'Cat Rides a Sheep' Is the Best YouTube Video

So sheepdogs are pretty good at math, and they don't even realize it!

Okay, so perhaps that's a bit simplistic, but the reality is even cooler: an August study in Journal of the Royal Society Interface made an algorithmic assessment of how sheepdogs herd their charges. What they found is that dogs take advantage of sheep flocking behavior, in which they tend to bunch up in response to an outside threat. When this was modeled algorithmically, the team's digital sheepdog was successful in bunching up a digital flock 100 percent of the time.

Motherboard weekend editor Michael Byrne wrote a lovely story about the research when it came out, which you really ought to read. What's stuck out to me since then—and I can't stop thinking about it, which is perhaps a bit weird—is one quote from the paper.

"Herding of larger groups (more than 40 individuals) typically requires multiple shepherds," the authors wrote. "However, single sheep dogs can successfully herd flocks of 80 or more sheep both in their everyday work and in competitive herding trials."

While it's tough to call anything "perfect" when it comes to science, when everything's messy and caveats abound, it's mind-boggling that sheepdogs have found a more-or-less mathematically perfect method of solving an incredibly problem: How do you round up a bunch of individuals when there's 80 of them and just one of you?

Okay, so let's approach this from the flip side for the hot second: Sheepdogs are good at herding sheep. No shit, right?

The natural response to such a sentiment runs something along the lines of but science!!!!!!, which remains valid here. And yet the idea that this shepherding system is so dialed-in offers an interesting case study for one of the most world-changing developments on the near horizon: the deployment of semi-autonomous, distributed systems. Yep, I'm talking about swarms.

Our robo-future isn't going to be ushered in by humanoid machines, because robo-humans are clunky and awkward to make. Instead, ambient bots will be skittering around everywhere. Take a Roomba—a fairly simple, single-purpose bot—and multiply that by the trillion-sensor scale projected for the Internet of Things.

Having trouble envisioning this? An ecological model fits well: Over time, little bots will evolve—or be developed by companies in return for your money, but you know what I mean—to exploit every ecological niche the Anthropocene has to offer. And again, ambient is key: All those sensors, wifi refrigerators, and interior climate-controlling dragonfly drones will work in concert to make sure that you, Person of the Future with Disposable Income, can live and work in maximum, data-driven comfort.

So the home of the future is crawling with bots. Great! Let's take this even bigger: We're entering the very early days of commercial drone usage, which is typified by Amazon's stunty delivery drone promise.

Now, battery life for UAVs still sucks, and the FAA is still dragging its heels when it comes to putting together commercial drone regulations, yet it seems assured that drones are eventually going to flood US skies. Don't believe me? Well, trust the boffins at NASA, who are already developing next-gen traffic control for our drone-covered skies.

The power of drones, like most robotics, doesn't lie in just one machine, but in making massive, networked swarms. Imagine a fleet of semi-autonomous drones creating a persistent aerial net over, say, a farm, constantly monitoring crops to manage irrigation in real-time. When one drone runs out of battery, it returns to base to charge, and another automatically takes its place, and the network flies essentially in perpetuity.

If this sounds wild, you're right. And yet I spoke with an acquaintance at a big UAV manufacturer about this very thing a couple weeks ago, and he said these types of drone networks aren't all that far off.

The swarm conversation isn't only taking place in bars, either. Here's a group of French researchers on the topic:

Actually, in the current generation of UV Systems, several ground operators operate a single vehicle with limited autonomous capabilities. Whereas, in the next generation of UV Systems, a ground operator will have to supervise a system of several cooperating vehicles performing a joint mission, i.e. a Multi-Agent System (MAS). In order to enable mission control, vehicles autonomy will increase and will require new forms of Human-system interaction.

Here's the kicker: That paper is from six years ago. The swarm conversation has advanced a lot since then, with a focus on how to actually get swarms to regulate themselves. Naturally, we come back to the sheepdogs (thanks to Byrne for pulling this quote).

"Our approach should support efficient designs for herding autonomous, interacting agents in a variety of contexts," the shepherd algorithm researchers wrote. "Obvious cases are robot-assisted herding of livestock, and keeping animals away from sensitive areas, but applications range from control of flocking robots, cleaning up of environments and human crowd control."

That human crowd control bit is a bit terrifying, but remember, we're talking about the power of massive, distributed swarm of robots that can work semi-autonomously with light human direction. A more uplifting application is search-and-rescue. Large scale searches are almost always limited by human resources, so why not use bots?

In a paper on the arXiv pre-print server, Alejandro López-Ortiz and Daniela Maftuleac present a mathematical approach to swarm searches at sea, and their models worked swimmingly. "We argue that such searches are also more cost effective, both from a net dollar perspective and in terms of speed of recovery for time critical searches," they write.

And the really cool thing here is that a lot of the top approaches for governing swarms find analogs in the biological world, as with our sheepdogs. If the goal is to have a single human be a shepherd for multiple drones, perhaps it's time to stop referring to drone networks as swarms, and instead call them flocks?

That's the terminology used in a paper presented at this year's IROS conference, in which researchers developed the current holy grail: a mostly autonomous, self-organizing flock.

"We present the first decentralized multi-copter flock that performs stable autonomous outdoor flight with up to 10 flying agents," they write by way of introduction. "By decentralized and autonomous we mean that all members navigate themselves based on the dynamic information received from other robots in the vicinity. We do not use central data processing or control; instead, all the necessary computations are carried out by miniature on-board computers."

While the algorithms powering the flock aren't purely organic, the concept is a huge step forward for all of this flocking, swarming research. It's pushing towards a big vision: A day when the ability of a single person, whether it be for searching or crop monitoring or whatever else, is multiplied across an entire group of self-organizing drones. When that happens, it will mean a whole lot less work for a whole lot of us humans. Losing jobs to drone swarms is pretty dystopic, sure, but hey, at least we'll have more time to watch videos of cats riding sheep.