This post is also available in:
עברית (Hebrew)
Autonomous flight at insect scale has long been a challenge in robotics. Existing microrobots can hover or move along controlled paths, but they typically lack the speed and maneuverability needed for real-world missions. This limitation is especially significant in scenarios such as collapsed buildings, narrow tunnels, or cluttered indoor environments—places where traditional drones cannot operate safely and where small, agile robots could support search-and-rescue or reconnaissance teams.
Researchers at MIT have now demonstrated an aerial microrobot capable of flying with speed and agility comparable to real insects. Using a new AI-enabled control scheme, the robot can execute rapid body flips, aggressive turns, and complex trajectories previously impossible at this scale. Its performance increases reflect the shift: speed improved by 447% , acceleration by 255%, and the robot completed 10 consecutive somersaults in just 11 seconds, even under wind disturbances.
The key problem the researchers tackled was control. Lightweight microrobots powered by fast-flapping artificial muscles are extremely sensitive to small disturbances, and earlier robots relied on hand-tuned controllers that struggled with anything beyond smooth flight paths. To overcome this, the team developed a two-part control architecture combining model-predictive control—which can plan difficult maneuvers—and a deep-learning policy trained through imitation learning. The resulting AI controller runs efficiently in real time while retaining the robustness of the more complex planner.
According to TechXplore, this approach allows the robot to perform behaviors that mirror the biomechanics of insects, such as saccades—rapid pitch-and-stop movements used for navigation and stabilization. These capabilities become especially valuable once onboard cameras or sensors are added, enabling microrobots to navigate autonomously outside motion-capture laboratories.
For defense and homeland security, insect-scale robots with high agility could support missions that require entering tight, hazardous spaces where larger drones cannot go. They could help map rubble after explosions, inspect structural voids, or silently gather information in confined interiors—tasks typically dangerous for human personnel.
Future development will focus on fully onboard computation and sensing, as well as multi-robot coordination and collision avoidance. With this advancement, microrobotics moves closer to field-ready platforms that combine extreme maneuverability with real operational value.
The research was published here.
