…]

The basic drone could only handle acceleration of about 3G, but a modified Firebee equipped with “Maneuverability Augmentation System for Tactical Air Combat Simulation” or MASTACS could pull 6G for several seconds at a time. This put it pretty much on a par with manned fighters. In 1971, the MASTACS developers challenged Commander John C. Smith, head of the Navy’s Top Gun combat training school – the “Top Gun” of the 1982 movie – to try and shoot MASTACS down.

Smith and his wingman, both flying F-4 Phantoms, made repeated attacks on the remotely controlled Firebee. It was far too agile for them. They fired two Sparrow radar-guided missiles and two Sidewinder heat-seekers without scoring a hit. Meanwhile, the Firebee kept circling around and lining itself up in firing position behind the Phantoms. Had it been armed, the Firebee would have had easy shots.

As usual, nobody liked a smart robot. MASTACS was deemed “too sophisticated” for training purposes.

[…]

Even the memory of the Fire Fly seems to have been lost. In 2014 the US Navy proudly announced in a press release that, “Truman will be the first aircraft carrier in naval aviation history to host test operations for an unmanned aircraft.” It seems that amnesia buried the 1969-70 Fire Fly operations from the USS Ranger, not to mention the TDR-1s flown from the USS Sable in 1943.

Since its founding in 2017, Anduril has argued that it’s a new type of defense contractor:

Instead of taking orders upfront from the US Department of Defense to fund development of products, Anduril has raised money from venture capitalists, including Peter Thiel’s Founders Fund, that it uses to build weapons it predicts the military will want. Its first product was an automated security tower designed for the US border in the early days of Donald Trump’s presidency. The company then began shipping early counter-drone aircraft to the US and UK militaries in 2019.

[…]

Anduril started work two years ago on the Roadrunner, a Looney Tunes-inspired dig at Raytheon’s Coyote, because it said the US would need a lower-cost, more nimble way to combat swarms. The tiny fighter jet has a carbon-fiber body and onboard electronics that let it track objects and perform maneuvers that’d be too dangerous for a human-piloted plane. One of its main advantages is that it can be reused, which makes it easier to launch at the first sign of an unknown object. “If you see a threat, you can launch multiple Roadrunners to go out to do a closer inspection of that threat and be loitering in case they’re needed,” says Christian Brose, the chief strategy officer at Anduril. “You can recall them, land them, refuel them and reuse them, so, essentially, you can launch without regret.”

[…]

To start the test, Anduril sent a fixed-wing drone into the air from a runway behind its compound. The sentry tower quickly detected the aircraft and fed information about its speed and trajectory into the company’s Lattice software. The test pilot received imagery of the drone and then manually marked it as a hostile threat. In an instant, the lid of the Roadrunner launch container opened, the turbines fired up and the craft zipped into the air. It took off toward the target and then began feeding its own sensor data and imagery into Lattice. As the Roadrunner closed in on the target, the test pilot gave a final command to destroy the fixed-wing craft, and, seconds later, the Lattice software displayed information showing that it had been a successful attack.

For the purposes of this demonstration, Anduril used proximity sensors to confirm that it would have taken out the target and didn’t actually blow up the fixed-wing craft. If it had, the Roadrunner wouldn’t have been able to do what it did next: It turned to fly back toward the Anduril compound, shifted into a vertical position and fired its thrusters toward the ground as landing legs kicked out from its side. During a maneuver lasting about a minute, the machine got ever closer to the ground before finally settling gently on a small concrete pad in a fashion very similar to a Space Exploration Technologies Corp. rocket. A future version of the Roadrunner will be able to land even after destroying a target, Luckey says.

The whole idea, as Anduril sees it, is to allow a single operator to manage dozens or more Roadrunners in the field with Lattice providing a full view of the surroundings, targets and weapons available. If a drone swarm approaches a base, Lattice will quickly see and identify all the drones, and, with a couple of clicks, the operator can send Roadrunners off to combat the threat. This is a major change from many of the other counter-drone weapons that require about a dozen people to operate them.

Anduril has raised $2.7 billion to date and is valued at almost $10 billion.

In the late 1950s, David Hambling notes (in Swarm Troopers), the latest sonar could detect a submarine more than twenty miles away, but the best anti-submarine weapons only had a range of a few miles:

The US Navy wanted to bridge the gap with a Drone Anti-Submarine Helicopter or DASH. This was a small helicopter capable of carrying a single weapon and dropping it at the required spot, guided by a controller back on board ship.

The DASH was based on a one-man helicopter called a “Rotorcycle” built by Gyrodyne Company. This had two rotor blades rotating in opposite directions for lift, and a propeller for forward motion. The drone version was the size of a small car and weighed just over a ton. By 1963, the US Navy had eighty of them.

[…]

DASH was designed to be expendable; when it dropped a Mk57 nuclear depth charge it would be within the lethal radius of the resulting explosion. The powerful warhead, from five to twenty kilotons, guaranteed that the sub would be destroyed, and losing one drone for one submarine was a good exchange rate. The idea that DASH should carry a non-nuclear homing torpedo and come back afterwards was a case of mission creep; according to the original design it was only supposed to make one flight.

[…]

Executive Officer Phil King of the USS Blue modified a DASH, adding a television camera for reconnaissance and gunnery direction. Known as SNOOPY missions, these involved the DASH flying out to find targets. The operator identified them via the television link, and the destroyer then opened up with its battery of five-inch guns. The drone operator could see where the shells were landing and tell the gunners how to adjust their aim.

Further developments followed, including NITE PANTHER and BLOW LOW versions equipped with additional fuel tanks for longer range, night-vision systems and airborne radar.

The next logical step was to convert the DASH from finding targets to attacking them. NITE GAZELLE, GUN SHIP, and ATTACK DRONE were all individual modified aircraft with a range of weaponry including a six-barreled minigun firing four thousand rounds a minute, grenade launchers, bomblet dispensers and bombs, as well as a laser designator for directing smart bombs. The idea was that drones with guns would deal with the ground defenses, leaving the way clear for the bomber drones to hit targets with pinpoint accuracy.

[…]

“It became quite evident that the Navy no longer wanted DASH and wanted to move onto LAMPS manned helicopters.”

LAMPS was the Light Airborne Multipurpose System, a new manned helicopter that would operate from destroyers and take over the role of DASH. Removing DASH from the picture meant there would be no competition, and nobody would be able to argue that LAMPS was unnecessary.

[…]

The LAMPS project became the SH-60 Sea Hawk, now a multibillion dollar success story.

The GBU-28 is a 5,000-pound laser-guided “bunker busting” bomb:

It was designed, manufactured, and deployed in less than three weeks due to an urgent need during Operation Desert Storm to penetrate hardened Iraqi command centers located deep underground.

[…]

The GBU-28 is unique in that time between the finalized design being approved to its first use in combat test took only two weeks between the 13th and 27th of February 1991.

The name apparently refers to the fact that this Guided Bomb Unit was designed, built, and ready to drop in four weeks:

The initial batch of GBU-28s was built from modified 8 inch/203 mm artillery barrels (principally from deactivated M110 howitzers), but later examples are purpose-built with the BLU-113 bomb body made by National Forge of Irvine, Pennsylvania. They weigh 5,000 pounds (2,268 kg) and contain 630 pounds (286 kg) of Tritonal explosive.

[…]

It proved capable of penetrating over 50 meters (164 ft) of earth or 5 meters (16 ft) of solid concrete; this was demonstrated when a test bomb, bolted to a missile sled, smashed through 22 ft (6.7 m) of reinforced concrete and still retained enough kinetic energy to travel a half-mile downrange.

It looks more like a missile than a bomb:

America’s first attack drones date back to World War 2, David Hambling notes (in Swarm Troopers):

Lt. Commander Delmar Farhney worked with the US Naval Research Laboratory in the 1930s building radio-controlled anti-aircraft targets for the Navy. It was an exciting era to be working with radio, and Farhney was convinced that unmanned aircraft would be devastatingly effective. By 1941, he had extended his work to aircraft capable of accurately dropping torpedoes and depth charges. Incidentally, Farhney was the first to officially refer to his aircraft as “drones,” a usage the military has since tried to suppress.

[…]

He could not use metal, so the TDN-1 was made from plywood.

[…]

Some of the work was carried out by organ makers Wurlitzer, with their long experience at shaping plywood. The TDR-1 had a wingspan of forty-eight feet, a speed of almost a hundred and fifty miles per hour, and awkward tricycle landing gear to give space for a 2,000-pound bomb or torpedo slung beneath the fuselage

As an airplane, the TDR-1 was unremarkable, but it was equipped with a remarkable technological breakthrough: remote control by television. Dr. Vladimir Zworykin of RCA was one of the inventors of the television, and he was keen to put it to use in drones. The prototype cameras weighed over three hundred pounds including the transmitter, but this was shrunk into a miniature system weighing ninety-seven pounds, packed into a box the size of a carry-on suitcase. The picture was monochrome with a respectable resolution (350 lines) and a refresh rate of forty hertz, but the image was poor by modern standards. The drone operator had to work under a black cloth to see the green, five-inch screen clearly in daylight.

[…]

It was controlled from a modified Avenger torpedo bomber flying up to eight miles away. The special Avenger had a crew of four, with pilot, radio operator, and gunner joined by a drone operator. The latter had a joystick, a television screen, and a rotary telephone dial. The dial controlled altitude and released weapons by dialing specific numbers, and the television gave a real sense of being in the drone.

[…]

The drones were eventually allowed to attack a derelict Japanese freighter called Yamazuki Maru off Guadalcanal. Three out of four drones hit the target, and, after some hesitation, the unit was sent into action.

The STAG-1 drones successfully attacked anti-aircraft sites, gun positions, ships, and even a lighthouse. Many of them were used in suicide attacks against challenging targets; the Japanese, not knowing they were unmanned, called them “American kamikazes.”

[…]

The commander of the STAG-1, Lt Commander Robert Jones, was convinced that their successes would prove the value of the drone concept. He believed drones would be an important weapon in the assault on mainland Japan. But the Navy top brass did not agree. The drones might be good for precision attacks, but what were needed were formations of heavy bombers. After the drones were all expended, STAG-1 was reassigned. Commander Jones watched unhappily as the thirty Avenger control planes were dumped overboard in Reynard Sound.

The television technology was a limitation, as anyone who has worked with monochrome images can appreciate. Targets that had a clear silhouette, like a ship on the water, showed up clearly and were easy to hit. But any target surrounded by jungle tended to be invisible on the small screen, as they blended in with the confused background.

(Meanwhile, experiments with larger radio-controlled aircraft as suicide bombers against major targets had limited success. In the most famous disaster, Lt Joseph Kennedy Junior was killed when a “robot” PB4Y-1 bomber blew up prematurely in 1944. This left his younger brother John F Kennedy as the family heir.)

Both Farhney and Jones continued the struggle to get drones recognised, and during the Korean War, unmanned attack planes were tried again. In 1952, six obsolete F6F Hellcats were converted to unmanned operation. They were controlled from nearby AD-2Q Skyraiders, with a television system developed from the one on the TDR-1. Flying from the aircraft carrier USS Boxer, the drones successfully hit a power plant, a railway tunnel, and a bridge. Jones wanted to continue operations and attack the Yalu River bridges, which had survived repeated attacks by US heavy bombers.

Farhney went on to become a Rear Admiral and headed the Navy’s guided missile research effort — and in the 1950s he made a number of public statements about UFOs, which he believed to be craft of extraterrestrial origin.

David Hambling opens Swarm Troopers with a history of drones — which is not a history of steady progress:

History shows that drones tend to be ruthlessly terminated by a military establishment that harbors a ferocious antipathy to anything that dares to compete with manned aircraft.

[…]

The cavalry officers who could not see the advantages of switching to motor vehicles at the start of the twentieth century were making a rational assessment based on the available evidence. Horses had centuries of successful service, while their motorized replacements had always been clumsy and unreliable, especially on rough going.

[…]

Even when they have proven successful, nobody wants them when the war is over: “The great broom of victory swept all new projects into the ashcan of forgotten dreams,” as Commander Delmar Farhney put it.

[…]

Drone prehistory goes back to 1849 and the use of bomb-dropping balloons in the siege of Venice. These were devised by the ingenious Lieutenant Uchatius of the Austrian army. It was not possible to bring siege artillery close enough to the city. Uchatius, better known to history as a photographic pioneer, rigged up hot-air balloons to release small bombs by remote control via a copper wire. About twenty of the balloons were launched. Austrian news reports suggested the bombs would turn Venice into rubble, but it seems that only one or two hit the city. The rest fell into the waters of the Venetian Lido or outside the city entirely.

[…]

In some ways the true inventor of unmanned warfare was Nikola Tesla, who demonstrated a miniature boat controlled by radio waves at Madison Square Garden in 1898. Tesla believed that a version armed with torpedoes could sink battleships and lead to a new age in which wars were fought between machines with no human combatants. As with many of his projects, Tesla never developed the idea beyond the initial demonstration.

Both Britain and the US developed their own drone aircraft in WWI. The British effort was headed by “Professor” Archibald Low — he used the title even though he was not actually a university professor. Low had a tremendous enthusiasm for remote control and was repeatedly distracted from one project by another. His project was known as “AT,” short for Aerial Target, a designation to mislead the enemy into thinking the device was simply a target for anti-aircraft practice. This name proved to be strangely prophetic.

The AT was a wooden biplane with a fourteen-foot wingspan and an explosive warhead, intended for use against both ground targets and zeppelins. It flew well initially, but the program was terminated after an unfortunate incident in 1917 when it was being demonstrated to a group of generals.

[…]

The first AT to be launched in the demonstration suffered engine failure during take-off and flopped into the ground. A Major Bell delivered the verdict quoted at the head of this chapter: “I could throw my bloody umbrella further than that!”

The second machine fared even worse. The operator lost control and the AT flew right at the audience, scattering them before veering off and crashing a few feet away. As a demonstration of controlled flight, it was unconvincing. Nobody present can have thought that it was advisable to put high explosives on drones.

[…]

By 1918, Sperry’s Aerial Torpedo was able to fly along a preset route and dive on a target, delivering a thousand-pound bomb or releasing a torpedo. The weapon was too late to be used in action, and after the war, the US Navy thought drones were useful only for gunnery practice.

[…]

Even when relegated to the ignominious role of targets, drones developed a knack for embarrassing humans. In the 1930s, the British Air Ministry decided to test the claim that battleships were vulnerable to air attack by flying a radio-controlled Fairey Queen target plane against the British Mediterranean fleet. After more than two hours of sustained anti-aircraft fire and numerous passes, the drone was undamaged.

The Royal Navy accepted that its air defenses needed upgrading. Large numbers of drones were built as a result — but only as targets. Nobody thought the test showed that an unmanned aircraft might be an effective weapon.

When it became clear that drones were playing a significant role in Ukraine, I decided to finally catch up on the topic, and I noticed that David Hambling, whose articles seemed reasonable and well written, had written a book on the topic back in 2013, Swarm Troopers: How small drones will conquer the world.

One of the first points he makes in the book is that the Pentagon used to always be 20 years ahead of the private sector:

Smartphone sales have accelerated from zero in 2006 to over a billion smartphones shipped in 2013.

[…]

Billions of dollars are spent annually on advancing technology just for small electronic devices.

[…]

These days soldiers are less likely to be awestruck at the gadgetry they are issued than shocked by how clunky it is compared to the sleek lightweight devices they have at home.

[…]

Selling to the military means extensive testing and certification, with the related delays and costs. Add to this a military bureaucracy that can take years to agree on the specification it wants in the first place, overseen by a political leadership that may cancel, delay, or divert any project depending on the shifting sands of expediency, and you have a recipe for a long time between generations.

Each generation of electronics roughly translates to a doubling of processing power, memory, pixels, or other relevant metrics. If a commercial product goes through a generation every two years, and the military cycle takes six years per generation, then in twelve years the military product goes from being four times as powerful as the competition to a quarter as powerful.

[…]

A drone is simply a smartphone with wings, and the wings are the cheap part.

The Lancet loitering munition is a standout success for Russia:

While other weapons have performed below expectation during the invasion of Ukraine, this 35-pound kamikaze drone has proven capable of taking out a wide range of targets, including main battle tanks and parked aircraft, from far over the horizon.

[…]

After being used on a trial basis in Syria in 2021, the Lancet was rushed into full-scale service for this conflict. The first known use in Ukraine was in July 2022, some five months into the invasion. Since then it has been used in small but growing numbers.

[…]

At first only a handful of Lancet strike videos were posted each month. But this January, 22 Lancet attack videos appeared. That number rose to 62 in May, and 124 in August. The makers claim they are mass-producing the weapon at a new facility, so what we are seeing now is only the start. This growth in production is taking place despite the fact that the Lancet uses Western-made electronics, which in theory should be impossible for Russia to obtain.

[…]

The Lancet is launched from a catapult rail and transmits video back to the operator. Lancets are reportedly flown in conjunction with reconnaissance drones which spot targets and relay coordinates. The Lancet operator flies to the target area, visually confirms the target, and carries out the strike.

An electric propeller drives the Lancet at around 70 miles per hour. This slow speed makes it an easier target than a guided missile or other munition.

“Every day we shoot down at least one or two of these Lancets,” Yuriy Sak, an adviser to Ukraine’s defense minister, told Reuters. “But it’s not a 100 percent interception rate, unfortunately.”

Early Lancet attacks were all on static targets. More recent videos have shown hits on moving vehicles. This may indicate a change in doctrine or an improvement in operator skill levels.

[…]

According to Lost Armor, as of Oct. 3 there are 667 Lancet strike videos. Of these, 210 are classed as target destroyed (31%), 355 target damaged (53%), 48 miss (7%), and 52 are unknown (7%) . In particular, the heavy armor of tanks sometimes shrugs off the Lancet’s relatively small warhead.

This suggests that around 2,000 Lancets have destroyed 200 targets and damaged hundreds. That may seem low, but with each Lancet costing perhaps $35,000 and each target costing millions, the Lancet is extremely cost-effective.

[…]

By far the largest number of Lancet strike videos show attacks on Ukrainian artillery, both towed and self-propelled guns. As a recent report from UK defense think tank RUSI notes, Russian forces now use the Lancet extensively as a counter-battery weapon. Artillery is the traditional means of striking enemy artillery, but the long range of the Lancet, and its ability to seek out hidden targets on the ground, give it real advantages. Additionally, the Lancet operator remains hidden and will not be targeted by counter-battery fire.

[…]

Towed artillery is much harder to destroy than a self-propelled gun, even when hit. The latter is a tracked vehicle with a store of flammable fuel and explosive ammunition on board, either of which can be set off by a Lancet strike. A towed artillery piece, by contrast, is a more solid piece of machinery able to survive the blast and minor shrapnel fragments of a Lancet hit.

“The lethality of Lancet is often insufficient,” according to the RUSI report. “One officer also said that although he had seen his gun ‘destroyed’ several times online, it remained alive and well.”

This tallies with previous conflicts in which towed artillery has proven more robust to counter-battery fire. Crews may be injured or killed, but the guns themselves tend to survive and remain serviceable. In WWII, the loss rate for self-propelled guns was two to three times higher than for towed artillery. So many of the Lancet hits on towed artillery likely did not result in kills.

Both drones and thermal imagers have been game changers in the Ukraine conflict, but fitting a thermal imager to a drone is not so simple:

These days high-end drones, like smartphones, have high quality video: and it is possible to shoot impressive 4K video at 60 frames per second from a drone that fits in your pocket. 2.7k and 1080p video are routine on lower-cost models. But thermal imagers are many years behind video cameras, and resolutions are much lower.

You can get a low-cost thermal imager like the Seek Thermal Compact for under $200, but the resolution is only 206 x 156 pixels – fine for checking insulation and finding leaks around the house, but no good for seeing objects hundreds of meters away. Going up to 320 x 240 will double the price, but you will still struggle to tell whether you are looking at a truck or a tank. Part of the problem is that while a video camera can show differences in brightness and color, a thermal image is monochrome and only shows temperature. The details which help identify objects visually may be missing, an issue highlighted by how difficult it is to recognize faces via thermal imaging.

When discussing the issue of thermal imager on reconnaissance drones, an expert from Ukraine’s Aerorozvidka drone unit noted on social media that a Matrice drone with a thermal imager costing several thousand dollars could only detect Russian vehicles at 3-4 miles distance and even then distinguishing types was difficult. The daylight camera could pick out targets from 15 miles. They suggested spending the money on more batteries and an additional ground control unit as a better way of boosting the drone’s usefulness.

This applies even more so with FPV drones. The drone flies at high speed and requires a skilled pilot to avoid obstacles and successfully hit the target, so good quality video with a rapid refresh rate, and cheap thermal imagers will not do the job.

[…]

“Ukrainian manufacturers also have all these technologies and can produce FPV drones with thermal imaging cameras, but the main problem is the price,” an Escadrone spokesman told Forbes. “If a regular FPV drone costs $500, then the same drone with a thermal imaging camera will cost about $2,500.”

[…]

This type of issue highlights the difference between military-grade loitering munitions like the U.S.-made SwitchBlade 300. This is similar in size to an FPV drone and has daylight and thermal imaging, plus a lock-on-to-target function and numerous other features, but costs around $50,000 per shot.

Larger, reusable drones costing in the tens of thousands of dollars make far more sense for thermal imagers.

An AI racing drone recently beat human pilots, raising the question of when AI drones will transform warfare:

“The AI is superhuman because it discovers and flies the best maneuvers, also it is consistent and precise, which humans are not,” says Scaramuzza. He notes that, as with AlphaGo, Swift was able to use moves — in this case flight trajectories — which the human champions did not even think were possible.

[…]

A $400 FPV with the warhead from an RPG rocket launcher can knock out a tank, personnel carrier or artillery piece from several miles away, or chase down and destroy a truck traveling at high speed. They are cheap enough to use against individual footsoldiers and can dive into trenches. But it requires a skilled human pilot. Ukrainian sources say the training takes around a month to achieve proficiency, and many people fail the course.

FPV success rates appear to vary wildly, with different sources citing 20%, 30%, 50% or 70% — much appears to depend on the exact situation, the presence of jamming, and the skill of the pilot. A super-skilled AI might push that rate far above 70%, negate any risk of jamming and enable fleets of smart FPV drones to attack simultaneously without human operators.

[…]

Swift relies on having reliable information on the speed, location and orientation of the drone in real time. This is far more challenging outdoors where there are changes of illumination, wind gusts and other variables to contend with.

Also, Swift has to learn the course ahead of time to work out its flight path.

“The current system only works for drone racing and for a specific racing track of which you perfectly know the map,” says Scaramuzza.

The neural network which navigates through the gates is trained specifically for that layout . The other problem is that Swift trains on a specific setup and if conditions change – for example the wind changes direction – all its learning may be wasted.

“Swift’s perception system and physics model assumes that the appearance of the environment and its physics are both consistent with what was observed during training,” says Scaramuzza. “If this assumption fails, the system can fail.”

Andrei Bogdanov, CEO of Barcelona-based drone company UAVHE, is not developing his Baduga flying rifle for the military:

The problem Bogdanov is trying to solve is the control of feral pigs. The Twitterverse mocked an American user who suggested that he needed an assault weapon to prevent his yard being invaded by “30-50 feral hogs” in 2019. But controlling these animals, which cause an estimated $1.5 bn in damage in the US alone each year, is a major challenge.

Hunters usually only kill a few in a pack, causing the rest to scatter. In Spain where Bogdanov is based, hunters shoot some 400,000 wild pigs every year, but this is not enough to stop the population rising.

[…]

Bogdanov, has developed Baduga, a hunting rifle mounted on a small drone. A smart suspension system keeps the weapon’s center of gravity below the point of attachment, and gyro-stabilization ensures that the barrel remains stable regardless of wind or motion. The sights, including a multispectral camera able to see in the dark, are mounted on the barrel.

Bogdanov says that the firing platform is effectively decoupled and independent from the drone, firing as easily as it would from a tripod. The system automatically compensates for recoil, and has a magazine of 60 rounds. A further development may see automated in-flight magazine changing.

Early versions of the design employed off-the-shelf gyros, the latest iteration is custom-built for this application and weighs around 4 kilos, with the rifle adding a similar weight. The platform is a standard heavy commercial drone, similar to those which carry movie cameras and survey instruments.

[…]

Bogdanov says his setup achieves an accuracy of better than 0.1 minutes of arc, so the limitation is the accuracy of the rifle and ammunition.

[…]

“Unlike ground shooters, a drone in the air has a more accurate picture of the direction and strength of the wind over the altitude spectrum — it is easily calculated from the drift of the aircraft relative to the ground,” says Bogdanov.

[…]

So why not go down the obvious route and develop this specifically as a weapon system for the defense sector?

“Despite a common myth, developments for the military do not bring in a lot of money,” says Bogdanov. “We have shown it many times, but so far the matter has not gone further than talks and interest from military customers connected with it.”