Mining Energy from ‘Empty’ Air

No matter where you are in the world, you're surrounded by signals: Wi-Fi signals, cellular signals, and other radio frequency (RF) signals are already in the environment. What could you do with them? That is the question ACM Grace Murray Hopper Award recipient Shyam Gollakota—a professor in the University of Washington's Paul G. Allen School of Computer Science and Engineering—set out to investigate. The answers took him from no-power Wi-Fi to a tracking sensor so small it can be worn on an insect's back. Here, he talks about the technology and its many applications.

Ambient backscatter harnesses wireless signals in the atmosphere to generate power and transmit encoded data. How did you come to work on it?

Communication is the biggest power hog in Internet of Things devices. That's because the act of generating electromagnetic signals consumes a huge amount of power. With ambient backscatter, devices communicate information by reflecting the ambient signals in the environment, instead of generating their own signals. The intuition is easy to explain: think about a heliograph, which uses mirrors to produce flashes of sunlight and communicate over large distances.

So the mirrors became antennas in your model, reflecting or absorbing signals from the environment.

Exactly.

Your work showed that it was possible for two battery-free devices to communicate with one another.

It's widely applicable, because by reducing the power consumption of communication, you can extend the life of battery-powered devices—or you can remove the battery and rely on a small amount of solar energy or radio signals to power the device.

You have even used ambient backscatter to create no-power Wi-Fi, right?

You might think that if all you're doing is reflecting signals, you can only send Morse code, but we showed that you can actually generate signals that look exactly like Wi-Fi signals. You can create Bluetooth, Zigbee, LoRa—pick your protocol. It's cool, but it's also very powerful, because now you can start talking to your phone or the millions of devices that are part of existing wireless infrastructure. You can talk to any device that has a Wi-Fi chip without modifying hardware.

"With ambient backscatter, devices communicate information by reflecting the ambient signals in the environment, instead of generating their own signals."

How does it work? Is there an algorithm that controls what gets reflected in order to mimic each signal?

The hardware is very simple. The only capability you have at the hardware level is deflecting and absorbing. Then, in software, you control algorithmically how much you're deflecting or absorbing—and how fast—to mimic whatever type of signal you want to create.

The possibilities for ambient backscatter are endless, and you have worked on a number of them, from battery-free cellphones to contact lens antennae. How do you select projects to pursue?

I think it's one of my responsibilities as a researcher to push the boundaries and demonstrate the possibilities, because once you do that, you can open people's minds and spark the imagination of engineers in each of these different industries.

You also have created startups like Jeeva Wireless to commercialize the technology.

I think the future of IoT is going to be low-power and potentially battery-free. Batteries are difficult to recycle and, in many environments, cumbersome to replace, so I think it's going to be a win for the environment and also fill a very practical need.

More recently, you have done a lot of work on the Internet of Bio-Nano Things (ioBNT), creating small robotic insects as well as camera and sensor systems that can ride on top of real insects.

Biology has a lot to teach us in terms of how we're designing our systems. A dandelion seed, for example, has the perfect structure to fly around in the air. The problem has always been that it takes a long time to design very tiny devices. The prototypes we created use off-the-shelf components to make super-small programmable platforms. Small microcontrollers are connected to an antenna and a backscatter switch or even a small solar harvester, and the whole thing weighs less than a hundred milligrams, which is so small that it can be carried by a bee.

These technologies have been enormously useful for biological tracking research.

There aren't many tools to track insects because of weight and size limitations. We worked with different groups at the university and even with the Washington State Department of Agriculture.

You are referring to the project to locate the nests of invasive giant Asian hornets, or so-called "murder hornets."

We used small wireless sensors to track hornet nests by trapping and tagging insects and looking at the wireless signals. It took a lot of effort, but everything happened within a span of two months, because the sensors are programmable. If you want to accelerate innovation, you need to make sure that the tools are accessible to a large set of people and, importantly, programmable in software. It shouldn't only be possible if you have the right connections and invest $1 million in fabrication.

"If you want to accelerate innovation, you need to make sure that the tools are accessible to a large set of people and, importantly, programmable in software."

You also have worked on several healthcare applications, using smart devices to do things like monitor people's heart rate and breathing.

We call the idea contactless sensing. In 2013, we showed you can use Wi-Fi signals to recognize gestures from anywhere in your house without having to wear a special device. Wi-Fi signals reflect off your body, and machine learning and signal processing algorithms can understand how you are actually moving.

One of the questions that always came up is, "What is a concrete and impactful application for this technology?" So, we started asking ourselves where you might need contactless tracking. Sleep is one example. We showed that you can reflect signals that are generated by your smartphone to track sleep apnea.

In tests, the app you built achieved accuracy levels that were comparable to thousand-dollar polysomnography tests.

The technology was licensed by ResMed, which makes sleep apnea machines. It's been used to track more than 70 million hours of sleep so far in the public.

You've since extended the work to smart speakers.

Smartphones are good, but what is the thing you don't need to plug in every night, that's always just there? For me, that's my smart speaker. What if it could track my sleep? What if it could track my heart rate or tell me whether I'm having a cardiac arrest? We've designed algorithms to do all those things in a completely contactless manner and tested it with real patients. So your smart speaker goes from something that just plays music to something that can transform people's health and save people's lives. I think that's incredible.

Many traditional healthcare companies were slow to recognize the power of digital health technologies. What has your experience been?

When we started off in 2014, mobile health was a big thing within the tech field, but it was not yet a big thing in healthcare. In the last five or six years, there's been a huge amount of change. The new crop of people who are graduating and becoming physicians have all grown up with cellphones. They're tech savvy, and they're questioning the status quo. A lot of change is happening within the community, which is great.

At the same time, there is a higher bar you need to get through when you make a medical claim. You need to get FDA (U.S. Food and Drug Administration) clearances and you need to prove the efficacy. But some of that bureaucracy and due diligence is useful, because you don't want someone to just start claiming that this is going to cure cancer.

How does this work fit into your research agenda for the next few years?

There are two main things I think the computer science community should focus on the next couple of decades. One of them is making technology more environmentally conscious—everything large, like datacenters, and small, like Internet of Thing devices, consumes a large amount of power. The second thing is health. Making devices and diagnostics more cost- and power-efficient is hugely important, particularly in the developing world, as well as in rural, resource-constrained settings.