A flying microrobot that achieves the speed and agility of insects

Searching for survivors under collapsed buildings after earthquakes poses a major challenge, as traditional rescue robots are too large to fit through narrow gaps. Tiny flying devices modeled after insects could easily navigate through the rubble, but until now such microrobots have been too slow and clumsy for serious use.

A team from MIT, led by professors Kevin Chen and Jonathan P. How, has successfully overcome this problem. Using a new artificial intelligence-based control system, they have created a miniature robot that can fly like real insects. The device, powered by soft artificial muscles, allows its wings to flap extremely quickly.

The main reason for the remarkable leap in performance lies in the two-stage software. The scientists used a sophisticated mathematical model for flight planning that predicts the movement and calculates complex maneuvers such as sharp turns and successive rolls. Since such calculations would use too much computing power during the flight itself, this system acted as a “teacher”. Through imitation learning, they created a lighter and faster artificial intelligence model that is executed directly during the flight in real time.

The test results are impressive. The new algorithm gave the robot a 447 percent faster speed and 255 percent greater acceleration compared to previous versions. The tiny robot successfully performed 10 consecutive flips in the air in just 11 seconds, staying within 4 to 5 cm of its intended path, despite occasional gusts of wind. It can also perform a rapid, powerful pitch-and-stop maneuver that insects use to stabilize their vision.

The system currently relies on external cameras to monitor its position in space. Before these robots can independently save lives in the field, researchers will need to successfully mount miniature sensors and cameras on them and ensure stable operation without external support.