Self-driving vehicles and drones require advanced technology that integrates the computing, or cyber, components as well as the physical components, such as sensors, motors or batteries. Two engineers at Washington University in St. Louis are taking a fresh approach to better design and manage power distribution and energy storage in these and other cyberphysical systems.
Xuan “Silvia” Zhang and Christopher Gill, both faculty in the School of Engineering & Applied Science, received a four-year, $936,504 grant from the National Science Foundation to study how to orchestrate modular power in a modular manner at the mesoscale, an area that has not yet been studied. Mesoscale systems, which can range from mobile devices to the International Space Station, falls between the microscale, which includes on-chip power delivery network in microprocessors, and the macroscale, such as an electric power grid.
“In a mesoscale system, you have a source of power, such as a battery; you have power distribution, such as a power converter or voltage regulator; and finally, you have the components that consume power, such as processors, motors, mechanical components or sensors, and they all behave very differently,” said Zhang, assistant professor of electrical & systems engineering. “This is very different from things at the micro- or macroscale because those are very homogenous systems.”
While having this heterogeneity is a challenge, Gill said it may also work to their benefit.
“Each component’s behavior can be captured and characterized as sort of a signature or a fingerprint in terms of power behavior,” said Gill, professor of computer science & engineering. “This also touches on the behavior of real-time and embedded systems where the timing behavior of each component also is an identifiable fingerprint. We’re really interested in taking these timing signatures and power signatures and reasoning about what are the different combinations of behaviors, which different components that are put together can use to maintain timing guarantees, longer battery life and safe power regulations.”
Gill said there also is unnecessary complexity currently in programming power and timing.
“We hope to create the right set of abstractions so that the power and timing behavior together are more programmable so someone could build a system without having to earn a PhD,” Gill said.
Zhang and Gill bring together their expertise to attack the problem from both the electrical and systems engineering angle and the computer science angle. Zhang and her lab are working on specially designed hardware, and Gill’s expertise in operating systems and other system software will allow them to build kernel modules, or other pieces of code, to orchestrate system behavior.
“Fundamentally, if we can create this principled approach to solve the power distribution and orchestration in the mesoscale system, it potentially has a wide range of applications,” Zhang said. “A lot of the general principle can be broadly applied to a system where the complexity of the interaction between the physical components and cyber components has to be intelligently managed. We need a method to solve this complexity.”
With $25,000 in seed funding they received from an Engineering school-funded Collaboration Initiation Grant in May, Zhang and Gill have already started building some prototypes. They also are working with the university’s Office of Technology Management to seek patent protection for some of their technology.
Zhang and Gill said this framework would be attractive to the robotics industry as well as to companies that design processors and system integration.
As part of the grant, they plan to work with undergraduate students in the Research Experience for Undergraduates (REU) program held in summers in the School of Engineering & Applied Science.