IN the future, tiny flying robots could be deployed to locate survivors trapped in rubble after a major earthquake. Like real insects, they would navigate cramped spaces unreachable by larger robots while avoiding static obstacles and falling debris.

Until now, most aerial microbots have been limited to slow, smooth flight paths — far from the swift, agile movement of real insects. But that may soon change.

Researchers at MIT have developed a flying microbot whose speed and agility are now comparable to those of its biological counterparts. A team co-led by Kevin Chen of the Department of Electrical Engineering and Computer Science created an AI-based controller that allows the tiny robot to perform acrobatic maneuvers, such as executing continuous flips in mid-air.

The new two-part control system — which balances high performance with computational efficiency — has increased the robot’s flight speed by about 450% and its acceleration by 250% compared to the team’s earlier best models.

In tests, the robot executed 10 consecutive somersaults in just 11 seconds, even under wind disturbances that would typically knock it off course.

“We want to deploy these robots in scenarios where traditional quadcopters struggle, but insects excel,” Chen said. “With our bioinspired control framework, our robot’s flight performance now matches insects in speed, acceleration, and pitch control. This is a significant step toward that goal.” Chen is also head of the Soft and Micro Robotics Laboratory within the Research Laboratory of Electronics (RLE) and co-senior author of a related paper recently published in Science Advances.

Chen’s group has been developing insect-inspired robots for over five years. Their latest iteration is about the size of a microcassette, weighs less than a paperclip, and features larger flapping wings for greater agility. The wings are powered by soft artificial muscles that drive them at extremely high frequencies.

However, the robot’s “brain” — the controller that guides its movement — had previously been manually tuned, limiting its performance. To achieve truly insect-like flight, the bot needed a controller robust enough to handle uncertainty and perform rapid, complex optimizations in real time. Yet such a controller would normally be too computationally demanding, especially given the complex aerodynamics of a lightweight flying machine.

To overcome this hurdle, Chen teamed up with Jonathan P. How of the Department of Aeronautics and Astronautics. Together, they designed an AI-driven, two-stage control scheme that delivers both the robustness needed for agile maneuvers and the efficiency required for real-time operation.

“Hardware improvements drove controller development, and advances in control unlocked new hardware capabilities,” How explained. “It’s a synergistic cycle: as Kevin’s team pushes the hardware forward, we find new ways to leverage those gains in software.”(SD-Agencies)