Scenarios: When Swarms Go To War

The following scenarios each describe a possible future conflict in which small drones working together play a key role. All of the technologies in these scenario already exist and are described in Swarm Troopers.

In the first scenario, a small nation fights off Russian invaders, in the second the US deals with an insurgency in North Africa, and in the third a new British aircraft carrier meets unexpected opposition.

Scenario 1: Defying The Bear

Russia annexed the Crimea in 2014, then supported for anti-government rebels in Ukraine. It was a warning to other countries in the region of an aggressive new foreign policy.

It was no great surprise when Russia moved in to occupy an enclave in a former Soviet state a few years later, on the pretext of protecting a Russian-speaking minority population. As with the Crimea, NATO failed to take decisive action. The difference this time was that there was strong resistance from the local population, in spite of the overwhelming Russian military presence.

It looked like Russia’s invasion of Afghanistan all over again. Massed firepower faced a superior knowledge of the local terrain. As in Afghanistan, the Russians sought to neutralize the opposition with airpower. The same Mil-24 Hind gunships known to the Mujahedeen as ‘devil’s chariots’ appeared swiftly to punish any rebel forces who stayed in the open.

In Afghanistan, the US had supplied Stinger missiles to defeat the Russian helicopters. The situation was politically different here, and the guerrillas could not rely on American missiles. Instead, they made their own anti-helicopter weapon.

A few years earlier the Ukrainians had built up an unmanned air force, based on modified commercial and hobby drones, for scouting and reconnaissance. Now the same idea was taken a step further.

The Osa (Wasp) was a simple triangular drone, modified from a consumer version sold over the Internet. The main hardware change was the addition of a small but powerful explosive charge. The flight controls also had an additional mode. The sense-and-avoid app normally used to prevent collisions between drones was modified so that the Osa would seek out other aircraft and attempt to ram them.

Pilots had been warning for years about the danger of collisions with small drones; now the guerrillas would use that danger to defend themselves.

Hundreds of Osas were manufactured in garages and workshops. They cost a few hundred dollars apiece and took a few hours to make.

The first test was a complete failure. Half a dozen Osas rose up from the launch site and headed towards incoming helicopters. Their slow speed made interception difficult and they trailed after the faster vehicles. More disappointingly, when an Osa did finally hit its target, the tiny warhead proved too small to damage the Russian helicopter. The Hind gunships were heavily armoured and could shrug off anything less than cannon fire.

The guerrilla technicians were not deterred. They modified the guidance software, and a few days later the first successful action occurred when an Osa crashed into the jet intake of a Hind. The helicopter lost power and made a heavy landing in thick woodland. The crew abandoned their aircraft, which was blown up by guerrillas minutes later.

Further successes followed. Thousands of drones and components sent by international supporters were smuggled over the borders, sometimes in the parcel bays of larger drones. Designs evolved in a matter of weeks with information shared over the Internet.

The Russian helicopters shot down many Osas, but a few always got through. Radio-frequency jamming was useless because, once released, the drones guided themselves and did not rely on ground control. Protective grilles on the jet intakes had limited effectiveness. The helicopter pilots took to circling at high speed, or breaking off attacks when they saw the ominous black dots of Osas spreading out in front of them.

With the threat from the air fading fast, the guerrillas started using their Osas more aggressively. Special units were tasked with taking the drones as close as they could to Russian air bases; extended range versions could travel twenty miles or more. Air operations were virtually shut down as even a handful of Osas presented too much of a risk to fly through. Even Russian jets were forced to fly at high altitude to avoid the drones, making it hard to hit targets on the ground.

Countless thousands of Osa were lost: shot down, crashed or failed to damage their target. But those that succeeded were deadly. More and more replaced those that were lost.

Faced with rising casualties and an increasingly confident opposition, the Russians started to look at diplomatic solutions for the crisis. Arrangements were made for a face-saving peace deal, and there was an orderly withdrawal of Russian forces less than a year after the invasion.

Scenario 2: No Boots On The Ground

Timeline: a year or two from today.

When a major Islamist insurgency erupted in North Africa, there seemed to be little that NATO could do to help stop it. There was no political appetite for another bloody and expensive war, and the only involvement would be in the form of airstrikes. Even then there was a desire to limit the risk of having a pilot killed for captured, so drones were the preferred approach.  But with new drones, the intervention was to prove vastly more effective than anyone expected — including the insurgents.

A typical engagement occurred as three pickup trucks packed with armed men sped across a dusty plain, heading for an isolated village far from the nearest government forces. The insurgents intended to bring the area under their control; they would follow the usual pattern of executing any village elders who resisted their rule, extorting money, and forcibly recruiting any young men.

The insurgents knew to watch out for American drones, and they scanned the skies for the big, slow-moving Reapers. None of them paid any attention to something that looked like looked like an old vulture perched on a telegraph wire. They had no idea this was a drone with what the USAF termed ‘biomorphic camouflage,’ meaning that it had folding wings and looked like a bird.  A large number of these drones had been dropped two nights previously from a USAF MQ-9 Reaper drone. The same Reaper, now with extra fuel tanks rather than drone pods, was now circling twenty miles away.

The perching drone registered the vehicles moving along the road, and at Creech Air Force Base in Nevada, operators received an alerts of possible insurgent activity.

Two drones circling nearby were tasked to investigate. Six more perching in the immediate area, and these were activated. Each of the drones had a six-foot wingspan, and from a few hundred feet away it was hard to distinguish from real vultures. At that distance you could not hear the electric motor. Solar cells recharged their batteries while they were perched, enabling the vultures to carry out missions lasting many days. The perched drones rose swiftly, riding thermals upwards like the birds they resembled, and at eighty miles an hour they converged swiftly on the convoy.

Minutes later, the drone operators were examining video footage showing the heavy weapons on the back of two of the pick-ups. The images had been beamed back via satellite from the Reaper which was networked with the drones on the ground. Bandwidth was one of the biggest problems; there might be a hundred video cameras in the area, but only enough capacity to send back three high-definition streams at a time.

At the same time, one of the drones equipped with a radio-frequency SIGINT package started sending back data. Its electronic eavesdropping allows it to scan through all the radio-frequency activity below, a technique which has proven invaluable since the early 2000s. The SIGINT drone was able to identify the SIM cards several cellphones below. One of them was a number known to be associated with a local insurgent commander.

Details of the three vehicles were passed up the chain of command, along with a request to carry out a strike. For the drone operators the wait seemed to last forever, as the Islamist vehicles edged ever closer to the village. Then, when insurgents were little more than a mile away from their goal, an instruction came from higher command. They wanted positive confirmation that weapons were present, and that all those in the vehicles were military-age males.

Three drones circled in closer. The Islamicists spotted them, and one started firing from the back with a Kalashnikov. All three vehicles stopped and men piled out, raising their assault rifles and firing bursts at the bird-sized targets two hundred yards overhead. The operators guided the drones to where they were shielded by the sun, but luck or good marksmanship paid off and one drone came tumbling down with a shattered wing. It landed with a soft plop on the dusty ground. Two insurgents immediately went over to collect their trophy, arguing over who had hit it.

Meanwhile the analysts at Creech were poring over zoomed images of the insurgents, confirming that they were all armed, adult males. The pictures were clear enough to identify the heavy weapons in the pickups as Russian-made 14.5mm machine-guns. A minute later as multiple photographs were compiled and merged to get a single clean image, the facial recognition system gave an 90% confidence match with the insurgent leader associated with the cellphone.

The strike was approved seconds later.

The insurgents had clambered back in their vehicles and were about to set out when two drones sped towards the lead vehicles. Backup drones followed a hundred yards behind. The drones were just a few feet above the ground and moving at attack speed of well over a hundred miles per hour. Their automated guidance was precise enough so they were not just aiming at a vehicle, they were aimed so the one-pound blast/fragmentation warhead would explode inside the passenger compartment between the driver and front passenger.

The first vehicle went up in a fireball and the others slammed to a halt. Locked on and under automatic control, the second drone adjusted its trajectory smoothly and hit its target two within inches of the aim point. As the second pickup exploded, the two backup drones peeled off and climbed high, already surveying the scene and sending back video for the operators to assess the success of the strike.

Both targets were burning wrecks, all occupants instantly killed outright. The small warheads were powerful and carefully tailored to exactly this type of target.

A minute later the third vehicle turned and headed back down the track, having stayed long enough to confirm there could be no survivors. Overhead, too high to be seen except by an observer who knew what to look for, two vulture-like drones were following it.

In a few hours, the drone operators would have located the caves of the insurgents mountain base. Close-up observations from the vulture drones would confirm the siting barracks and headquarters. Every vehicle arriving and leaving would be scanned and tracked.

Further drone strikes would follow, not just on the caves and vehicles but also against individual insurgents fleeing on foot. The drones were plentiful, each of them a fraction the cost of a Hellfire missile, and well worth expending when it guaranteed a kill.

The insurgents now knew what the miniature drones looked like, but by the time they saw one it was usually much too late. The drones were everywhere, perched everywhere, flew everywhere, saw everything. They waited on rocks and telegraph poles by the roadside, and circled overhead. Unlike the Reaper drones which came and went, the vulture drones stayed permanently. If you shot down one six more would come and the shooter was a dead man, often within seconds.

Before long the shooting ceased. Killing an enemy soldier was one thing, but there was no glory in being a martyr for shooting a drone

The insurgency did not end immediately. In retaliation for the drone strikes the insurgents committed a series of increasingly violent atrocities, but those who committed them did not live long. The watchful eyes of the many drones missed nothing. Within weeks, the first young insurgent approached a perching drone, and, to the surprise of the watchers in Creech, laid down his Kalashnikov and held up his hands to offer his surrender. Many more were to follow.


Scenario 3: The Queen in Trouble

The aircraft carrier HMS Queen Elizabeth was the pride of the Royal Navy when she was commissioned in 2020. Britain had not had a carrier since HMS Invincible retired in 2005, and the politicians were eager to show what a difference the new ship would make. Within months the carrier was dispatched with a task force to West Africa. Britain was force again, and Britannia ruled the waves.

At least, that was the plan.

The task force was there to assist with a political situation typical for the region. A friendly democratic government had been toppled in a military coup. Rumour suggested that the coup had been covertly encouraged by Chinese interests, who had their eyes on rich mineral reserves in the country’s interior.

Gangs of soldiers ran wild in the capital, killing and looting. Foreign nationals lay low in their villas and hotels or took refuge in the embassies.

Britain was part of the multinational effort to restore order. If military action should prove necessary, the HMS Queen Elizabeth would provide air power, in the form of thirty-two brand-new F-35B Lighting II combat aircraft. These would assist ground troops from neighbouring African countries. Britain would play a part without putting boots on the ground, and the Prime Minister looked forward to a swift victory untarnished by any British casualties.

Shortly after the Queen Elizabeth and her Royal Navy escorts arrived, unidentified aircraft were detected on radar. F-35 air patrols made visual contact and confirmed that the carrier group was being shadowed by dozens of small, unmanned aircraft.

The drones were quickly identified as a design originally manufactured in China under licence from Israel, but copied by Taiwan and several other countries. Known as Hong-Jian (Red Sword), the drones had an eight-foot wingspan and solar panels on their upper surface which, combined with soaring software to take advantage of wind shear, allowed them to fly for days on end. The drone was made of Kevlar with carbon fiber reinforcement, with an on-board Android autopilot and an explosive charge about the size of a hand grenade. They were disposable, costing a few thousand dollars each on the international market, and as cheap but deadly kamikazes the Hong Jian had already made their mark in several minor conflicts.

Satellite imagery supplied by the US showed crates of Hong Jian drones being offloaded in the capital’s port from a Panamanian-registered cargo ship. The minimum estimate was that several thousand were already on the dockside or in nearby warehouses. Technicians were busy assembling the drones and launching them from catapults, watched by a small crown of curious locals. The nationality of the technicians was the subject of much speculation.

The self-styled “Field Marshal and President for Life” who led the coup issued a statement that his air force had the British Task Force surrounded. He denounced “all forms of colonial imperialism” and demanded the British withdraw to an “exclusion zone” some two hundred nautical miles away — or face the consequences.

The official response was an elaborately diplomatic refusal. The British Admiral commanding the Task Force made an unofficial but widely-reported response:

“I’m damned if we’re going to run away from some tinpot dictator with a lot of toy aircraft.”

The first wave of Hong Jian drones attacked just after dawn. There were over two hundred of them, and they converged from all points of the compass. They flew straight at the vulnerable parts of the ships, the radar domes, radio masts and antenna arrays. The straight lines and flat planes of the ships were simple geometric patterns that made it easy for the drones’ cameras to locate their programmed point of attack.

Although too small to be hit by anti-aircraft missiles, many of the drones fell victim to the radar-guided 30mm Oerlikon cannon and multibarrel Phalanx guns on the British destroyers, as well as the numerous rapid-fire miniguns mounted on deck rails and manned by sailors.

Video analysis showed that about a dozen of the attackers got through. There was virtually no damage, except for an F-35 which has been preparing for take-off on the flight deck of the HMS Queen Elizabeth. A drone had skimmed over the carrier’s deck and struck one side of the plane. The subsequent fire had been quickly brought under control and there were no casualties, but the £100m aircraft would require days of repairs before it could fly again.

The Royal Navy Electronics Warfare officers were experimenting with jamming the drones’ communications, but met with limited success. The drones barely used any signals at all, and with just a few pulses spread over a wide swathe of the spectrum. The swarm was flying itself with minimal human input.

Two hours later radar detected a second force of drones assembling to the West of similar size to the first. The drones were spaced about a hundred meters apart, forming a spherical cloud almost a kilometre across.

When an aircraft was sent up to monitor them, the entire cloud started converging on it. The pilot flew around the swarm and watched it gradually change direction to chase him. The drones could never catch the fast jet, and the pilot shot down a couple of drones with cannon fire, but he had to be wary of flying too close to the swarm.

The pilot was ordered to break off. He could not do significant damage to a swarm of several hundred drones, and there was too great a risk of an accident against so many. A goose or duck could damage a jet turbine; a drone with an explosive charge would wreck it. And these drones were determined to intercept the plane if they could.

A smaller cloud of several dozen drones then appeared in a loose formation between the carrier group and the airborne F-35. They had been skimming the sea at low level and had not been appeared on radar until they were a mile or two away. They were set on ambushing the pilot as he tried to return to the HMS Queen Elizabeth. When the pilot was redirected to approach from the opposite direction, half of the drones moved to block his approach.

The F-35’s fuel was approaching a critical level. Rather than run any risk of losing a plane for no advantage, the pilot was ordered to divert away from the carrier group and land in a neighbouring African country. The plane sped away from the swarm at four hundred miles an hour while the necessary diplomatic arrangements were made.

Running away might look bad, but losing an aircraft would be worse, and the Admiral could always say that the plane was diverted for technical reasons. The plane might be saved, but with the increasing number of Hong Jian, now forming several swarms in all directions, it was not safe to fly from the carrier.

Bad news was to follow: several hours after the F-35 landed, twenty drones caught up with it while it was parked on the tarmac. A film crew had just arrived to shoot a wildlife documentary, and were filming the plane and trying to interview the pilot when they spotted small drones circling overhead. The drones made several passes, apparently making sure of their target before diving en masse at the F-35.   After the tenth hit the plane disappeared in a massive fireball.

Shortly afterwards a second wave of drones was picked up on radar as it prepared to attack the task force. This time the drones zig-zagged as they came in, making them harder to hit, and they all headed for one ship, the destroyer HMS Griffin. Again hundreds of drones were shot down, but this time forty got through and each one headed unerringly for a target.  The attack was made more deadly by drone which had climbed to several thousand feet and dived vertically, making them almost impossible to stop.

Several crewmen manning machine guns on deck were killed and many more injured, and there were multiple hits on the fire control radar for air defence and on the Phalanx guns themselves. By the end of the attack, HMS Griffin had lost eighty per cent of her anti-aircraft capability. The order was given that no sailor should risk their life on deck during a drone attack. The machine-guns were silenced; more drones would get through next time.

The next morning, British newspaper headlines announced the greatest loss of life in the Royal Navy since the Falklands Conflict, and demanded retaliation. By now the Task Force was at the centre of a rotating cloud of unmanned aircraft, a cyclone of Hong Jian drones staying at the extreme range of gunfire, its numbers steadily growing. Any sign of movement on the carrier flight deck brought  drones spiralling in to make passes over it, ready to make attack runs at the first sign of an aircraft on the flight deck.

With air operations impossible, the only way of striking back would be to use cruise missiles, or to sail closer and bombard the shore with gunfire from the destroyers’ 4.5” guns. Both options were rejected as too indiscriminate.

The Admiral could not be seen to withdraw. Remaining in place seemed safe enough, until intelligence analysis of some of the downed drones showed they were carrying thermite charges rather than fragmentation warheads. Thermite, a powdered mixture of metal and metal oxide, burns at thousands of degrees.  If enough charges were concentrated in one place, they might burn through the carrier’s flight deck, putting it out of action. Or perhaps doing even more serious damage. It all depended on how detailed a knowledge the attackers had of the ship’s exact layout. Nobody dared mention the reductions in armor that had been made to cut costs during the carrier’s lengthy construction.

The Admiral responded angrily when told of the possible danger from the drones, which he still called ‘toy aircraft’. He refused to believe that a 65,000-ton carrier could be seriously harmed by such attacks.

“This isn’t the ****ing Death Star,” the Admiral snapped at one subordinate before storming out.

The number of drones was now estimated at over eight thousand – in pure cost terms, the entire swarm cost less than a tenth as much as the destroyed F-35. It seemed likely that the whole lot would be used in one massive assault on the carrier; the question was whether there would be preliminary attacks to reduce the effectiveness of her escorts. Whoever was directing the swarm was probably doing the same frantic analysis as the defenders.

As communications flew between the task force and Whitehall, the day was saved by a small group of raiders from the Special Boat Service. Transported by submarine, they had silently landed by the docks, located and stormed the drone control facility, killing numerous soldiers in the process. The communications antenna and ground station were destroyed with explosive charges, along with a large number of crated drones. Those technicians who survived — ‘illegal combatants’  — were captured and taken back to the task force, along with electronic equipment which would reveal how the drones could be jammed.

By the next morning, the drone swarm was still circling, but now it was without guidance, no more than an airborne obstacle. So long as the pilots flew carefully, they would be able to carry out their missions, and Britain could assert control of the air once more. But the reputation of the HMS Queen Elizabeth had sustained permanent damage, and her role in future conflicts was uncertain – except perhaps as a drone carrier.

Scenario 4: Swarm Apocalypse

The Near Future: The first reports came from fishing boats a thousand miles out in the Pacific. One of the fishermen took some photos and soon they were buzzing around the Internet: a sky full of tiny dots, which, at maximum digital zoom, looked like tiny grey triangles. These reports merged with rumours of Chinese action against the US into a story that an attack was under way.

The NRO confirmed that a swarm of small unmanned aircraft was heading across the Pacific towards the West Coast. Each drone was around six feet across, and there was a separation of about two hundred yards between them, flying in formation like a flock of birds. The swarm was about two miles across and estimated to contain up to four thousand drones.

The drones were powered by batteries charged by solar cells, aided by soaring algorithms which allowed them to ride air currents. The swarm was moving at approximately fifty knots or more than 1200 miles a day.
No communication was detected between the swarm and any external controller: the swarm was an autonomous, self-guided entity with no central control, but members each guided by the same set of rules, allowing them to co-ordinate their actions like a flock or birds or a shoal of fish.

This came at a time when situation over Taiwan had deteriorated sharply. A US Fleet had been sent to show support for the Taiwanese government, which was locked in an increasingly fractious dispute with China. There had been a stand offs in the South China Sea with warning shots fired after a destroyer from one side allegedly cut across the path of a frigate from the other.

Taiwan alleged that the drone swarm had been launched by China. The Chinese promptly released radar and satellite imagery claimed to prove that the swarm had come from a Taiwanese-owned cargo ship fifty miles off the Chinese coast.

The launch site was hard to local because small drones do not need runways like aircraft or elaborate launch pads or vehicles like ballistic missiles.
The launch could have been carried out by twenty teams of four men, three of them unpacking and carrying out pre-flight checks while the fourth hand-launched one drone every few minutes, over a single shift. The drones circled overhead, their numbers growing until the swarm was fully formed.

A swarm could be launched much more rapidly by having tube-launched drones like the Switchblade or Hero-30 which could be simply ripple-fired in a matter of minutes, but this would have added considerably to the cost and complexity of manufacture. Evidently the approach was for mechanically simple, cheap drones which might have a significant failure rate at launch but which could be launched in massive numbers. Consumer technology means that drone preparation and launch is now a simple process which requires minimal training.

F/A-18s from the US Navy established visual and radar contact and confirmed the swarms speed, bearing and composition, providing high-resolution pictures from close up. Unofficial sources said that the drones were carrying explosives, small explosive charges wrapped with Zirconium rings little larger than a hand grenade which would produce considerable incendiary effect and some blast, but little dangerous shrapnel.

Comparisons were made with the Japanese Fugo balloons of WWII. In 1944 Japan had launched several thousand mulberry-paper balloons, which were carried across the Pacific on the jet stream. These dropped incendiary bombs on a timer. Intended to set fire to the forests of the Pacific Northwest, the Fugos had minimal impact. Only a few hundred reached the US and they started a few minor fires. The only casualties occurred when children on a picnic in Oregon found a cashed Fugo and it blew up.

There were immediate demands from all quarters for the swarm to be intercepted. Defense officials patiently explained that the drones were too small and too numerous to hit with air-to-air missiles, and cannon fire was impractical. The also mentioned the risk of a plane running into a drone with disastrous results – it would be like walking through a minefield. Bombing was not feasible; only a nuclear warhead would be big enough to take out a significant fraction of the swarm.

The swarm was headed directly for San Diego, the centre of US Pacific Naval power.

Some suggested that the Aegis-equipped warships would throw a defensive screen around the city, shooting down any drone that entered, but the limited number of missiles and short range of the Phalanx guns used to shoot down incoming missiles made this a forlorn hope.
However, other means were available. Four E/A-18 Growler electronic warfare aircraft were vectored in on the swarm. Equipped with the latest in radio-frequency jamming technology, the Growlers could produce a high-power beam capable of burning out the electronics of a small drone.
“A turkey shoot with microwaves — we’re going to roast them mid-air,” claimed one office in an unofficial briefing.

The results of the jamming attack were not stated in detail; hundreds or thousands of drones had been downed, but a large fraction of the swarm continued on its path . The swarm had broken up when it came under electromagnetic attack, and it seems that the small drones had an unusual degree of ‘hardening’ against high-power microwaves.

Briefings were held, both on and off the record, to damp down levels of panic in the media. One statistic quoted was that the entire swarm carried less destructive power than a single one of the Air Force’s most powerful bombs. The consensus was that that attack was purely intended as a show of force to demonstrate that the enemy — whoever it was — could deliver chemical or biological weapons to the US mainland at will.

Discussion was already shifting to threat posed by ‘the next swarm’ when the attacks started.

The first strikes occurred at San Diego Airport. Although the airport was closed and all flights had been cancelled for the duration of the emergency, there were dozens of airliners parked on the apron. Drones targeted a number of the parked aircraft, selecting larger airliners and aiming for fuel tanks for their suicide dives. Within minutes several airliners were ablaze, sending huge columns of black smoke rising into the air.

Next came reports of wildfires in several areas around the city. Conditions were dry, and the drones had been able to identify the most dangerous areas to start fires. They also and took advantage of favourable winds, aiming for points where so the fires would spread and combine.
Firefighters were already overstretched with the situation at the airport, where fuel storage tanks had been ignited by coordinated efforts of multiple drones, when a new phase started. As fire trucks raced to deal with emergencies, they were attacked. The profile was the same each time: a drone spotted a big red vehicle, looped around and dived at the front windshield, detonating on impact and killing or injuring the occupants.
As the casualty reports started pouring in, authorities agonised over whether to order firefighters to abandon their vehicles and give up trying to fight fires. The fire trucks were such an easy target, and there were so many drones hunting them, that to continue looked more like suicide than heroism. By this point there were over a hundred wildfires in progress, several threatening residential areas.

Then the downtown assault commenced. Again, the attack was precise and coordinated. Certain skyscrapers had been selected, and each received the same treatment, with twenty to forty drones involved depending on the buildings’ size. One drone would smash into a plate glass window and detonate. Fortunately most office blocks had already been evacuated. A second drone would fly through the open window and explode in a fireball in heart of the building.

Sprinkler systems damped down some of these attacks, but the overall volume was far too great for them to handle. Across the grid of city streets the number of burning towers steadily increased. After 9/11 planners were only too aware of the effects of an uncontrolled fire in a skyscraper, but more and more buildings were burning and there were less resources to tackle them.

There were other attacks as well, but these did not get so much play in the media. Several cargo ships were drifting after their bridges had been gutted by explosions; these included one carrying hazardous chemicals. In the San Diego Naval base, communications antenna and radar equipment were destroyed, but the drones stayed well clear of warships.
The rail system was at a standstill after the drivers of dozen locomotives had been killed at key points. The attacks on the electrical infrastructure went almost unnoticed as electrical substations were degraded and destroyed, more drones diving in at two-minute intervals until enough damage had been done and city blocks started to go dark.

This seemed to trigger a new phase in the attack. Now every moving police car was a target.

Even as the carnage played out across San Diego, the global media went into overdrive. Many, especially in Taiwan,loudly demanded immediate US reprisals against Taiwan. Others – especially Russia – pointed at elements in Taiwan as the real culprit. Forensic reports from crashed and captured drones had so far failed to prove anything. They were made from off-the-shelf commercial components, not military electronics from an identifiable source, and the airframes came straight off an industrial 3D printer. Some camera lenses were traced to a Taiwanese manufacturer; but this might easily have been a deliberate false lead.

Fires continued to rage, and it would be many days before the last of them burned itself out and the total cost of the attack could be calculated. The loss of human life was a not on the same scale as 9/11, but the damage to the city was widespread. It was clear that a larger swarm could have set large sections of the city on fire so that the fires combined into a single huge conflagration: the sort of firestorm that devastated Dresden and Tokyo in WWII with the he loss of tens of thousands of lives.
One thing was abundantly clear: continued involvement in the China/Taiwan situation might lead to further attacks on the US homeland on a much larger scale.

A sidelight of the forensic analysis was that the estimated cost of manufacturing the drones was just a few hundred dollars each – less than two million for the entire swarm.

Two lessons stand out from the drone attack.

One was the swam was extremely hard to kill with existing weapons. Even a few thousand drones were far too much to handle, hundreds of thousands could be fielded the cost of a single F-35. A single swarm that size could wreak untold damage.

The other was the attacks were deadly because of their precision.
Vision algorithms and pattern recognition allowed the drones to identify targets and aim for the most vulnerable spot: the fuel tanks of an airliner, the upper storeys of a skyscraper, the cab of a locomotive. A small quantity of explosive and incendiary in the right place proved far more destructive than much larger munitions delivered with less precision. The swarm software allowed the drones to assess the effect of each attack and then deliver further attacks where necessary, to the same or different aimpoints.

(Note – all of the technologies mentioned above – including small drones with endurance, swarming software, precision attack software for small munitions, and low-cost military drones made from commercial, off-the shelf electronics, are already in place.