-- Control Engineering, 6/20/2008
Rosemont, IL – It seems odd to say that software for swarm robotics has been debugged naturally for 120 million years, but that information and 24 robots smaller than a toaster helped James McLurkin, MIT roboticist, explain three globally important things engineers should do (beyond creative motion control and improve system integration).
Enthusiastically claiming an engineer’s right to “nerd pride,” McLurkin, an inventor, researcher, and teacher, told the Sensors Expo keynote audience on June 10 they should share engineering joy to:
1) Support the engineering community. Engineering is fun, and young people need to know that, he advised. Don’t say, “I’m an engineer.” Instead, “Tell young people what you do. That’s cool,” McLurkin said.
Volunteer to help in engineering related activities for youth, “or if you’re brave, teach,” he said.
McLurkin grew up acquiring engineering skills, by making things out of boxes, building things with Legos, working with model trains, and figuring angles and balance on BMX stunt bikes, which he still demonstrates.
2) Help our country. We’re not building enough engineers in the U.S. To help, donate time, hardware, or software to educational institutions. Eclipse open-source programming toolkit, for instance, has been very useful at universities.
3) Save the world. We can engineer our way out of global warming. Only engineers can do this—ourselves, and by inspiring the next generation of engineers to solve the problem.
As for robotics and swarm-type software, researchers have broken down bees’ nectar-collecting decisions into software (Cornell, 1995), McLurkin said, explaining the comment about 120 million years of debugging. How might that apply to robotics? As a graduate student at the MIT Computer Science and Artificial Intelligence Laboratory, McLurkin works with professor Leslie Kaelbling researching distributed algorithms for swarms of mobile robots.
Goal is to understand how to use local interactions among nearby robots to produce large-scale group behaviors from the entire swarm. All the intelligence doesn't have to be embedded in one set of controls; interactions create new knowledge.
Imagine how a swarm of 10,000 small robots could work together to do something useful. Suppose a few hundred heat-seeking robots swarm to locate survivors at an earthquake site, and 20,000 cockroach-sized bots scurry through the rubble confirming locations, and report to rat-sized structural robots that analyze data, and direct a few brontosaurus-sized rubble-removing robots.
What could 2,000 robots do on Mars? Two are there now. McLurkin’s stage demonstration simulated how a group getting off a space ship might handle challenges.
On stage (see photos) the robots demonstrated a dispersion algorithm, transfer of leadership, formation of groups according to a determined need, perimeter security, and return to “home.”
The demonstration included applause and levity. “I don’t know what this one thinks he’s doing,” as one wandered off; McLurkin put him back in the pack. He kept a watchful eye over the brood throughout the demonstration. Without sensors installed that would detect an edge, “they will happily plummet off the stage, so I’m trying to keep them toward the center,” he noted.
Later, he had them line up by identification number, a clump of little robots formed, still trying to move and bump out of the way, without much success. “They don’t have knot-rectifier software,” he smiled, amid laughter. “Must be too many engineers in the room” for this part of the demonstration to work right. He grinned, then moved a few of the robots well outside the clump so they could line up, and they did.