Archives

Researchers are working to make sure a seeing, hearing, door-opening robot called “ANYmal” can also function in extreme conditions—a mission that takes them to the labyrinth of drains and tunnels below Zurich.

Their aim is to determine whether ANYmal—a robot that ANYbotics, an ETH Zurich spin-off company, jointly developed with Robotic Systems Lab—could one day be useful in sewer systems. It might, for instance, help city of Zurich employees who regularly have to walk or crawl through the some 62 miles of accessible shafts and drains underneath the city and whose job it is to check the walls and floors for damage.

ANYmal robot — ANYmal on its first tour of inspection in Zurich’s sewer system (Credit: Daniel Winkler/ETH Zurich)

This work not only poses a health risk, but is also potentially lethal, given that the drains can fill up with water very quickly without warning. Another advantage of robots in such an environment is that they could operate in narrow sewers that are not accessible with current technology.

Going down

The researchers place the robot upright at the bottom of the shaft. It is about 19.5 inches tall and has four articulated legs as well as something resembling a head that consists of a camera and various sensors.

Peter Fankhauser, cofounder of ANYbotics, radios his colleagues on the surface, who are responsible for coordinating the test and sending commands to the robot. Fankhauser then twiddles with a joystick and the robot plods forward. As this is the first test in unknown terrain, he takes partial control of the robot even though it is capable of moving autonomously.

“It’s a precautionary measure,” says Fankhauser, “Just because something works in the lab doesn’t always mean it will in the real world.”

ANYmal — (Credit: Daniel Winkler/ETH Zurich)

After all, the conditions underground are not what the robot is used to: the chamber is wet and slippery, with lower temperatures and higher humidity than in the lab. What’s more, it’s very, very dark.

“It’s hard to distinguish much down here,” says Fankhauser, almost with a hint of resignation in his voice, as the robot moves at a slow pace through the roughly 10-foot by 16-foot tunnel.

The robot emits a uniform electromechanical sound—a kind of rhythmic whirring—that blends with the sound of rushing water emanating from the main sewer nearby. Given that the robot is on its maiden test run 13 feet below ground level, the researchers have taken the precaution of avoiding large volumes of water.

In the dark

The goal of the three-year research project, called THING (sub-Terranean Haptic InvestiGator), is to design robots that can move about on their own and are better able to identify their surroundings.

Robots generally use 3D cameras and laser sensors for orientation. But such devices can malfunction in adverse conditions—such as when the ground surface is wet or the air full of dust. That’s why the researchers consider enhanced haptic perception—orientation by touch—to be a possible solution.

Researchers at other institutions are also working with and testing ANYmal. In addition to the tests in the sewerage system, next year researchers will deploy the robot in a Polish copper mine. That will determine whether it can function in an entirely different microclimate, one characterized by hot, dusty air and gravel surfaces.

One of the key questions on this first day of testing is whether the robot can find its way around at all in the darkness of the sewerage system. Initially, two helpers with big LED lamps illuminate the surroundings so that the researchers can clearly see what’s going on.

Then, Fankhauser asks the helpers to turn off the lamps and radios his colleagues on the surface to tell the robot to use its own lights. The robot’s sense of touch isn’t the only thing that helps it find its way in the dark, as Marco Hutter, who has been conducting research into legged robots for many years at ETH Zurich, explains: “The robot uses laser sensors and cameras to scan its surroundings. By identifying irregularities in the surface of the concrete, it can determine where it is at any given moment.”

All that the researchers can see in the darkness now are the small round LEDs in the robot’s “head.” The atmosphere is other-worldly: the darkness, the sound of rushing water, the electromechanical whirring, the robot’s LED eyes. Then someone breaks the eerie silence with a droll comment: “Its eyes are a bit like a Rottweiler.”

ANYmal’s future missions

Researchers completed the first ANYmal prototype was completed in 2015. They carry out on-site tests two or three times a month. For instance, Fankhauser and some members of his team recently headed to an offshore platform in the middle of the North Sea. The hope is that robots could one day perform inspections on such platforms. On its pilot run at least, ANYmal autonomously completed several inspection routes with flying colors.

After almost ten years of research, there’s a lot ANYmal can do. It can not only walk autonomously, but also boasts the sensory capabilities of sight, hearing, and touch. These enable it, for instance, to read the air pressure display on a machine, identify sounds, and recognize objects—for example to determine whether or not a fire extinguisher is in the right place. The robot can even perform certain manual tasks on its own.

Equipped with an additional gripping arm, it can open doors, dispose of refuse, or press a lift button. It also delivers data that is more precise than our own eyes, ears, and noses can perceive. It can identify the ambient temperature and detect the presence of gases in the air. Its latest trick is recognizing the composition of the ground beneath it. “Some of its powers are superhuman,” says Fankhauser.

Despite the lack of light in the sewer, the robot seems to be finding its way quite well, plodding through the shallow channel at a leisurely pace. When the high-tech machine reaches a 20-centimeter-high ledge (just under 8 inches) in a dry side arm of the sewer, Fankhauser brings it to a halt with a flick of the joystick. Initially, he is reluctant to give the robot the command to climb over the ledge. Although it has easily mastered this maneuver in laboratory conditions, down here it is a risky undertaking.

“It’s an expensive machine,” says Fankhauser. But he gives it a try anyway. ANYmal doesn’t manage it at its first attempt. It stops at the ledge like a horse balking at a jump. “Default, start again,” radios Fankhauser. Now the robot elegantly places one leg after another over the ledge.

Watching from the surface

Sitting on a bench under a white canopy on the surface, two assistants have their eyes firmly fixed on a laptop. A generator is buzzing and a router is blinking—and many a cyclist passes by along the main road looking on in bemusement at the hubbub around the open manhole at the side of the road.

Looking over the researchers’ shoulders, you can see an almost constant stream of data flickering across the screen. And thanks to state-of-the-art 3D and laser technologies, live images the robot constantly transmits from underground are visible on a separate monitor.

When Fankhauser radios from below that he wants the robot to touch the wall of the sewer with one of its legs, the two assistants have their work cut out for them. The researchers haven’t programmed software they are using hasn’t been programmed for this. They respond quickly, however, taking an algorithm originally programmed to teach ANYmal to shake hands. But to make sure the robot doesn’t hit the wall with force, the researchers have to adapt the parameters.

In this case, the problem is the angle at which the robot is to raise its leg. One of the assistants types in 100 and then gradually ratchets up the number. At 180 the perfect level is reached and the robot’s maneuver is successful.

“The robot was in nonstop operation and collected a lot of data,” says Fankhauser as he undoes his high rubber boots and removes his protective clothing.

Hutter is satisfied, too: “All the teams will be taking home a huge volume of data to incorporate in their research.”

They are now one step closer to their goal of delivering a robot that can function properly in challenging conditions underground. But their work is far from finished. The robot recorded 500,000 measurements per second over the course of the day. “That’s enough data to keep us busy for six months,” says Fankhauser with a laugh.

Hutter received support from ETH soon after embarking on this research in the form of an ESOP scholarship and a Pioneer fellowship.

Source: ETH Zurich

Bees wearing backpacks could replace some drones

‘Smart skin’ warns of strain in bridges and airplanes

Sound and vibrations let smart devices know where they are

Methods that analyze sounds and vibrations can make smart devices more aware of their surroundings, according to new research.

“A smart speaker sitting on a kitchen countertop cannot figure out if it is in a kitchen, let alone know what a person is doing in a kitchen,” says Chris Harrison, assistant professor in Carnegie Mellon University’s Human-Computer Interaction Institute (HCII). “But if these devices understood what was happening around them, they could be much more helpful.”

Harrison and colleagues in the Future Interfaces Group worked on two approaches to the problem—one that uses the most ubiquitous of sensors, the microphone, and another that employs a modern-day version of eavesdropping technology used by the KGB in the 1950s.

Borrowing from Hollywood

In the first case, the researchers have sought to develop a sound-based activity recognition system, called Ubicoustics. This system would use the existing microphones in smart speakers, smartphones, and smartwatches, enabling them to recognize sounds associated with places, such as bedrooms, kitchens, workshops, entrances, and offices.

“The main idea here is to leverage the professional sound-effect libraries typically used in the entertainment industry,” says Gierad Laput, a PhD student in HCII. “They are clean, properly labeled, well-segmented, and diverse. Plus, we can transform and project them into hundreds of different variations, creating volumes of data perfect for training deep-learning models.

“This system could be deployed to an existing device as a software update and work immediately,” he adds.

The plug-and-play system could work in any environment. It could alert the user when someone knocks on the front door, for instance, or move to the next step in a recipe when it detects an activity, such as running a blender or chopping.

The researchers began with an existing model for labeling sounds and tuned it using sound effects from the professional libraries, such as kitchen appliances, power tools, hair dryers, keyboards, and other context-specific sounds. They then synthetically altered the sounds to create hundreds of variations.

Laput says recognizing sounds and placing them in the correct context is challenging, in part because multiple sounds are often present and can interfere with each other. In their tests, Ubicoustics had an accuracy of about 80 percent—competitive with human accuracy, but not yet good enough to support user applications. Better microphones, higher sampling rates, and different model architectures all might increase accuracy with further research.

Good vibrations

In a separate paper, PhD student Yang Zhang, along with Laput and Harrison, describe what they call Vibrosight, which can detect vibrations in specific locations in a room using laser vibrometry. It is similar to the light-based devices the KGB once used to detect vibrations on reflective surfaces such as windows, allowing them to listen in on the conversations that generated the vibrations.

“The cool thing about vibration is that it is a byproduct of most human activity,” Zhang says. Running on a treadmill, pounding a hammer, or typing on a keyboard all create vibrations that can be detected at a distance.

“The other cool thing is that vibrations are localized to a surface,” he adds. Unlike microphones, the vibrations of one activity don’t interfere with vibrations from another. And unlike microphones and cameras, monitoring vibrations in specific locations makes this technique discreet and preserves privacy.

This method does require a special sensor, a low-power laser combined with a motorized, steerable mirror. The researchers built their experimental device for about $80. They affixed reflective tags—the same material used to make bikes and pedestrians more visible at night—to the objects they wanted to monitor. They can mount the sensor in a corner of a room and can monitor vibrations for multiple objects.

The sensor can detect whether a device is on or off with 98 percent accuracy and identify the device with 92 percent accuracy, based on the object’s vibration profile, Zhang says. It can also detect movement, such as that of a chair when someone sits in it, and it knows when someone has blocked the sensor’s view of a tag, such as when someone is using a sink or an eyewash station.

The researchers presented their findings at the Association for Computing Machinery’s User Interface Software and Technology Symposium in Berlin. The Packard Foundation, the Sloan Foundation, and Qualcomm supported the work on Ubicoustics and Vibrosight, with additional funding from the Google PhD Fellowship for Ubicoustics.

Source: Carnegie Mellon University