Swarm Intelligence

What is it, and how armies can use it other than for drones.

Aug 24, 2024

Before stumbling across the theory of swarm intelligence, I was only familiar with the concept of swarming as a tactic or in the future use of drones. The context and explanations that were given were mostly focused on mass, as in having large numbers, in order to overwhelm the enemy, or be everywhere at once but still act as a swarm. The real value of swarms is not in their numbers, it is in the emergent intelligence it can create to solve problems. It is better to outsmart the enemy, instead of only being more numerous. Moreover, we should not limit ourselves to the application of swarm intelligence to drones. In this post I will argue that we can also apply these concepts to mission command and logistical planning. Let’s first dive into what swarm intelligence is actually about.

Here is an explanation of swarm intelligence by the brilliant scientist Robert Sapolsky in his latest book named Determined: ‘An ant forages for food, checking eight different places. Little ant legs get tired, and ideally the ant visits each site only once, and in the shortest possible path of the 5,040 possible ones (i.e. seven factorial). This is a version of the famed “traveling salesman problem”, which has kept mathematicians busy for centuries, fruitlessly searching for a general solution. One strategy for solving the problem is with brute force – examine every possible route, compare them all, and pick the best one. This takes a ton of work and computational power – by the time you’re up to ten places to visit, there are more than 360,000 possible ways to do it, more than 80 billion with fifteen places to visit. Impossible. But take the roughly ten thousand ants in a typical colony, set them loose on the eight-feeding-site version, and they’ll come up with something close to the optimal solution out of the 5,040 possibilities in a fraction of the time it would take you to brute-force it, with no ant knowing anything more than the path that it took plus two rules (which we’ll get to). This works so well that computer scientist can solve problems like this with “virtual ants”, making use of what is now known as swarm intelligence.”[1] Sapolsky explains that this adaptiveness can also be found in trees wanting to branch skyward most efficiently to maximize the exposure to sunlight, the branching of our circulatory system and even the branching of neurons in our brains. There is much we can learn from how our brains work (see earlier Post), but apparently also about how they wire themselves physically. What happens is that through simple rules, more complex systems ‘emerge’.

Sapolsky provides an example such rules by describing how bee scouts check their environment for food sources. “They each find one, come back to the hive to report; they broadcast their news by way of the famed bee waggle dance, where the features of the dance communicate the direction and distance of the food. Crucially the better the food source a scout found, the longer it carries out one part of the dance – this is how quality is being broadcast. At the second phase, other bees wander about randomly in the hive, and if they bump into a dancing scout, they fly away to check out the food source the scout is broadcasting about …. And then return to dance the news as well. And because a better potential site = longer dancing, it’s more likely that one of those random bees bumps into the great-news bee than the good-news bee. .. until the entire colony converges to the optimal site.”[2]

Now here is the interesting conclusion: “there is no decision making bee that gets information about both sites, compares the two options, picks the better one, and leads everyone to it. Instead, longer dancing recruits bees that dance longer, and the comparison and optimal choice emerge implicitly; this is the essence of swarm intelligence.”[3]

You might wonder, how do bees decide how long they should dance? Isn’t the qualification of a food source as being ‘fairly good’ to ‘really good’ subjective? Can bees be wrong in their qualification? Yes, of course they can and it is likely subjective. However, this subjective interpretation of one bee gets checked by many others bees when they visit the food source. If one (or only a few) are wrong, then the majority of bees will not reproduce the dance and thus follow other bees. Errors will remain an outlier and not become a problem as long as you have large numbers, hence the need for a swarm.

Future military applications of swarm intelligence
As explained in the beginning of this post, much of the current applications of swarm intelligence is related to making drone swarming effective. Getting drones to swarm is being achieved by three simple rules: separate, align, and cohere. These rules were created by Craig Reynolds in the 1980’s when creating simulations based on swarming behaviours. These rules make sure that large groups of drones can maintain a minimum distance to each other (ie. separate), have the same heading and speed (ie. align) and remain in a coherent group (ie. cohere).[4] Yet again, what these rules achieve is that the drones can act as a swarm. It does not give them the power to solve other problems or reach an optimal solution where normally we would need brute force computing power.

There are already numerous companies that sell military applications of swarm intelligence, mainly related to drones.[5] [6] There are approximately 60 vendors that are working on drone swarming technologies and over 200,000 patents linked to aerospace and defence industry on this topic.[7] Combining swarm intelligence with drones does seem like a logical first step for testing and applying the concept with technology. One (somewhat outdated study) predicts that around 2028 swarm weapons will be feasible.[8] Considering the recent developments in Ukraine have spurred the innovation in drones, combined with the enormous strides made in online AI-language programs such as ChatGPT, I think this prediction of 2028 might be fairly accurate.

Countries like China are also researching and investing in swarm intelligence. This report from 2017 already highlights the developments made by the Chinese People’s Liberation Army. Its describes how China sees potential to use drones for saturation assaults to overwhelm the enemy defences of high-value targets.[9] But China is not alone in this, more recently Russia is also claiming to work on drone swarms.[10]

We cannot fall behind and we should continue to research and invest in swarm intelligence. I might even add that it can be the key to unlocking the true potential of what I have called ‘The Deadly Triad’ in one of my earlier posts. In this post, I argue that if we want to make unmanned systems valuable on the battlefield, they will need to become autonomous and they need to come in great numbers. I argued that they don’t need to be very intelligent, but when we program them to act according to the principles of swarm intelligence, they can become very deadly indeed.

However, we should not limit ourselves to the application of swarm intelligence to machines. If organisms like bees, ants and even neurons can use swarm intelligence, why not humans?

Mission Command or Swarm intelligence?
My first thought when reading this explanation of Sapolsky was that this is not only applicable to directing swarms of drones, but it could perhaps also be used to direct our current troops, consisting of human soldiers. Western Armies pride themselves on their mission command, in which unit commanders receive a mission with an intent from their commander. This intent describes the reason why the mission needs to be executed and provides a guideline for the unit commander to act accordingly when things do not go as planned (which they never do). The unit commander then knows what his intended goal is and can act accordingly without first asking or consulting his superiors. In practice, mission command often still requires a lot of coordination and superiors often give so many coordination measures that the actual freedom of action can be severely limited. Decisions are often still made at higher echelons. What if we could learn from swarm intelligence? There is, I think, a link with mission command. Mission command is meant to result in the most optimal solution, without the (or any) higher echelon having all the information or even taking part in decision-making.

Sapolsky: “there is no decision making bee that gets information about both sites, compares the two options, picks the better one, and leads everyone to it. Instead, longer dancing recruits bees that dance longer, and the comparison and optimal choice emerge implicitly; this is the essence of swarm intelligence.”

Imagine a Brigade attempting to breach enemy defensive positions. The frontal platoons report where they are most successful in breaching the enemy lines, but they only report this to their nearest units, which in their turn report this to their flanking or rear units. What if we only communicated something like a static burst containing a direction (as with the bees)? And that the duration of the burst would automatically override shorter bursts of less successful breaching attempts (as with the bees)? Communication could be limited to a short range since only nearby units need this information, and they can relay it on to their adjoining units. Yet if you communicate over longer distances, than higher command would also know where to allocate the main effort of the fire support, to name one example. Of course, this thought experiment has its downsides and most importantly, it requires large numbers. The main point is that we can perhaps reach optimal solutions without any central command. The only thing that we need to enhance is our interpretation and application of mission command. Perhaps we can add some simple rules for communicating success or failure, but also opportunities or risks, just like the bees have their simple dancing communication.

Ironically, Western Armies are working towards solution in the opposite direction. What many Western armies are now focussing on are communication and information systems (CIS) that can transmit information faster and to as many levels as possible, where everyone has access to all the information (with large consequences for bandwidth). We think we need more information in order to make better decisions. In their book, Swarming & the Future of Conflict, John Arquilla and David Ronfeldt argue that technology first needs to mature in order to make swarm warfare possible. Unless I misunderstand their intention, I disagree that we need more sophisticated (communication) technology. Biology shows us that simple organisms or even parts of organisms like singular neurons apply simple rules to achieve emergent swarm intelligence. The core strength of swarm intelligence is the lack of central decision-making. Why would we need sophisticated CIS? What we need is to become less reliant on CIS, because what we can learn from the recent battles in Ukraine is that exactly these systems will be jammed by our adversary. Command posts and other crucial nodes will be targeted first. But what if we weren’t reliant on our command posts in certain phases of the operation?

More information is not always necessary, nor better. What matters is that you have the necessary information at the exact right moment, which can sometimes be surprisingly little information. What if we don’t need more information, but instead better mission command with simple rules that can be communicated between separate unit commanders? This will not only provide more optimal solutions, but also quicken our decision-making. Moreover, it would be very difficult for the enemy to eavesdrop or otherwise induce any plans communicated by the higher echelon. But there is more we can learn from the evolution of simple organism.

What logisticians can learn from slime molds.
Yes you read that correctly, slime molds. These are slimy, fungal, amoeboid, single-cell protists, that grow across a surface like a carpet in search of food.[11] They can join forces and merge into a large ‘organism’ in search of food, because one single slime mold cell can ooze little distance in search of food. Collectively, they can solve problems and thus survive. Sapolsky describes in his book how they can find food in a maze for example. But the most brilliant experiment has been done by Atsushi Tero at Hokaido University. What he did was place oat flakes (food) at very specific locations in a walled-off area. What the mold initially does is expand across the entire surface and connect with all the different food sources. Sometime later, most tubules (tubes of clustered slime molds) slowly retract, leaving the shortest total paths of connecting tubes between each food source. This is a perfect simulation of the problem of the ‘travelling salesman’, as discussed earlier. Now the brilliant part. The locations of the oat flakes are not random, they represent the exact locations of the different train stations of Tokyo, whereas the entire surface of the walled-off area represent the map of Tokyo. The pattern that emerges from this slime molds is statistically similar to the actual train lines between the different train stations. See picture below: The Travelling Slime Mold[12]).

This simple slime molds can arrive at the same solution as scores of urban (human) planners are capable of. Why? Because of the emergent swarm intelligence.

Military logisticians also have to deal with the problem of the travelling salesman when arranging supply routes to and from supply areas and units. It will be somewhat impractical to have actual slime molds do military logistical planning, yet computers can use the same mechanism in algorithms, and this is already being done with ‘virtual ants’.[13] This is different than simple brute force computation because it comprises the actual simulation of swarms. An application like this would come in very handy for our logistical branches. But let us not forget that they are also of use to the intelligence branch. I would love to set it at work and predict what the most optimal logistical network of the adversary would look like on the map. Although it comes with dangers, since obviously the adversary can do the same, you at least have an optimal solution from which you can intentionally divert because of cover and concealment reasons.

What I really like about this example is that it concretely describes the problem and how algorithms, or call it artificial intelligence, can possibly solve it. This technology is within our reach and can possibly easily be integrated in our current battlefield management systems, which already holds all the information on terrain and the location of units and supply areas.

Conclusion
We should further explore the possibilities of using swarm intelligence for military tactics and decision-making and apply them for human beings. There is much potential in using swarm intelligence in large numbers of autonomous, unmanned systems (The Deadly Triad), or in short, drones. However, we should not limit ourselves to applying these rules only to technology, especially since this magnificent concept is itself a product of biological evolution. If we want to be smarter than our adversaries, perhaps we should pay more attention to what evolution has to offer us, even simple organisms like bees and slime molds. Another fine example of looking Beyond the Art of War.