Stable as bumblebees: new algorithm helps flapping-wing micro-robots keep balance in wind gusts

Researchers from the Moscow Institute of Physics and Technology (MIPT) and Skoltech have created an algorithm for stabilizing the flight of miniature robots with flapping wings, along with a computational environment for testing various control algorithms in gusty winds. Bumblebees are among the most maneuverable creatures in nature, thanks to vortex aerodynamics where each wingbeat forms stable vortex structures that generate lift. However, flapping flight is inherently unstable, and transferring this aerodynamics to a robot or its digital model causes it to lose balance at the slightest wind gust. The new stabilization system controls three key wing parameters: wingbeat amplitude, mean trajectory position, and wing rotation angle around its own axis. The algorithm uses a nonlinear controller based on a discrete Lyapunov function to calculate changes in wing kinematics that reduce deviation from the desired course. The algorithm was tested in a digital wind tunnel with strong vertical gusts: a linear controller could only hold the insect if wind force did not exceed 50-60% of body weight, while the new hybrid algorithm stabilized flight even when gusts were 1.5 times the object's weight, with stabilization time of 3 to 6 seconds. The computational environment follows a 'perception-computation-action' architecture and is compatible with reinforcement learning tools, allowing fair comparison of classical regulators, neural network algorithms, and hybrid schemes. The team plans to further complicate the model and test the hybrid controller with neural networks for fully autonomous navigation of flapping-wing micro-robots.