Tuesday, February 20, 2018

Secret Science Club Post-Lecture Recap: Simon and the Swarmbots

Last night, I headed down to the beautiful Bell House, in the Gowanus section of Brooklyn, for this month's Secret Science Club lecture, featuring the triumphant return of biologist, transportation expert, and all-around fantastic human being, my great and good friend Dr Simon Garnier of New Jersey Institute of Technology's Swarm Lab. Dr Garnier delivered the December 2015 SSC lecture and has been a regular attendee of lectures since then. Over the years, I have become an unabashed fanboi, but don't let him know that... he has a knack for self-deprecation, so he might be embarrassed.

I have to confess that I arrived at the lecture late, after a two-and-a-half-hour MTA ordeal. I boarded a 4 Train at the Woodlawn station, as usual, but train traffic was delayed because of a fire at 51st St in Manhattan which closed down the Lexington Avenue subway line. At 161st St, an announcement was made, suggesting transfer to the B/D line, which ends up going down 4th Avenue Brooklyn, which is exactly where I was headed. It was a no-brainer, but for the brainlessness of the MTA holiday second-stringers who were on duty- the platform for the B/D trains was closed, no trains. I had to wait for another 4 train, and take it to 149th St, to transfer to the 2 Train, which ordinarily would have taken me to Brooklyn, but for construction. I had to transfer to the N/R train at Times Square. The N train was slow-going, but I might have made the lecture on time but for the boobery of the conductor of the train, who never mentioned that there was no local service along 4th Ave. I exited the N at Atlantic Avenue and I dutifully waited on the platform for an R train to take me two measly stops... no R trains were forthcoming. On a hunch, I boarded the third N train that stopped at the station, whereupon I learned that I would have to take the train to 36th St, then transfer to a Manhattan bound local train to get to 9th St. Facchinello!

I finally arrived at the beautiful Bell House and I have to say that EVERYBODY had been worried about me, because the only lecture I'd ever missed was on the date of my father's death. I was greeted effusively by staff, Dorian and Margaret, and the regular crowd- here's a big hug and molto amore to everybody- thanks for being so great. I had arrived about halfway through Simon's lecture, and he was showing a video of swarm-robots combining to pull a child across a floor:

As usual, don't read the YouTube comments, it seems that there are too many people who refuse to see the practical applications of this technology to post-disaster rescue efforts.

One conundrum in coordinating robot 'swarms' is whether a central control should be implemented- a centralized control system would require a lot of computing power and would be hard-pressed to adapt to changing conditions. For example, a centrally-controlled swarm of aerial drones could be stymied by the presence of birds. To illustrate the capability of swarming drones, Dr Garnier showed us a video of Intel's drone exhibition for the opening of the 2018 Winter Olympics:

There are three rules to being able to swarm like a flock of seagulls starlings: If too close to another individual, move away. If too far from another individual, move closer. Align movement.

Using these three roles, the European Union is developing swarming drones with no central control, in a venture called the COLLMOT Project:

Predictably, the project is receiving major funding from military organizstions. To illustrate the potential dangers of weaponized swarming drones, Dr Garnier showed a brief clip of a fictional short titled Slaughterbots:

Suddenly, the idea of package delivery by drones is a lot less appealing...

The topic of the lecture then shifted to transportation infrastructure (my particular travail of the evening). It is estimated that $120-305 billion worth of productivity is lost to traffic each year in the United States. Dr Garnier joked that, if there were a percentage bounty for solving the problem, he'd take it. Much of the transportation infrastructure in the country is too old, was designed for less traffic, and bad behavior on the part of users compounds traffic problems. Dr Garnier displayed a video of a simulation of phantom traffic jams similar in subject matter to this video. He followed this up with Hyundai's 'Empty Car Convoy' commercial, which illustrates the 'too close slow down but not too much' approach to traffic control:

During the video, he called our attention to the 'terrible moment that didn't happen', and noted that undisciplined drivers could be helped by robots. He capped off the segment on robots by posing the question: Autonomous swarms- problem or benefit?

Dr Garnier then pivoted to the subject of morphogenesis- processes leading to the formation of functional structures. He emphasized the importance of functionality. This portion of the lecture involved discussion of structures formed by social insects. Dr Garnier displayed a diagram of an African termite mound. These structures, built by blind workers with tiny brains, can be several meters in height and have a complex 'architecture'- the queen is housed in the core of the mound, and there are multiple nurseries and fungus gardens. A central shaft provides ventilation, 'breathing' in or out in response to temperature changes. Dr Garnier noted that such ventilation systems could be incorporated into skyscrapers.

Dr Garnier then brought up to topic of stigmergy, the mechanism of communication through modification of the environment. He joked that every internet post ever is an example of stigmergy. To illustrate stigmergy among social insects, he began with the example of paper wasps- the first cell made by an individual wasp stimulates the construction of the next cell, and the process is continued until an entire nest is constructed. Put simply, the rule is 'observe what is done and make the best next step'. The simple probabilistic rules used by the wasps result in regular structures. Different rules result in different shapes, and in computer simulations, rules are modified to create new structures.

A potential application for this type of construction would be the use of swarming robots to build structures on Mars before the arrival of human explorers or colonists. Dr Garnier displayed a video of Harvard University's robot swarm construction crew:

In typically droll fashion, Dr Garnier joked 'robots are better at Lego than you' and followed it up with a hashtag: #childhoodruined. Other robots are pre-programmed self-assembly units, with the body of the robots being the actual construction materials:

Dr Garnier then brought up his particular specialty- ants, which are capable of using their own bodies with much flexibility. He opened with Australian weaver ants, which form chains in order to form nests:

He joked, "do not let these ants get on your body, they bite and they inject acid into you and they are hard to get off- I am speaking from experience."

He then displayed a video of fire ants forming rafts in order to survive floods, then shifted to the topic of army ants, which form nests out of the own bodies to protect their queen, then move and reassemble the nest. These temporary nests can conform to any environmental condition- filling tree cavities or depressions in the forest floor. The Harvard University kilobot swarm mimics the ability of army ants to form specific shaped structures:

One particularly useful skill of army ants is their ability to form self-repairing bridges:

The Harvard lab developed a climbing microbot which could possibly be developed into temporary patches for weak spots in bridges until the money is available for a more permanent fix:

Self-repairing infrastructure, it's what's needed when nobody is repairing infrastructure.

Dr Garnier finished his talk by noting that these robots in development are individually stupid, but collectively capable. I would editorialize that humans tend to be the opposite.

The lecture was followed by a Q&A session. One individual asked about ant castes- ants often have a diversity in body shape along with a diversity in purpose, but swarming robots tend to be uniform. Regarding the danger posed by robots, Dr Garnier noted that the danger lies in how humans use them- will the robots be delivering packages or less nice things? Dr Garnier noted that he wasn't a complex AI expert, but that swarming robots were dumb, easily controllable. He noted that swarm behavior can be exploited- for example, dolphins and sharks can often exploit swarm behavior to catch fish. Regarding computer viruses, they can exhibit swarm behavior, overwhelming computer systems through sheer volume. Some bastard in the audience asked if swarming organisms changed their behavior when introduced to novel environments, and Dr Garnier brought up the example of invasive Argentine ants- in their native environment, ant colonies are territorial, and fight each other. Introduced to alien environments, a lack of genetic diversity has led to the formation of megacolonies, such as one which stretches from the Portugese littoral along southern Europe to Italy. These colonies seem to be collapsing though, probably due to the lack of genetic diversity which led to their formation. One final question concerned Von Neumann machines, which Dr Garnier noted are far from development.

Once again, the Secret Science Club has delivered a fantastic lecture, so thanks to Simon, Dorian and Margaret, and the staff of the beautiful Bell House... and sincerely, thanks to everybody for your concern- I love you all. I also have to note that SSC alum, good guy, and rockstar Mark Siddall was in the audience as was the NJIT crew and some Rockefeller University scholars. The ride to the Bell House was horrible, but I would not have turned around for any reason!


mikey said...

Interesting stuff!

I immediately thought about 2 things, both military related.

First, there have been a pretty large number of 'war games' - computer based simulations - in which the US and Iran clash in the Persian Gulf. A few years ago, the US officer who was in charge of the 'Iranians' in the simulation discovered he could 'swarm' small, fast boats armed with anti-ship missiles against the larger, much more capable American ships. The thing is, he won, and that tactic has won every time since. Given enough fast boats and sufficient coordination, there is no effective counter to the tactic. You get some - perhaps MOST - of them, but they always get you in the end.

Second, all the modern combat vehicles - fighter jets, helicopters, tanks, AFVs, Drones and even weapons - are networked. All the information from all their sensors is instantly available to everyone else. The stealth jets never even have to switch on their sensors - they have a complete awareness provided by the other vehicles. The power of this kind of networked system is much greater than the sum of its parts....

Big Bad Bald Bastard said...

That war game sounds like a swarm of army ants overwhelming a cane toad. I doubt the DoD will implement swarms, they like their big, expensive toys.

Emma said...

These are my favorite of your posts. It's really awesome that the staff at the lecture hall was worried about you because you're such a well-loved regular.

I've been reading Great Mambo Chicken and the Transhuman Condition, and I'm impressed that science has come this far down the path toward realizing the dreams of all those weird crazy people from the 80s & 90s. A robot that makes other robots is the fundamental basis of the coming nanotechnology revolution, apparently, and here we have it all happening. Kind of scary! If somebody tries to turn me into a bush robot, Imma slap 'em tho. Ain't no robots turning me into no robot!!!

The next time you see Dr. Garnier, tell him I love his root-lifting spray mousse. (That was a terrible joke, and I apologize.)