Battery-free, energy-harvesting perpetual machines: the weird future of computing

Custom designed to run entirely without batteries, the hand-held gaming device is powered by small solar panels as well as the button presses of the person playing it. That’s right: Even after the apocalypse, survivors will at least have “Tetris."

The implications of this demonstration are potentially huge, and not just for videogame junkies. In our battery-free future, carbon, moisture and light sensors that last for decades could be scattered by drones across farms; smart cities might be inundated with all-seeing, all-hearing surveillance devices; vehicles and buildings will use artificial intelligence to anticipate needs and perform simple tasks; and “implantables" in our bodies will more tightly integrate humans with everything else connected to the internet.

Nvidia Corp. Chief Executive Jensen Huang has predicted this future of computing will eventually include trillions of devices. “I hope for God’s sake they’re not all powered by batteries," says Josiah Hester, an assistant professor of computer engineering at Northwestern University, and a co-lead on the Game Boy project.

There are many practical and environmental reasons to hope for battery-less sensors, in everything from bridges (to monitor their safety) to human bodies (to monitor our health). But battery or no, a key concern is what happens to a sensor’s data when it runs out of power.

To address this problem, the Game Boy research team upended a fundamental rule of computers: If you turn it off, you lose unsaved work. Their system, by contrast, can lose power completely, even many times a second, and the instant it gets enough power again—say, from a player impatiently mashing buttons—it picks up right where it left off.

Known as “intermittent computing," this system relies on a still-exotic kind of memory chip. Almost every computer in history has had two separate forms of memory: volatile RAM and more permanent, but harder to access nonvolatile storage, which includes anything from punch cards and magnetic tape to hard drives and flash memory. But these researchers are using a new type of RAM—ferroelectric RAM or F-RAM—that erases the distinction. It’s as quickly and easily accessible as typical RAM, but as persistent as any permanent storage medium. It also takes only a minuscule amount of electricity to make it work, and it doesn’t degrade over time, like flash memory does.

Jasper de Winkel, a Ph.D. candidate at Delft University of Technology in the Netherlands, and the technical lead on the batteryless Game Boy project, married this power-sipping, nonvolatile memory to a power-sipping processor from Ambiq, a 10-year-old Austin-based company that specializes in processors for smartwatches, industrial sensors and other ultralow power devices.

The total package—including the memory, processor and display—draws on average 11.5 milliwatts of power. This makes it, according to the researcher’s calculations, about 20 times more power efficient than the original Game Boy from 1989. By comparison, a typical smartphone draws 1 to 3 watts of power from its battery when in use, or around a hundred times more power.

It’s this combination of traits—never needing to reboot, using very little power, and harvesting energy from the environment—that yields a system that could be a “perpetual" computer, says Dr. Hester. The goal of perpetual computing is tiny sensors, radios and other devices that gather, process and transmit data until at last they physically break down. Not to be confused with a mythical “perpetual motion machine," these could be very real additions to our environment, scattered across earth and sea, providing an infrastructure of data collection that could outlive its creators.

Energy harvesting hasn’t improved dramatically in the past few decades, but “what has changed is what you can do with these very tiny amounts of energy," says Joshua R. Smith, a professor at the University of Washington, where he heads up the Sensor Systems research group. (He wasn’t involved with the Game Boy project.)

Dr. Smith and his collaborators have demonstrated it’s possible to use the radio waves already coursing through our environment to power tiny sensors and computers. In 2005, his group was the first to show off a small microcontroller powered by radio waves beamed over a considerable distance. In one of the lab’s latest projects, the team wirelessly powered a small, batteryless video camera. In early 2021, Jeeva Wireless, a startup founded in order to commercialize the underlying technology, will release its first chip.

As enticing as this technology sounds, it will always be limited, especially compared with the ever more powerful computers we’ve grown accustomed to carrying around. It’s just physics: Tiny systems that use very little power might someday become clever in the way the genius of the invertebrate world, the fringed jumping spider, has managed to cram an impressive amount of smarts into a small body, but it’s not about to build a spider civilization and put other spiders on the moon.

“Something like your phone is probably always going to have a battery," says Dr. Smith. “But maybe when that battery runs out, it will still be usable in a reduced-functionality mode, using energy harvesting."

Still, with a growing array of processors and sensors that can sustain themselves on as little as a few hundred microwatts of power—less than half of the power generated by a house fly in flight—the number of possible ambient energy sources multiplies significantly.

In addition to existing ways to harvest energy, from radio waves, solar power and vibration, there are some in development that sound more sci-fi. For example, researchers at Northwestern recently demonstrated a novel thread that can turn body heat, or any thermal energy, into electricity. The result could be, for example, a hat that powers health sensors, or a ski jacket that trickle-charges your phone. Other researchers at a variety of institutions are working on ways to produce electricity from the microbes that live in soil.

And when you combine multiple energy sources, you get a package that could go places few computers have before—like inside construction material. Researchers have previously proposed putting wireless sensors into freshly poured concrete, where they could monitor strain more or less indefinitely.

“A bridge is supposed to have a lifetime of 50 years, and in the U.S. we’ll leave it up for 200 years, because that’s how we do infrastructure," says Dr. Hester. “Imagine getting stress and strain data at high resolution across a bridge for that entire time."

In other words, someday we might know it’s time for a repair when the bridge itself cries out for help.

—For more WSJ Technology analysis, reviews, advice and headlines, sign up for our weekly newsletter.

Write to Christopher Mims at christopher.mims@wsj.com