Happiness Is A Swarm Gun

Perhaps we can be grateful at least that someone in the US Navy has a sense of humour as plans were recently announced to field test a new swarm robotics program called  'Low Cost UAV Swarming Technology' codename LOCUST.

Biblical references aside, the aim of their program is to create disposable self-organising drones swarms, as a cheaper and more dangerous alternative to $1M missiles. Each 'missile' is actually an airborne pack of 30 drones, individually shot from a canon before forming into co-operative flock in mid-air under the general guidance of a single sheepherding pilot.

As a program, LOCUST isn't flying into uncharted territory, swarm robotics is one of the boom research areas in labs across the world, with engineers tackling a full spectrum of scales, from nano robotics, at a cellular level, insect sized "Kilobots", choreographed dancing spheres, micro quadcopters and even boats.

There are some remarkable possibilities, especially in medicine and nano-robotics, but as the physical size of the robots increases, the military potential becomes more perturbing. While civilian applications from disaster relief to agriculture might be significant, there less reflection about the outcome of projects specifically designed to exponentially scale, not just to tens or hundreds of members, but quite possibly hundreds of thousands.

Pitched as a weapon, these numbers presents some spooky military options - a mass of robots that can confuse and overwhelm an enemy, guard a convoy, transport objects, and solve problems as a pack. The stated aim is to form self-organising clouds with behaviour that is adaptable to the situation in which they find themselves, a squad formulating solutions for goals together rather than simply moving from A to B. By stealing a few of ideas from nature, robot teams are able to achieve en masse what seem like impossible feats, for instance 20 tiny robots moving a 1800kg car.

While single drones are usually associated with surveillance or the long-range kill - operating unseen, as a spy or a sniper - a major 'advantage' of the swarm is it’s physical presence. It is not a single target, but a headless, amorphous beast, growing into huge flocks, impossible to individually target or disarm.

In someways the creation of the swarm only represents the next logical step after the deployment of physically reliable autonomous robots. That drones are now cheap enough to be on sale in Argos is testament to the massive leaps that has been made in UAV technology over the last few decades; a progression that has relied the simultaneous solution of several non-trivial problems including the miniaturisation of microprocessors and sensor technology, wireless communication, GPS, sophisticated libraries for movement stabilisation and even battery technology.

These first challenges of drone robotics are perhaps similar to those a baby encounters in infancy; the development of basic motor skills, moving and not falling over, of limited communication, or navigating obstacles in the environment, knowing when to refuel. And of course single drones, like children have only limited autonomy.

For the drones to work together effectively however requires a different set of hard problems to be overcome that are dependent on first ones. A good list of the challenges is defined in more detail in this paper but includes more abstract ideas such as team goal setting, the leaving or joining of the flock, pattern formation, collective movement, task allocation and collective mapping. Extending this slightly dry set of technical terms we might see this next stage of robots akin to a child’s first exposure to society, and the development of social skills and and appreciation of the relationships and hierarchies of their community. In general a far more nuanced set of challenges, possibly without defined 'solutions'.

An intriguing aspect of this next stage is that while the challenges of actually getting a drone into the air is generally a hardware problem, the creation of stable, collaborative swarm is in the software and some of these problems have had people interested for decades already. The most acute insights into crowd behaviours actually originate entirely in computer code, for instance the first simulations of flocking algorithms, developed by Craig Reynolds in 1986. Reynolds developed a set of simple rules governing patterns of behaviour of digital agents that generates simulations that at least appear to have they're own, hypnotic and anthropomorphic logic. Craig Reynolds termed these agents "boids", and defined a series of crowd behaviours, which in turned became OpenSteer, an open source library widely used in the computer games. ¹  These path-finding algorithms have a brought range of uses in simulation, not least the crowd-dynamics of CGI films - one of the first being the stampeded scene is The Lion King.

It's worth considering the fundamental metaphorical difference between a swarm and a crowd. While a crowd might be defined by a shared purpose, as in the crowd-sourced, crowd-funded web and it's varying degrees of utility, as a group is not necessarily adaptive, rational or even logical - it's simply strength in numbers.

However it's not just volume that gives the swarm more terrifying potential - reliant on organisational principles we cannot easily predict or outsmart - in the case of a weapon, something that might can attack from all directions. While it also relies on the group having some shared purpose and is defined by relationships between those within it - co-operation of some kind between it's members is essential. While crowds and mobs dissipate or run out of energy - the members of a swarm are essentially dependent on it for identity, for safety, for survival or create conditions where individuals are dispensable for the benefit of the whole. ²

The most obvious and powerful exploration of this is The Birds, Hitchcock’s terrifying portrayal of the reversal of natural order. Another film which explores this motif in incredible visual splendour is Phase VI, the only feature-film directed by Saul Bass, the famed designer responsible for amongst many other projects, the iconic poster for The Birds. In Phase IV, ant species of the world begin to unite and before turning against the scientists sent to research them. What makes these behaviour of the creatures in both these films so acutely uncanny is that we are free to imagine there is a pervasive and collective purpose to their actions but we cannot comprehend what it is - in fact we can do little but run. _³

Since it's helpful to work at various levels of abstraction within the field of swarm robotics and everything is simulated before it's built, techniques for controlling the swarm can be borrowed from disciplines as distinct as particle physics or sociobiology. Agent based simulation is already a well established research technique in social sciences and macroeconomics, as well as the bedrock of computer game AI.

Of course the ‘agents' inside an economic model might operate on axes labelled 'incentives' or 'productivity' - but the exploration and decision making in these ’solution spaces’ might be entirely portable. Strategies based on video games or emergent phenomena found in networks, simulations and "Big Data" become fair territory to exploit in honing the intelligence of the swarm and in a sense weaponising the research itself.

What’s more it's a two way street - in one study of real-world locusts researchers observed the locusts “spontaneously generate the same formations that physicists observe in statistical models.”  Some of these principles appear essentially interchangeable between real animal behaviour, mathematical simulations and artificial intelligence - which in a world of drone warfare, might provide a genuine military advantage.

So while the drone and it’s ability to remotely induce terror, has slowly but surely entered the mainstream, the image of the weaponised swarm hasn't quite permeated the public consciousness yet - although this may soon be about to change. Some of the research projects are making it out of the lab and into the interactive art world - including this video by Intel demonstrating the choreography of  100 drones set to an orchestra. It's a corporate promo so it hardly critical about the military potential of the technology but it does represent a 21-gun salute for the launch of the swarm age.

Aside from the obvious question about how Beethoven might feel about Intel mashing the five notes of their jingle into his fifth symphony, the video appears to be a celebration that we are even able to maintain control of this many drones, presupposing the swarm as a wild force that can be tamed but only just.

Perhaps as the drone has become a focal point for a discussion surveillance, military power overseas and the technological disconnect between remote control and effect, the evolution of swarm dynamics may come to represent something beyond that. If robotic swarms are going to be part of the future of our 'security apparatus'  they represent a demonstration of not just a new weapon but the cultivation of a different kind of fear, fuelled by a ravenous appetite for emergent properties and Big Data. Less an exercise in soft power but something altogether more 'fuzzy'.

Notes