Two Clemson University researchers reap $1 million in funding
CLEMSON — Two Clemson University faculty members are receiving a grand total of $1 million in funding as part of the nation’s highest honor for scientists and engineers in the early stages of their research careers.
Jacob Sorber and Yue “Sophie” Wang were among the honorees in this year’s National Science Foundation’s Faculty Early Career Development (CAREER) Program. They are working on separate projects and each has been awarded $500,000 for research.
Sorber’s research enables low-cost, low-power sensors to gather data for long periods of time. The sensors would be powered by energy from environmental sources, such as the sun, with no need for batteries or manual recharging.
He said the sensors have the potential to transform science and society. They could, for example, but used to monitor human health, growing conditions in greenhouses or the behavior patterns of animal populations in the wild.
Wang is focusing on two distinctly human attributes — trust and regret — to develop new “control algorithms” and decision-making strategies that would help humans and robots work together to be more productive. She sees big opportunities for humans and robots to collaborate in manufacturing.
Wang also sees high potential for “human-supervised mobile sensor networks.” Robots could begin doing low-level simple and repetitive tasks while humans could be involved in high-level complex tasks, she said.
Anand Gramopadhye, dean of the College of Engineering and Science, said the awards are a clear testament to the hard work and creative ideas that Sorber and Wang bring to Clemson.
“We are very proud of them both,” Gramopadhye said. “The award means that two of the nation’s brightest emerging researchers are here at Clemson. The funding will enable Jacob and Sophie to develop significant programs for maximum impact. It’s a job well done for both.”
While research is central to the award, winners must also be excellent teachers. Awardees have proven themselves exemplary in integrating research and education. Selection is highly competitive.
Clemson now has multiple NSF CAREER award-winners. Among the more recent winners is Sandra D. Ekşioğlu, an associate professor of industrial engineering. She won her award at Mississippi State before moving to Clemson.
Here is more on the research Sorber and Wang are doing:
Jacob Sorber, assistant professor, School of Computing
Sorber’s award is for part of his quest to design low-cost, wireless sensors that can operate maintenance-free for decades without a battery. They would be powered solely by energy harvested from the sun, vibrations and other environmental sources.
“I envision devices that are super small, super cheap that you can deploy for 50 years,” Sorber said. “I want the lifetime of my sensing devices to be limited by the natural decay of the hardware. When do the chips stop working? I’ve never run a computer that long.
“Today, the battery always wears out first, and replacing batteries is not sustainable in many applications, and expensive and tedious in others.”
But eliminating the battery makes energy flow unpredictable, which poses enormous challenges for application designers. The device is prone to shutting down and rebooting, sometimes several times a second.
A computer that reboots frequently may not have time to complete a task before its next power failure. When power returns, it may not remember what it was doing when it shut down and valuable computation and data can be lost. In the worst case, the device may not ever complete the task it was programmed to do.
Sorber is proposing a new computing platform, called Mayfly, that will make it easier to develop applications for devices that fail frequently. Like the short-lived insects, Mayfly devices will “come alive” for short periods of time.
Even if there are frequent power failures, Mayfly’s new hardware and software techniques will help devices make consistent progress toward finishing its tasks.
Mayfly will include a new programming language to help sensors figure out how much power is available and prioritize how to use it so that the most important functions get done first.
Mayfly’s hardware will also prioritize and manage harvested energy automatically. The hardware will simplify software decisions and improve harvesting efficiency, allowing the device more time to do useful work.
Sorber is also developing new ways of programming computers to predict when they will crash and save just enough information so that they continue where they left off when they come back on.
“We’re trying to develop a system that makes it easy for programmers to write these kinds of programs,” Sorber said. “Right now, this is really a nightmare for a programmer. It’s possible, but extremely difficult.”
The Sorber team will test prototype sensors in the university’s greenhouses in hopes of transforming how greenhouse gases are monitored.
As many as 30 sensors will measure ambient light, temperature, humidity and leaf wetness and send the readings to a “gateway device” that will upload them to the Internet, where they will be made available to horticulture researchers.
Sorber also sees big possibilities in wildlife tracking and plans to deploy Mayfly sensors to help with two separate studies.
One is an ongoing study of the social patterns of the desert tortoise in Nevada. Another is an early-stage effort to study invasive mongooses in the dense subtropical brush of Sandy Point National Wildlife Refuge on St. Croix.
The sensors will allow biologists to unobtrusively monitor animal activity levels, movement patterns and socialization. Sorber’s team hopes to help biologists better understand how genes flow through tortoise populations and how mongoose populations affect local fauna and sea turtle nesting sites.
Sorber plans to tightly integrate his research with graduate and undergraduate education throughout the project. He will use his research to offer hands-on computing experience to students at R.C. Edwards Middle School in Central. The curriculum could also be used by other schools.
“The ideas, tools and techniques we’re developing will have a lasting impact only if they are adopted by the next generation of engineers and scientists,” Sorber said. “Mayfly-style computing is forcing us to rethink a number of time-honored computing techniques. I have some ideas, but I’m hoping that younger minds with fresh insights will help fill in the remaining gaps.”
Eileen Kraemer, the C. Tycho Howle Director of the School of Computing, said, “Harvested energy to support mobile sensing is a giant step forward, but the complexity of programming such systems could be overwhelming. The tools that Dr. Sorber proposes to create will make implementation manageable by programmers. We are lucky to have such a dedicated and enthusiastic researcher as Dr. Sorber here at Clemson.”
Yue “Sophie” Wang, assistant professor, mechanical engineering
Yue “Sophie” Wang is working on new “control algorithms” and decision-making strategies that would help humans and robots work together to be more productive. But she is looking squarely at two human traits to make it happen.
Trust and regret.
Her lab is a tech enthusiast’s dream. Wang has an obstacle course for flying drones, the kind of robotic arm you would find in a manufacturing plant and small vehicles that are programmed to follow each other like a train but without the couplings.
For her research on trust, Wang wants to develop algorithms that will help workers determine which tasks they should do and for how long and when they should hand the job to robots.
Humans can often do more complicated and flexible jobs than robots, but when humans work too long and are under stress, their performance fades.
Figuring out when a human is too tired or stressed to work can often be a judgment call, but Wang is developing models that quantitatively measure performance in real time as humans and robots work collaboratively.
“We will show this information on the computer screen and provide suggestions: ‘Now is the time you should take over or now is a time you should rest,’” Wang said.
“The workload balance between the human and robot is governed by the trust that the human has for the robot. That’s key as we determine how much autonomy the robot has.
“We are going to mathematically model the trust. The models will be based on how much prior trust the human had for robot, the rate of improvement in performance, as well as the rate of decrease in the number of mistakes of both human and robot collaborators.
“These will help us develop what we researchers call ‘trust-based control algorithms.’ They will help take some of the burden off humans and guarantee top performance.”
Wang is also working to make robots seem more human by teaching them regret.
No human likes to feel regret, but it’s how we learn. Wang wants robots to have the same experience, except they will base their regret on mathematical formulas instead of feelings.
At each moment, robots need to decide between tasks, Wang said.
“They need to think, ‘What if I choose task A? How much will I lose? How much will I gain?’” she said.
As part of the decision-making strategies Wang is working to develop, the robot would calculate the probability of each event and make a decision based on the difference. The goal is to minimize the possibility of a regretful decision.
For example, a robot may need to pick up a square part but isn’t sure if the part is square or rectangular due to the robot’s camera position. The robot would calculate the risk in picking up the wrong shape.
If it’s worth risking the regret in picking up a wrong shape, the robot would go ahead and pick up the part.
The “trust-based control algorithms” and “regret-based decision-making strategies” will be programmed into a humanoid manufacturing robot, as well as a “heterogeneous multi-robot testbed.”
“They will be programmed into autonomous robots so that humans and robots can better collaborate in assembly,” Wang said. “Our work could also be used in human-supervised mobile sensor networks.”
To help advance the field, Wang is using her research and lab in a new undergraduate course, “Human-Robot Collaborative Manufacturing using Humanoid Robots.” She also plans to offer a new graduate course on control of human-robot collaboration systems.
Wang has also participated in outreach activities, including the Math Excellence Workshop for incoming freshmen and a summer reading program for K-12 students at the Pickens County Library.
Among those congratulating Wang was Melur K. “Ram” Ramasubramanian, the chair of the mechanical engineering department.
“I’m happy to hear that a mechanical engineering faculty member has won an NSF CAREER award,” he said. “This leaves no doubt we have some of the brightest young minds here at Clemson. No one is more deserving than Sophie.”
This material is based upon work supported by the National Science Foundation under award numbers 1453607 and 1454139. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.