Imagine you're running and you sprain your ankle. The pain makes you gingerly limp the rest of the way home. This is a great example of how nature adapts to failures in a system. The pain tells you: 'if you continue running like normal, the injury will only get worse'. So you naturally adjust the way you run. Drones currently cannot do this with a worn-out propellor. Researchers from Delft University of Technology and Wageningen University & Research have now demonstrated that a concept we learned from nature, which was originally developed to predict collapse in ecosystems, can also help detect when engineered systems are heading towards failure. This is crucial for ensuring drone and autonomous vehicle safety as they increasingly become part of everyday life.
"You can compare our approach to the way humans experience pain. After an injury, pain provides immediate feedback about our condition and helps us judge what actions remain safe," says Jasper van Beers, researcher at Delft University of Technology. "Machines generally lack this form of self-awareness. The new indicators, derived from real-time measurement data, offer a first step towards giving engineered systems a similar ability to recognise when they are approaching their limits."
From healthy forest to collapse
The study, published in Proceedings of the National Academy of Sciences, applies early warning indicators based on a phenomenon known as critical slowing down. In nature, critical slowing down occurs when a system becomes less resilient and takes longer to recover from disturbances, often signaling that it is approaching a tipping point. For instance, a healthy forest can quickly recover after a dry season. But after each dry period, this recovery starts taking longer. Eventually, a relatively small climate extreme can trigger large-scale forest dieback. Scientists can observe recovery time to see whether an ecosystem is indeed approaching a critical tipping point.
While these methods have been widely used in ecology and climate science, it remained unclear whether they could also work for actively controlled systems such as drones, aircraft and autonomous robots. Unlike natural systems, these technologies are constantly regulated by controllers that respond to changing conditions in real time. Yet the researchers found that early-warning signals from ecology and climate science still reliably indicate when a controlled system is approaching instability.
Detecting Trouble Before It Becomes Critical
To test and validate the approach, the research team worked in the CyberZoo, a unique drone research facility within the Faculty of Aerospace Engineering. There, scientists can safely push drones to the edge of loss of control, deliberately introduce damage, and collect the data needed to understand how failures develop.
By combining simulations, flight data analysis and extensive experimental testing, the researchers were able to identify which combinations of damage, flight conditions and manoeuvres are most likely to result in loss of control. In addition, the indicators can be used not only to detect instability, but also to adapt system behaviour in real-time. For example, they could help determine strategies that maintain flight despite damage to an aircraft's wing—much in the same way humans accommodate ankle injuries by limping in order to keep walking.
Beyond drones
Van Beers: "By bringing together knowledge from different scientific disciplines—in this case aerospace engineering and ecology—we continue to drive breakthroughs that help translate fundamental research into practical technologies."
A key advantage of the method is that it does not rely on detailed physical models of the drone. Instead, it uses data from inexpensive onboard sensors to identify subtle changes in system behaviour. Because of this, the technology could potentially be applied to a wide range of engineered systems beyond drones, as it offers a generic way to monitor resilience and identify problems before they lead to accidents.
Applications could include monitoring critical infrastructure, supporting predictive maintenance in aircraft and other vehicles, improving quality control during manufacturing, and enhancing the reliability of autonomous systems such as self-driving cars. The first real-world impact is expected in the rapidly growing drone sector, where the technology could help prevent accidents as drones are deployed in growing numbers across industries.
