MIT’s Aerial Microrobot Matches Insect Speed with AI-Controlled Flight
MIT engineers developed an AI-controlled aerial microrobot capable of flying with speed and agility comparable to insects, marking a major advance in micro-scale robotics.
Engineers at MIT have demonstrated an aerial microrobot capable of flying with speed and agility approaching that of real insects, overcoming a long-standing limitation in micro-scale robotics. The insect-sized robot can execute aggressive maneuvers, including repeated midair somersaults, while maintaining stability even in windy conditions. The breakthrough was enabled by a new artificial intelligence-based control system that dramatically improves flight performance.
Tiny flying robots have long been viewed as promising tools for applications such as search-and-rescue operations, where they could navigate through narrow gaps and unstable environments inaccessible to larger drones. Until now, however, aerial microrobots have been constrained to slow, smooth flight paths due to limitations in control systems and onboard computation.
AI-Based Control Unlocks Agile Flight
The MIT team designed a two-part, AI-driven control framework that balances high performance with real-time efficiency. Compared to previous versions of the robot, the new system increased flight speed by approximately 450 percent and acceleration by about 250 percent. In testing, the robot completed 10 consecutive somersaults in just 11 seconds while staying within a few centimeters of its intended trajectory.
“We want to be able to use these robots in scenarios that more traditional quadcopter robots would have trouble flying into, but that insects could navigate,” said Kevin Chen, an associate professor of electrical engineering and computer science at MIT and co-senior author of the study. “With our bioinspired control framework, the flight performance of our robot is comparable to insects in terms of speed, acceleration, and pitching angle.”
The microrobot itself is roughly the size of a microcassette and weighs less than a paperclip. It uses flexible artificial muscles to rapidly flap oversized wings, generating lift and enabling sharp directional changes. While recent hardware improvements made more aggressive flight possible, earlier versions of the robot relied on manually tuned controllers that limited overall performance.
From Predictive Planning to Real-Time Action
To overcome these constraints, Chen’s team collaborated with researchers led by Jonathan P. How in MIT’s Department of Aeronautics and Astronautics. Together, they developed a two-step control strategy that combines predictive planning with machine learning.
The first component is a model-predictive controller that uses a mathematical model of the robot’s dynamics to plan optimal flight maneuvers. This controller can account for uncertainties in aerodynamics, external disturbances such as wind, and physical limits on force and torque. While powerful, it is too computationally demanding to run continuously in real time.
To address this, the researchers used the predictive controller to train a deep-learning-based policy through imitation learning. The resulting AI model captures the behavior of the high-performance planner in a compact form that can run extremely fast, allowing the robot to respond instantly to changing conditions.
“If small errors creep in and you try to repeat a flip multiple times, the robot will crash,” How said. “We need robust flight control.”
Toward Real-World Deployment
The robot’s agility allowed it to demonstrate flight behaviors commonly seen in insects, including rapid pitch-and-stop maneuvers known as saccades. These movements help insects stabilize their vision and orient themselves in space, and similar capabilities could support future onboard sensing.
“This bio-mimicking flight behavior could help us when we start putting cameras and sensors on board the robot,” Chen said. “This work signals a paradigm shift. It shows that we can build control architectures that are both high-performing and computationally efficient, even at insect scale.”
The research team plans to focus next on adding onboard sensors and autonomy so the robots can operate outside laboratory motion-capture environments. Coordinated flight among multiple microrobots and collision avoidance are also areas of active investigation.
The study was published in the journal Science Advances and highlights how advanced AI control methods are enabling microrobots to move beyond experimental demonstrations toward practical, real-world use.
