November 7, 2011
The Non-Robotic Robot
In popular parlance, the term “robotic” verges on the insulting. It implies uninspired, unthinking and repetitive. Yet at the same time, robots themselves are cool—striding through popular culture in science fiction and reality shows, alongside (or in place of) military troops, serving as science ambassadors in classrooms or pluckily plodding though nuclear disasters in Japan.
At UC Santa Barbara, the fluidity and diversity of real life inspires the approach to this new-old technology, bridging the divide between “clone” and “cool.” Many robotics projects seem to draw more from the menagerie than the hardware store, with rats and mice, dogs and dragonflies, even maple seeds and bacteria, influencing the design of hardware and software, and how robotics is taught in academia.
While specifics vary, robots are generally defined as having some mobility, whether as an arm attached to a post or an unmanned aerial vehicle (UAV) soaring overhead. They have the ability to make some decisions on their own, even if that only goes as far as how hard to screw in a bolt, and they can do this more or less autonomously. In short, a robot can move, it can think and it can make itself useful.
Ignoring fiction, explained professor Katie Byl, the first age of robots dawned in the 1980s when stationary robots doing repetitive tasks with high positional accuracy defined the term (and conjured the pejorative aspects of “robotic”). “It’s a robot that is designed for something like machining a part where the tool has to be pressed with high force and high accuracy against an end effector—not the kind of dynamics for a robot to interface with your grandmother,” explained Byl, whose working group at UCSB is named the Robotics Lab.
While these industrial robots remain and grow in importance, the state of the art moved toward robots that could move around on their own, initially over artificially smooth terrain and now over increasingly genuine—i.e. uneven and unknown —ground, sea or sky. Once they can ramble successfully, they should be able to exploit their environment and work without much—or any—real-time human guidance. The progress ranges from the very theoretical to the very practical. Professor João Hespanha, for example, has just demonstrated aerial robots the U.S. army would like to use to track down insurgents firing mortar shells in Afghanistan.
“In the future,” said professor Francesco Bullo, whose own work centers on robotic coordination, “people envision robots everywhere. You can imagine small, almost like ‘smart dust,’ agents in a very, very small space that move, sense, transit, maybe perform an action as well. But the reality is that a robot needs a battery. One can imagine everything, but the reality is that we’re nowhere near having anything on the microscale, even less at the nanoscale.”
So while the sky is the limit, there are still some very basic hurdles to step over, like bipedal walking.
Turning to fiction, from the homicidal NS-5s of the film “I, Robot” to the droids of “Star Wars” to Marvin the Paranoid Android of “Hitchhikers Guide to the Galaxy,” they all shared the ability to walk.
Having been raised on a diet of Hollywood robots that are as agile as ballerinas, “a lot of my grad students were surprised that that wasn’t a solved project yet because everyone saw it in movies,” Byl reflected. “They make it look like we already have walking robots.” Adding to this perception are the odd Japanese humanoid robots that exhibit a large range of life-like motions—albeit “on the same stage, the same set of stairs, three times a day doing the same choreographed motions,” as she noted.
“The real challenge is to get a robot to perform as well as my two year old son, Pieter, who can kind of crawl up and down stairs with his own technique and can make a stack of blocks 20-high. Those are the kind of problems that are extremely challenging right now, akin to playing grand master chess for a computer, because there’s so much uncertainty about what you’re going to encounter.”
This challenge, more broadly addressed as locomotion (and its cousin manipulation), is an area of special interest for Byl. Before coming to UCSB last year, Byl was at Harvard University, where she worked with private industry robot maker Boston Dynamics on Little Dog. This four-legged motorized beast (and its brother Big Dog) remains a trend-setting investigator of “the fundamental relationships among motor learning, dynamic control, perception of the environment, and rough-terrain locomotion,” the company explains.
“The real challenge is to get a robot to perform as well as my two year old son, Pieter, who can kind of crawl up and down stairs with his own technique and can make a stack of blocks 20-high. Those are the kind of problems that are extremely challenging right now, akin to playing grand master chess for a computer, because there’s so much uncertainty about what you’re going to encounter.” Katie Byl, UCSB
At UCSB, Byl and her lab have extended that work on traversing so-called rough terrain, on two legs as well as four, and have also creating small, flapping-wing robots that can not only move but can sense terrain and make their own paths. In her office at Harold Frank Hall she demonstrates a variety of small robots, from solar-powered “micro-aerial vehicles” to a purely mechanical device that, using passive dynamics, could almost walk downhill forever.
These “fun, dynamic motion problems” in engineering are often bio-inspired, she continued—drawn from nature without exactly mimicking nature.
Byl dislikes the term biomimicry, saying it suggests “taking something that works and trying to reverse engineer what the structure is, as opposed to looking at a particular principle that has been successfully used in nature and trying to boil it down to a small set of principles you can use as building blocks. You can fall into a trap of trying to copy something that works without understanding why it works.”
Wright Brothers as a pair of engineers who, realizing birds had conquered yaw, pitch and roll, set about figuring out those challenges rather creating a flapping, feathered flying machine” Katie Byl, UCSB
Byl’s group is currently working on semi-autonomous control of the LittleDog robot for rough terrain locomotion. LittleDog is a small quadruped manufactured by Boston Dynamics. Shown above are doctoral student Brian Satzinger (at left) and Katie Byl.
Picking up a miniature fly developed at Harvard, she noted that its flapping wings recall those of a real fly, but use a piezo-electric movement rather than trying to replicate the musculature of a fly. She points to the Wright Brothers as a pair of engineers who, realizing birds had conquered yaw, pitch and roll, set about figuring out those challenges rather creating a flapping, feathered flying machine.
At a different extreme, she fears the legacy of “stiff” industrial robots might lead some to take the long way around a robotics problem, with researchers trying to emulate the tradition of high positional accuracy and aiming to ratchet up the degrees of freedom rather than addressing the “basic impedance problems” of locomotion.
Like many of the researchers studying robotics and its cousins at UCSB, Hespanha was drawn here by the strength of the work going on in control engineering.
“Controlling systems is so general you find it in many departments—electrical engineering, mechanical, chemical engineering, civil engineering—but in most universities this group is spread across many departments,” he explained.
“Even though UCSB is a small school compared to others,” he continued, giving some examples of larger Midwestern universities, “at UCSB all these things are working together, so we can offer more courses than many of the big schools.”
The hotbed of this multidisciplinary effort is the Center for Control, Dynamical Systems and Computation, which professor Petar V. Kokotovic founded two decades ago; Hespanha is its director, and Bullo is the associate director. Few schools have a dedicated department of robotics or its brother discipline of mechatronics, and even at the center its eclectic focuses offer only a few “or pure-play” robotics options. Byl’s Robotics Lab is the rare entity at UCSB where “robot” actually appears in the name; the lab is part of both the center and the Department of Electrical and Computer Engineering.
The Institute for Collaborative Biotechnologies at UCSB also studies some robotics issues—like Hespanha’s mortar-seeking drones—that have defense applications, and develops bio-inspired roots. And other departments have their eyes on robotic applications: in Computer Science, for example, researchers like Matt Turk and Tobias Höllerer, who are working on the human and computer interface, see immediate robotic implications.
This all suggests the birthing of a robot requires many parents, many of whom—such as those working on wireless communications—may not have realized how entrenched they were in the robot family. As Bullo explained, a plethora of technological breakthroughs in computers, cameras, gyroscopes, GPS, sensors, batteries, and more computing that, combined with substantial theoretical developments, have broken robots out of their structured environments into the workaday world.
His own area of expertise lies in crafting models and theories of how robots would perform the useful tasks these breakthroughs allow. “The applications you’re seeing now,” Bullo said, “are the ones that bring together the best of breed from all of these practical technologies but also theoretical algorithms.
“The robots that I work with are really airplanes, but what makes them robots as opposed to traditional airplanes is they are intelligent systems, they make decisions by themselves. They are fully autonomous and they work together.”
“The Unicorn UAV is basically a foam wing powered by an electric motor. It has an onboard auto-pilot fed by a GPS unit, three-axis rate gyros and accelerometers, differential and absolute air pressure sensors, and a magnetometer. The autopilot communicates with a ground station through a radio link. We use Unicorns to test our cooperative control algorithms at Camp Roberts, a California National Guard Facility near Paso Robles. The Toyon Research Corporation in Santa Barbara has been a UCSB long-term research partner in this area.” - João Hespanha, UCSB
“Once you have the perception of the environment, you have to ask yourself, ‘How are my robots going to act among themselves, how are they going to interact with the environment, how are they going to interact with their human operators and with humans moving in the environment?’ ” His algorithms in the burgeoning field of network science tend to blur neat distinctions between robotics and the natural world, since the underlying principles that might govern a robotic network would might also find a home in activities as far afield as running a power grid or divining the intricacies of human society.’
The connection becomes more intuitive as Bullo points out that “your brain is composed of a massive network of interacting neurons,” and that in biology, the models, methodology and tools for understanding multi-agent systems are common and broadly applicable. So as robots learn to live in the environment, their masters learn to draw from it.
One U.S. Army-funded project in which Hespanha has been deeply involved is creating UAVs that can spot exactly where a small explosion—such as that from a mortar—has taken place.
Using microphones placed on the ground, these prototypes identify the location of the firing, then vector in a second UAV to snap a picture.
Because in places like Afghanistan these very mobile weapons can be fired and moved quickly, often from amid urban areas crowded with innocents, nailing down where these weapons are would be a boon for the military and the civilians.
Unlike Byl’s flapping-winged robots, Hespanha’s drones look like fixed-wing aircraft. “The robots that I work with are really airplanes,” he explained, “but what makes them robots as opposed to traditional airplanes is they are intelligent systems, they make decisions by themselves. They are fully autonomous and they work together.”
In another project, he’s working on tiny aircraft shaped like maple-tree seeds, but with antenna and tiny propellers, that a GI can pull from his pocket and throw in the air and that will follow him “like a dog,” reporting the soldier’s status and action back to headquarters.
While these projects have easily understood military applications—as do many of the robot projects, which are often funded by the Army or the Pentagon’s Defense Advanced Research Projects Agency—they also have considerable potential to be useful in civilian life.
“The robots are collecting information, sharing it and adapting in real time to things that are happening in the real world.” - Francesco Bullo, UCSB
Some of Bullo’s algorithms are also inspired by nature. The strategy that drives the search algorithm has been inspired by the movements of bacteria. Other efforts have been inspired by fish. Bullo’s work on gossiping robots has been inspired by animal herd communication.
Rather than sharing choice tidbits about who’s oiling who, groups of these gossiping robots communicate small bits of immediately applicable information, doing it with the lowest possible communication requirements to nearby robots, essentially whispering in each other’s ears. Look no further than lions, ant and bird flocks for examples, he added. The robots—or ducks—could be shouting this information, but that effort uses lots more energy and creates interference for others.
Like those Japanese humanoid robots, Bullo said, “it is relatively easy to coordinate groups of robots if every robot knows every bit of useful information. Rather it’s much more realistic—and much more challenging—if you’re not so sure where is a safe place for a robot to travel, where there is an obstacle, where there is a phenomenon of interest to measure, and so forth. Basically the robots are collecting information, sharing it and adapting in real time to things that are happening in the real world.”
The multidisciplinary and hands-on nature of robotics make it a hit among students—and a valuable tool for teachers.
At UCSB, undergraduate and graduate robotics courses are routinely oversubscribed, and courses taught by Brad Paden and Katie Byl must turn away some of the eager. Part of the attraction is likely its fun quotient; Legos and Microsoft Kinect, after all, are legitimate teaching tools.
Professor Brad Paden’s introductory robotics control class, for example, is described as an “overview of robot control technology from open-loop manipulators and sensing systems, to single-joint servovalves and servomotors, to integrated adaptive force and position control using feedback form machine vision and touch sensing systems. Design emphasis on accurate tracking accomplished with minimal algorithm complexity.”
That boils down to splitting into teams to make “roborats,” which compete against each other at the end of the course—vying for the honor of collecting the most “cheese.”
Those aren’t the only robotic rodents roaming around campus. The UCSB student branch of the Institute of Electrical and Electronics Engineers (IEEE) recently completed its “micromouse” which, in the best traditions of the lab, had to negotiate a maze of which it had no prior knowledge. To address this challenge, the students created three sub-teams to tackle the software, mechanics and electronics of the beast. (The organization’s mouse joined 13 others in a first-ever California Micromouse Competition at UC San Diego this May.)
In another project, the students created an “autonomous package delivery robot” built around an electric wheelchair. The project involved upgrading the chair’s electronics, fitting it with the appropriate sensors to help it navigate and writing the software to help it navigate, move and operate on its own.
That kind of hands-on learning will be a larger part of the undergraduate curriculum thanks to instructional improvement grants Byl and Bullo received.
The resulting curriculum, Bullo said, “really redesigned the undergraduate course from scratch.” While robotics is very graphical, he said, the classic texts really focused on factory-floor type robots, which, while valuable, don’t reflect future need to understand the kinematics and dynamics of life outside the shop.
Because robotics draws from so many disciplines, it’s an excellent vehicle to teach them. That’s what drives Amir Abo-Shaeer, a UCSB mechanical engineering graduate who founded the Dos Pueblos (High School) Engineering Academy at his Goleta alma mater in 2002. His success in engaging students in science, technology, engineering and math garnered Abo-Shaeer a “genius grant” from the John D. and Catherine T. MacArthur Foundation last year—the first given to a public school teacher. His program is also the focus of a new book—“The New Cool: A Visionary Teacher, His FIRST Robotics Team, and the Ultimate Battle of Smarts”—by best-selling author Neil Bascomb.
The capping event for the Dos Pueblos academy’s seniors, much as it is for the UCSB undergrads and the IEEE students, is entering their robots in an international competition. In this case the contest is Dean Kamen’s For Inspiration and Recognition of Science and Technology (FIRST) Robotics Competition. The academy has twice taken the Best Quality award at the championships.
“The thing about robotics that’s nice at the high school level is that the stuff the kids have learned in high school is what they actually use to design and build the robot,” Abo-Shaeer said. “There are other things that would be harder to teach from an engineering standpoint—it would be hard to teach chemical engineering in a high school setting, for example— but kids at that age can understand how to put together a pretty sophisticated robot with what they’ve learned.”
The end result, he said, “is a more nuanced metric, but no doubt when you look at that educationally we’re achieving success.”