CSL researchers work to improve resiliency of aircraft, power grids

When an airplane faces mechanical issues mid-flight, pilots must account for not only the component failure, but also its potential impact on the rest of the plane. Understanding the potential impact of a malfunctioning part, and the capabilities of the airplane to proceed with the flight or safely land, is the focus of recently published research from the Coordinated Science Lab’s Melkior Ornik and his student, Jean-Baptiste Bouvier.

“If you have a system, an airplane or submarine for example, and part of it is damaged, that part is now producing inputs you don’t control,” said Bouvier, who holds a master’s degree in aerospace engineering from the University of Illinois Urbana-Champaign and is a currently a doctoral student. “The goal of our research is to look at what kind of targets we can reach after the damage. We built a new method for reachability to ensure the aircraft can reach the target despite any undesirable input. If we can measure that undesirable input in real time, then we can add that information to our control.”

Bouvier recently presented the some of his work on the project, “Safety-constrained and efficient learning for resilient autonomous space systems,” at the 21^st IFAC World Congress

under the title, “Resilient reachability for linear systems.” The overall work builds on math theory research done in the ‘60s.

“Our goal is to make the systems more resilient,” said Bouvier. “By ‘resilient,’ we mean the system can lose control of any of its parts and still be able to reach its goal. For a submarine that has three propellers, if it’s resilient, it can lose any one of them and still succeed.”

After determining the resiliency of an aircraft, the pair worked to quantify what makes it resilient in order to be properly prepared for a malfunction. If an airplane is resilient to the loss of one engine, this means the other engine is not only powerful enough to support the plane on its own, but also to counteract the adversarial actions of the malfunctioning engine. Bouvier showed that for a certain system to be resilient to the loss of one actuator, the system needed to have at least two-times the necessary actuators plus one, or 2N+1.

While this specific project has a focus on aerial applications, Ornik believes it can be adapted for power grid or computer network research as well.

“If someone takes over a part of the network, they are able to do with it whatever they want,” said Ornik, an assistant professor in aerospace engineering. “With someone else controlling a part of the network, can we preserve the stability of the network? Can it still perform in a way that works for us? Those are the questions we’re hoping to answer.”

This research is partially funded by NASA.