image: Brad Campbell, University of Virginia assistant professor of computer science and electrical and computer engineering view more
Credit: Tom Cogill/University of Virginia School of Engineering and Applied Science
Even the most frugal among us upgrade our cell phones and laptops every now and again. They’re relatively easy to replace.
That’s not true of the internet-connected sensors increasingly installed in buildings and other structures. Small monitoring devices in hard-to-reach places continuously collect and communicate data – everything from temperature and humidity levels in office buildings to the pressure on bridges.
Sensors make up a sizable slice of the broader network researchers call the “internet of things,” comprising the billions of connected products that make our cars, homes, businesses and cities more efficient. These networks promise enormous societal benefits, such as medical sensors that help doctors give patients better care and environmental sensors that help people save energy.
But one big question holds the technology back from reaching its potential: How do you keep all those tiny computers working through the march of time?
“Internet of things devices are often monitoring infrastructure we expect to last for decades, or centuries. We don’t replace all of our buildings every five years,” said Brad Campbell, an assistant professor of computer science and electrical and computer engineering at the University of Virginia School of Engineering and Applied Science.
He has earned a five-year, $700,000 National Science Foundation CAREER Award for his research, “Repurposable Devices for a Greener Internet of Things,” to answer the question.
The CAREER program, one of the NSF’s most prestigious awards for early-career faculty, recognizes the recipient’s potential for leadership in research and education. Campbell is a member of UVA Engineering’s Link Lab, an interdisciplinary cyber-physical systems research center where he is one of several faculty on the leading edge of research focused on how humans can use connected systems to interact with their surroundings.
Consider a smart-lighting system that turns lights on and off in a building depending upon where people are moving around inside. One building may use up to 30,000 sensors. Fully exploiting the good things that connected devices could do means devices could quickly run into the trillions. There’s no practical way to upgrade or replace all of them with the same frequency we do personal devices.
Imagine just trying to keep batteries in the sensors – one of the biggest maintenance headaches of all, and one Campbell has lots of experience with going back to his days doing research for his Ph.D. in computer science at the University of Michigan.
“Not to mention what do we do with all those old devices that were working reasonably well, but they maybe just didn’t keep up with the times? Now those basically become e-waste,” Campbell said, adding that, unlike larger computers and appliances, small electronic devices are difficult to recycle.
“What we need is something that can let us use already deployed devices and keep them useful decades into the future,” Campbell said. “But it’s challenging because the arc of technology doesn’t go backward. We don’t tend to want things with fewer features, or that are less secure or that provide less utility. We always want more, more, more.”
Campbell and his team, including Ph.D. students Nurani Saoda and Nabeel Nasir, are designing a new class of sensors and what they call the “ecosystem” in which the sensors will operate. By “ecosystem” Campbell means software the team is developing to run on existing commercial hardware platforms, such as the Raspberry Pi. The goal is that the sensors would be capable of adapting to whatever the future brings.
The first hurdle is powering the sensors without wires or batteries and without knowing what their surroundings will be in 20 years.
The researchers already know how to design sensors to draw enough energy from nearby sources, such as the sun, indoor lighting or vibrations, to operate. But those sources might not always be there as building occupants and uses change.
“Part of this project is a new design for the energy-harvesting power supply that can encapsulate that complexity, so it can manage what happens when the energy characteristics and needs change,” said Nurani, who is leading the power supply work. “That frees up the application-level processor to focus on the sensing task, which simplifies development and creates more adaptive devices.”
A second challenge is adding software on old networked devices to perform new tasks without crashing existing applications.
“The status quo today is you just update everything,” Campbell said. “You replace all of the code that’s running on your devices, and if something goes wrong, maybe there’s a way to revert back to the old version, or if there’s a bug, hopefully you can update it again.”
The older the hardware gets, the scarier updates become – until operators must ignore any new security issues and are stuck with the device’s current operation.
To solve the problem, Campbell’s team is working on new software architectures to make the devices’ software configuration modular, essentially isolating software components from one another. The components can then be upgraded individually, without reprogramming the entire device.
This software modularity is one of the methods developed from the project Campbell will incorporate into graduate and undergraduate courses for the educational piece of the CAREER Award. He is also looking for graduate students from backgrounds historically underrepresented in engineering and first-year undergraduates to collaborate on the project.
The goal is training engineers skilled in fundamental techniques who understand the intersection of the internet of things and its cross-disciplinary applications, as well as the ethical implications for all stakeholders.
The team will address one more limitation on the functional lifespan of small connected devices: “computational obsolescence.” As hardware improvements make it possible to get more performance and run increasingly complex software using the same amount of energy, older devices’ computing powers diminish relative to the new demands.
Campbell looked to the way wireless communication technology prioritizes backward compatibility. For example, a modern Bluetooth 5 device can still pair with a Bluetooth device from the mid-2000s.
“This suggests that while today’s microcontrollers may not be sufficient for tomorrow’s software, today’s devices will be able to communicate for decades to come,” Campbell wrote in his CAREER Award proposal.
The insight led to the idea of offloading tasks too complex for an aging sensor to a nearby gateway hub capable of doing the job. Because there will be far fewer gateways – roughly five for every 300 sensors – they’re feasible to upgrade or replace as technology evolves.
It’s kind of like using a 2007 iPhone in 2027 – with all the speed and functionality of the latest model.
Nasir is developing the gateway software to support the sensors.
“In effect, it’s as if we have the newest hardware deployed for the devices themselves, but without having to actually replace a huge number of sensors,” Nasir said.
Importantly to Campbell, future application developers who want to add new functions to the sensors should never have think about or even be aware of the transaction between the sensors and gateway. He wants to remove barriers preventing decision-makers across all sectors of society from maximizing the benefits of the internet of things.
“Having these devices not follow the path of smartphones where we replace them all the time, I think is important,” Campbell said. “But enabling better-operated equipment, better-managed buildings, more efficient infrastructure is also important.
“We need to find ways to not just have the individual technology pieces or the components or the devices themselves, but the larger ecosystem actually scalable,” he said. “That’s what led to this idea of repurposable devices.’