We, robots

We, robots

By John Deever

Quadrotor control: Tracking work with, from left, Anne Mahacek M.S. '13, Christian Zempel M.S. '13, Alicia Sherban M.S. '13, and doctoral candidate Jasmine Cashbaugh. Photo by Charles Barry
Adventures with the Robotics Systems Laboratory by land, sea, and sky. And in orbit.

On the third floor of the Bannan Engineering building, an undergrad facing several giant screens monitors a NASA satellite that’s circling the globe. Down the hall, his classmates are designing and building wheeled, flying, and marine robots, which a master’s degree engineer will experiment with for her thesis. A few steps away, Ph.D. candidates working on their dissertations under professor Chris Kitts develop advanced software and control techniques, allowing multiple robots to operate autonomously in complex formations.

“We’ve got everything up here,” Kitts says, “from research desks to a more traditional lab room with soldering irons and electronics all over the place. Up on the roof is our tracking antenna for the spacecraft.” The Robotics Systems Laboratory (RSL) even includes a robot graveyard students use to “cannibalize older robots for parts.”

In the RSL, the coolness factor is never in short supply. Nor is a sense of how this work might apply: The National Academy of Engineering recently recognized Santa Clara’s field robotics program as one of the top “real-world” engineering programs in the nation.

A few projects the robotics teams are running these days: a robotic boat called SeaWasp, a handful of large underwater robots, and a flotilla of robot kayaks—not to mention wheeled robots, robot arms, and a new squadron of quadrotors four-bladed helicopters.

Kitts credits the breadth and depth of the program to a lot of mentoring by folks in the Silicon Valley aerospace community—NASA Ames, Lockheed-Martin, and others—plus grants, contracts, and collaborative projects. Students can take equipment into the corporate setting and put it through tests. “But the design engineering all happens at Santa Clara,” Kitts says.

Most projects are not designated undergrad or graduate student only. Undergraduates might design, build, and learn how to operate a system. Graduate students help with the more sophisticated design challenges, and they also use the robots as experimental platforms to demonstrate their own research techniques.

Up the road at the NASA Ames center at Moffett Field, Kitts secured an unused warehouse (a former home and garden center, to be precise) with a large shop floor and high ceilings that’s ideal for projects like flying those quadrotors around in formations that require complex autopilot software. Wheeled rovers zip around the shop floor, undergoing tests with tracking systems that allow students to perform controlled experiments. And the tethers of the underwater robots—a thousand feet long—can be unrolled and inspected.
 

SeaWasp and the robot kayaks

Deep blue: The Triton robot explores the lakebed of Tahoe with a little help from Giovanni Minelli '06, M.S. '11. Photo by Charles Barry

Then there’s Lake Tahoe. Every May, Kitts and the SeaWasp team head for the lovely alpine lake for some deepwater robotic research. Partnering with geologists from the U.S. Geological Survey and the University of Nevada-Reno, students map Tahoe’s underwater geology, photograph fault lines and features, and collect samples. The geologists bring questions—perhaps about volcanic flows and earthquakes in the distant past; Kitts brings robots and operator know-how to collect the data that answers these questions.

The SeaWasp vehicle automatically sails back and forth throughout a section of the lake where the geologists hope to see something. The robot maps lake bed topography to a half meter, revealing any major geologic features. “Afterward, the geologists look at the map and say ‘Oh, there—that looks interesting. Let’s go look at that,’ then we’ll use an underwater robot to take a look,” Kitts says. Discoveries in recent years have included evidence of an earthquake and landslide-induced tsunami within Lake Tahoe.

The latest RSL project has been to “robotify” kayaks. Student teams acquire a standard kayak hull, affix a trolling motor like the kind you might buy for a rowboat, then add electronics, computers, and the necessary communications equipment.

Repeat a few times, and you’ve got a robotic kayak flotilla—which also ties in with Kitts’ research on multi-robot systems: finding ways to get robots to work together to accomplish tasks more precisely or more cost-effectively. “There are some things that you don’t see robots doing at all that this new technology will enable,” Kitts says. “We have a new technique that allows us to do cost-effective environmental sensing. We can use a group of robots to rapidly ‘sniff out’ the high or low point of a pollutant.”

The practical implications for robot kayaks are manifold. Beyond tracking problems such as oil spills, they can follow algal blooms and find oxygen-depleted regions caused by fertilizer runoff—which lead to “dead” regions of estuaries such as those in the San Francisco Bay Area. Nitrogen or oxygen sensors on the kayaks would allow the fleet to drive straight to a source instead of mapping the whole area to find the highest concentration of a contaminant. Deployed quickly enough, they could pinpoint the problem and even help prevent a massive die-off of marine life.

“It all comes back to this idea of getting more than one robot to work together,” Kitts says. “The application for this technology is ultimately to help people or the environment—that’s the direction we try to go with a lot of our projects.”
 

Rogue robot!

What happens when robots go bad? “We joke that we build a lot of systems that have failures,” Kitts says, “which is great, since fault diagnosis is one of our research areas!”

Multiple robots offer redundancy in case something goes wrong. But low-level safety precautions are built in as well. If the team loses the wireless link to a kayak that’s traveled out of range, the robot will turn off its motors automatically. “With a wheeled robot, it just stops,” Kitts explains. “It’s not gonna go anywhere. But with a kayak, it’ll stop, but the wind may still blow it away or the current may take it. We always bring an extra kayak along so that we can go out and manually retrieve something and bring it back.”

As with the kayaks, the quadrotor robot helicopters require a software architecture that allows them to operate cooperatively.

“You can command things like, ‘I want this cluster of robots to go over here,’” Kitts says. “As an example, we have the Frisbee maneuver, which is: Get in a circle, go in that direction, and spin while you do it. From an individual robot point of view, it’s a very complicated maneuver. But the technique we have allows you to say that very simply, and then the automated control system takes care of all the details.”

This past year has seen increasing talk (and concern about) flying drones being deployed domestically in the United States. While the RSL began experimenting with large radio-controlled planes some years back, that program has been mothballed since the Federal Aviation Administration put a moratorium on autopiloted planes without a human in the loop. But by 2015, the FAA is expected to have guidelines for flying autopiloted aerial vehicles.

Meanwhile, the formations of quadrotors can be flown indoors and out at NASA Ames. When the FAA issues its new guidelines, expect a proliferation of those machines. And when that time comes, Kitts and the RSL want to be ready—with all systems go.

comments powered by Disqus