When you sprain your ankle in the middle of a run, your body sends a pain signal to your brain, forcing you to stop. Essentially, the ability to sense pain stops you from pushing through the injury and causing further self-harm.
Researchers at Delft University of Technology and Wageningen University have applied this exact concept to drones, giving them a digital equivalent of a nervous system that recognizes a faulty part and triggers a pain-like warning signal. What’s even more interesting is that the technology could find use in self-driving cars.
Unsplash
So how does the “pain” system actually work?
The team developed early warning indicators, something they call “critical slowing down” signals, borrowed from a concept originally used to predict ecosystem collapse in ecology. Their study is published in the Proceedings of the National Academy of Sciences (via TechXplore).
Any complex system, biological or engineered, begins to show subtle changes in its sensor data before it actually fails. This particular system also detects those changes, using only real-time data, without needing predictive models or historical baselines.
They tested it on quadrotors at the CyberZoo drone research facility by incrementally damaging rotor blades from healthy up to 55% tip damage. In their testing, loss of control occurred at 15% blade-tip damage on the front-right rotor, and the system successfully flagged the instability as it gradually built up.
“You can compare our approach to the way humans experience pain,” said lead researcher Jasper van Beers. “After an injury, pain provides immediate feedback about our condition and helps us judge what actions remain safe. Machines generally lack this form of self-awareness.”
Cruise
How could this help your car?
The same concept translates to autonomous vehicles and advanced driver-assistance systems, especially the ones deployed commercially as robotaxis.
A self-driving car dealing with a degrading sensor, a failing actuator, or unfavorable road conditions pushing it toward its handling limits faces the exact same problem. It has no way to feel a warning before it loses control.
Since the system works on real-time data alone, it doesn’t require any retrofits or new hardware: it processes what’s already there. The researchers explicitly mention self-driving cars as a target application, which sounds quite appealing to me.
