Bee-Inspired AI Navigates Drones with Low Power

An international team from the Netherlands and Germany developed Bee-Nav, an AI system that lets drones navigate using very little computing power. Researchers at Delft University of Technology and Wageningen University, along with colleagues in Germany, modeled the system on how honeybees return home. The AI uses a small neural network of just 42 kilobytes to learn flights and recognize environments, reducing the need for heavy hardware or GPS.

Bee-Nav combines distance tracking with visual memory, similar to bees using odometry and landmarks. Testing showed a perfect success rate indoors, but outdoor performance dropped to roughly 70% due to wind distorting images. The team believes further refinement will improve real-world robustness.

This technology could enable lightweight drones for tasks like greenhouse monitoring and industrial inspections, helping to increase efficiency and reduce waste. Understanding how bees navigate also provides new biological insights. The work appears in the journal Nature, and researchers plan to continue improving the system’s resilience.