Roberto Furfaro
Roberto Furfaro
Tyler O'Neal, Staff Editor SYSTEMS

UA Prof. Furfaro wins $4.5M to develop AI-powered hypersonic guidance, navigation systems

As countries around the world work to advance weapons traveling at Mach 5 and faster, a team led by University of Arizona experts builds a "brain" for high-speed vehicles and interceptors.

Roberto Furfaro, a University of Arizona professor of systems and industrial engineering, has been awarded $4.5 million to lead the development of improved guidance, navigation, and control systems for autonomous vehicles operating at hypersonic speeds. The three-year proposed research is sponsored by the Joint Hypersonic Transition Office through the University Consortium for Applied Hypersonics (UCAH).

Hypersonic speed – Mach 5 or higher – is the speed that exceeds five times the speed of sound. As the United States works to develop hypersonic technologies, research in the field has never been more important.

"Many conventional systems are designed using linear theory, and are not designed to fly or intercept at that speed," Furfaro said. "There are a lot of things happening in hypersonic flow that are so nonlinear that they are not fully understood, and that we need to characterize if we want to design systems that work under these conditions."

Consider how, when a car is moving at 80 mph, a one-second delay in the driver's decision-making can have catastrophic results. Hypersonic vehicles, which travel thousands of miles per hour and face additional factors such as shockwaves and extreme heat, have even less room for error.

UArizona is home to the Arizona Research Center for Hypersonics, where researchers conduct supercomputer simulations and wind tunnel tests to learn more about how vehicles behave in extreme environments. The lab develops and employs novel CFD codes for hypersonic vehicle simulations. 

The artificial intelligence-powered guidance, control, and navigation methods Furfaro and his team develop will act as the "brain" of hypersonic vehicles – including interceptors, which are high-speed, maneuverable vehicles designed for defense against enemy aircraft.

"This investment is a major win for our burgeoning hypersonic research program," said David W. Hahn, the Craig M. Berge Dean of the College of Engineering. "Roberto has a broad range of expertise in areas including space flight mechanics and machine learning, making him and his team exceptionally well qualified to lead this effort."

To train hypersonic systems to navigate and react to extremely complex, high-speed situations on their own, the team is using a type of machine learning called meta-reinforcement learning.

"With meta-learning, we can train it not only on one scenario but on many scenarios," Furfaro said. "The system is able to learn over a distribution environment, and every time it converges faster to the next one. By enabling this continuous learning, we are basically able to have a system that continually adapts."

A strong team builds a test environment

UArizona alumnus Brian Gaudet, a research engineer in the university's Space Systems Engineering Laboratory, is playing a critical role in developing and implementing the AI system. Other collaborators and co-investigators include aerospace and mechanical engineering professor Samy Missoum, who is working on Department of Defense-funded work to characterize hypersonic environments; and materials science and engineering professor Erica Corral, who serves as the co-director of industrial and national lab engagement and workforce development for UCAH's Consortium Engagement Board.

Furfaro will also work with aerospace and mechanical engineering faculty members Alex Craig and Jesse Little, who work in experimental aerodynamics, and Kyle Hanquist, an assistant professor in the same department who specializes in computational fluid dynamics. Other collaborators are at the University of Texas at Austin and Raytheon Missiles and Defense.

"The University of Arizona has a nationally prominent hypersonics research program, which received $10 million in federal and state support in 2021 to enhance our research facilities," said University of Arizona President Robert C. Robbins. "Many of the field's top experts agree that artificial intelligence will play an increasingly important role in the advancement of the field, and Professor Furfaro's receipt of this highly competitive grant will bring together many areas of expertise to advance this critical area."

The researchers will use this data – gathered from simulations and wind tunnel tests about how vehicles behave in hypersonic flow – to characterize and create a simulated environment for training the adaptive brain of the system.

"We're incredibly supportive of the University of Arizona's work in developing hypersonic technologies and talent," said Wes Kremer, president of Raytheon Missiles & Defense. "The advancements Professor Furfaro and his team will make to guidance, navigation, and control systems will directly impact our nation’s ability to develop advanced hypersonic capabilities."

UArk prof Panda wins grant to improve recovery of critical systems after cyber attacks

Tyler O'Neal, Staff Editor SYSTEMS

Brajendra PandaThe National Centers of Academic Excellence in Cybersecurity, located within the National Security Agency, awarded $637,223 to Brajendra Panda, a professor of computer science and computer engineering, to improve recovery methods for critical infrastructure systems following a cyber-attack. 

Critical infrastructure includes things like the power grid, gas and oil pipelines, military installations, and hospitals. An example of a recent attack on critical infrastructure is the ransomware attack on the Colonial Pipeline last year. The attack on computerized equipment left the pipeline down for six days while the company forked over an estimated $4.4 million in Bitcoin to pay the attackers (though much of it was subsequently recovered). 

In his proposal, Panda notes that the interdependence and interconnection of CI systems make them more vulnerable to cyber-attacks and can cause initial damage to spread quickly to other systems. “Thus, a small vulnerability in one of these systems can result in crippling a large number of them,” Panda noted. “These systems are of heterogeneous type by nature, meaning they contain both heterogeneous software and data.”

Due to the complexity of CI systems, recovering them can cause significant delays, which is concerning given the time-sensitive nature of the functions these systems provide, such as electricity. 

Panda’s goal is to develop fast, accurate, and efficient recovery mechanisms that, when coupled with the expeditious damage assessment techniques he has already developed, will offer an “integrated suite solution.” This will allow affected CI systems to continue running while providing as many critical functionalities as possible.

The two-year grant, with an option for a third year, builds on a previous $287,000 grant from the same funding agency that focused on expediting the assessment of damage following a cyber-attack. 

German physicist presents a theory that enables simulation of biological systems in a matter of minutes instead of months

Tyler O'Neal, Staff Editor SYSTEMS

For many processes important for life such as cell division, cell migration, and the development of organs, the spatially and temporally correct formation of biological patterns is essential. To understand these processes, the principal task consists not in explaining how patterns form out of a homogeneous initial condition, but in explaining how simple patterns change into increasingly complex ones. Illuminating the mechanisms of this complex self-organization on various spatial and temporal scales is a key challenge for science. So-called “coarse-graining” techniques allow such multiscale systems to be simplified, such that they can be described with a reduced model at large length and time scales. “The price you pay for coarse-graining, however, is that important information about the patterns on small scales – like the pattern type – is lost. But the thing is that these patterns play a decisive role in biological systems. To give one example, they control important cellular processes,” explains Laeschkir Würthner, a member of the team led by LMU physicist Prof. Erwin Frey and the lead writer of a new study that overcomes this issue. In collaboration with the research group of Prof. Cees Dekker (TU Delft), Frey’s team has developed a new coarse-graining approach for so-called mass-conserving reaction-diffusion systems, in which the large-scale analysis of the total densities of the particles involved enables the prediction of patterns on small scales.

The scientists illustrated the potential of their approach with the Min protein system, a paradigmatic model for biological pattern formation. The bacterium E. coli uses various Min proteins circulating in a cell to determine at which location cell division takes place. A decisive factor here is that the proteins involved occur at different frequencies depending on their location in the cell and chemical state – which is to say, they have a variety of different densities. “We’ve now managed to reduce the complexity of this system by developing a theory that is based solely on the total densities of the proteins, such that we can completely mirror the dynamics of pattern formation,” says Frey. “This is a huge reduction. The numerical computations are now accomplished in a matter of minutes instead of months.”

The researchers were able to experimentally confirm theoretical predictions of the model, according to which distribution of the proteins depends on the geometry of the environment. They did this by reconstructing the Min protein system in an in-vitro flow cell, with the results showing the same protein patterns as were revealed in the simulation. “Such reconstruction of information at a small length scale from reduced dynamics at the macroscopic level opens up new pathways for a better understanding of complex multiscale systems, which occur in a broad range of physical systems,” says Frey.