The First World War saw flights of fighters coordinating among themselves, most doing so without even a means of electronic communication. The Second World War saw bomber raids of many aircraft, formed together under the control of their pilots, and coordinated by radio, flying exactly where and when they were told to multiply and mass effects.
Formation flying multiple aircraft is not new; military aircraft have been flown en masse for more than a century.
Since the 100-drone formation flights by Intel in 2015, I have observed a growing trend that many – particularly in the media – seem to consider formation flying by multiple unmanned aircraft to be something sensational. They talk of the ‘swarm’ and use it frequently to generate sensationalist headlines such as ‘swarming killer drones’. The connotation is extremely negative. However, they are wrong to use the term ‘swarm’.
In ‘swarm robotics’ robust, scalable and flexible collective behaviours are designed for the coordination of large numbers of robots acting as a collective that takes inspiration from the self-organised behaviours of social animals.
In the military context, swarming is when autonomous or partially autonomous units of action attack an enemy from several different directions and then regroup without being managed by a central command. Some might say that this is just manoeuvre warfare.
Swarm is defined as ‘moving in or forming a large or dense group’. It is a collective noun, but that is not how the term is being used in the media: ‘swarm’ is being used as an adjective and is conflated with ‘swarming’ in recent discussions. Swarm and swarming are being used interchangeably to describe a collective behaviour, which is quite different from simply describing the movement of a large group of things.
What is key in swarming behaviour or swarm as an adjective, as opposed to ‘a large group’ noun, is the automated or autonomous sensing of and adaptation to the environment by the members of the swarm.
The swarm members have their own detection, programming, reasoning, and decision-making abilities and perform these roles themselves, independent of a central control unit.
One of the most well-known contemporary examples of the incorrect use of ‘swarm’ is the multiple UAV attack against a Russian airbase in Syria on 5 and 6 January 2018.
Journalists often quote the incident as the ‘first-ever drone swarm attack’. Each of 13 unmanned aircraft involved in that mission was loaded with a pre-programmed flight plan and explosives and launched to complete its mission. Individual aircraft did not sense anything other than their location.
Importantly they did not sense that they were being tracked and targeted (and neutralised). They did not reason and did not adapt to or make decisions about their environment.
The attack failed: six were forced to land at assigned coordinates using electronic warfare equipment and seven destroyed by anti-air missiles.
This ‘swarm attack’ was no more technologically advanced than the German V-1 buzz bomb attacks on London and surrounds during the Second World War, or the multiple Tomahawk missile strikes of the First and Second Gulf Wars, or the coordinated missile strikes against Syrian facilities in 2017 and 2018. The buzz bomb was sensational 75 years ago, but this swarm attack was hardly a ‘first-ever’.
The fact is that all UAV out there today are either quite dumb — programmed and have no or limited-sensing — or very smart because they are under the direct control of a human operator.
When these smart, controlled UAVs are grouped into a formation or are executing a common mission, they are no different to aircraft formations of 100 or 75 years ago: the human makes the sensing, reasoning, and adaptation of the flight.
This is not to say that unmanned-aircraft swarming is not coming.
True swarming is coming, but it is not here yet. To achieve this, governments, industries, and academics alike are working on research to entrust swarms of robots and vehicles with behaviours and abilities to self-organise, to collaborate, and to complete multiple tasks together.
Some pieces of propaganda show swarm ‘behaviours’ that are actually pre-programmed actions with logical rigidity: this is programming, not swarming.
What does true swarming look like?
Imagine UAVs are operating across the sky, providing aerial observation, targeting, data networks, delivery of even smaller UAV and precision navigation and timing services, with ground robots that can be tasked to take action in a wide spectrum from logistics to combat and casualty evacuation.
That robotic swarm is a heterogeneous cross-domain team, consisting of dynamic configurations, sensing capabilities, spatial footprints and behavioural strategies, independent of centralised control, synchronised to work with, and cued by, their human teammates.
Imagining a more expansive vignette of robotic swarming is not too difficult:
It is 2030, and an Australian joint task force (JTF) is deployed on stabilisation operations in the near region against a force of insurgents who have been equipped and trained by a technologically sophisticated, militarised nation-state seeking to gain power at a regional pivot point.
The Australian JTF includes swarming machines in support of an Army brigade. Multiple unmanned assets come and go with trusted permission from the networked combat teams, and they operate in all five domains: on and underwater, on land, in the air, in space, and interacting with the cyber/electromagnetic spectrum.
These assets started their capability life cycle in the 2010s as small tactical unmanned aerial systems (UAS), ground robots, teleoperated armoured vehicles, and armed medium altitude long endurance (MALE) remotely piloted aircraft systems (RPAS). They are now semi- and fully autonomous.
Their configurations are dynamic, changing which assets are leading or following and adapting routes to account for unpredictable weather, changes which are frequent and difficult to predict in the Pacific.
The systems take evasive action from insurgent threats in the kinetic, electromagnetic, and cyber spectrum.
The insurgents are well equipped with mobile, radar-cued surface-to-air missiles and counter-UAS systems. At higher altitudes, a Loyal Wingman swarm protects the crewed Wedgetail by changing flight altitudes and /profiles to account for radar threats. Down at ground level, machines in the team sense themselves and their surroundings to adapt to conceal their signatures, and or to exploit the signatures of threat forces.
This adaptation occurs across a wide spectrum of sound, vibration, colour, light, electromagnetic, radar, and particulate sensing.
The machine sensing can algorithmically adjust its behaviour depending on the tactical and operational scenario and mission guidance: passive, reactive, overt, covert, offensive, defensive, or population interactive.
Humans issue the orders and the mission commands, and, as the team rolls through the area of operations, the machines are cued and prioritised by the humans and their robot teammates.
Robots are sacrificed, they use automated/autonomous kinetic engagement to shield their machine and human teammates, and they undertake the dull, dirty, and dangerous roles to enable the humans in the team to do what they do best.
This is science fiction becoming science fact.
The advent of true machine swarming behaviour is coming: an armada of machines, evolved algorithms, distributed intelligence, and complex autonomous behaviours – just as in a colony of bees.
However, true swarming is not here yet.
In the meantime, we need to dial down the use of the term ‘swarm’ when discussing multiple unmanned aircraft.
Lieutenant Colonel Keirin Joyce, CSC is an Australian Army officer who has been supporting UAS technology development within the ADF for the last 14 years. Keirin sends his thanks to article collaborators Trav Hallen, Chris McInnes and Jacob Choi. The opinions expressed are his alone and do not represent the views of the Australian Army, the Australian Defence Force, or the Australian Government.
This article was published by Central Blue on May 17, 2020.