James McLurkin: Dances with Robots

James_mclurkin_2[Image credit: Geoff Oliver Bugbee]

James McLurkin is speaking. The title to his presentation is "Dances with Robots: The Story of One Engineer, 112 Little Robots and the Toys, Insects, and Star Wars Movies that made it all Possible."

He begins lightly with the three laws from I, Robot and suggests to laughter that robots don't care. They can't read the rules.

He asks three questions to begin.

  • What is intelligence? Would we know it if we encountered it? He says that we don't know exactly how to define intelligence. Do we attribute intelligence or is it something we possess?
  • Offering himself as proof, he says intelligence can arise from the interaction of unintelligent components. No, he says, we don't know how it happens.
  • Does intellect need a body? Probably. Intelligence doesn't just happen in the brain.

He cites military and Mars rovers as two of his favorite robots. The military robots save lives and the Mars rovers endure extreme circumstances.

Swarm in Nutshell. He spends his time working on multi-robot swarms, hundreds at a time. How can we get them to cooperate? They can tackle the dull, dirty and dangerous jobs and cites several examples, like finding earthquake survivors, where they would perform very valuable services.

What if we sent 200 instead of 2 robots to Mars?

In his first demonstration, he shows how they might move away from the lander and form up to go somewhere. Seven depart for a different location, leaving a few behind. Giving a command, he sends the remainder to join the first cluster.

He then spreads them out to the sound of music, having them flock toward a point on the stage. One of the biggest problems, he says, is figuring out how the group can define boundaries so that they remain coherent.

Each robot is sharing information with infrared transmitters so that they can stay together. The shared information contains location information. They can also relay information among themselves to send alerts or to dynamically react to events.

He next has the robots orbit around an important find, perhaps a newly-discovered Martian fossil.

To leave the Red Planet the robots line up behind the highest ID'd robot. The lowest designated robot brings up the rear and the whole blinking convoy proceeds toward a landmark, perhaps a recovery ship.

Much of his inspiration comes from the insect world, which exhibits complex group behaviors. There is a couple of hundred years of biological observation to draw on. Discussing Honey Bees he calls it "Beeware." Other insect software can be derived from ant scent trails used during foraging.

The "magic of complexity" arises from local interactions, which develop into complex group behaviors.

He then breaks down each bot into its parts - radio, camera, and IR inter-robot communications. The equipment is used for activities like "directed dispersion," a maneuver useful for finding hidden objects.

McLurkin's humor is winning. He has a very outgoing personality.

Having covered distributed algorithms, he asks for four women and four men to come forward and warns them "not to kick the robots."

Facing the audience, each person holds up a card with a number. He asks them pair up and to average their numbers, seek out another partner and re-average. Group and repeat. Group and repeat.

He explained how this was an example of a distributed algorithm. One person asks how, if the ants find through scent, do robots do similar things? The answer is synergy and it's difficult to code. How to take implicit cues from the environment is something he wants to explore later in his career, he adds.

Each person ends up citing the correct pair-wise number. By computing successive pair-wise numbers, the group moves toward the mean. The resulting numbers will always wind up in a tight range. This is very useful if all the numbers aren't known, or if you don't know what a number is.

This is how honeybees share food. The may make an effective local decision based on global data.

McLurkin recounts how he became interested in robotics. Cardboard boxes are great for building stuff. Model railroading taught him precision crafting. LEGOs taught him mechanical engineering. Pac-man taught him computer science. A video game is a complex piece of coding. MS Word will never tax your machine, he says. BMX bicycles taught him real mechanical systems. As he got older radio control cars from the hobby store taught him more about mobility and how, he adds, to avoid heavy, ankle-damaging radio control cars.

He details his undergraduate and graduate work quickly and then proceeds to his second robotic demonstration, which he dubs the "Martian band." The group begins to slowly beeping and chiming, syncopated at first, and then later, in time.

The group then reforms to into like groups, much like a symphony. Together they play the Star Wars theme. And with that, he concludes.