1.29-Gram Flying Robot Navigated With Its Own Sensors but Still Needed Wired Power
A 1.29-gram flapping-wing robot used onboard sensing and computation to avoid obstacles and land on a sunflower without external motion tracking. The 30-second test achieved centimetre-scale positioning, but a human supplied high-level commands and the robot still relied on wired power.
A 1.29-gram flapping-wing robot has flown for 30 seconds using sensors and computation carried on its own body, avoiding obstacles before landing on a sunflower. The Science Robotics result removes the external motion-capture cameras that normally tell insect-scale aerial robots where they are, but the machine still drew electrical power through a wire.
The 30-second summary
- What happened? MIT and Cornell researchers integrated a microcontroller, inertial sensor, optical-flow sensor and height sensor into a 244-milligram electronics package carried by a four-winged microrobot.
- Why does it matter? The robot could estimate its own motion with centimetre-scale accuracy outside a motion-capture arena, a prerequisite for field uses where external cameras are unavailable.
- What is the catch? A human supplied high-level commands and an offboard source supplied power, so this was not fully autonomous or untethered outdoor flight.
Key number: The complete robot weighed 1.29 grams, while its onboard sensing and computing package weighed 244 milligrams.
A built-in sense of motion
Laboratory microrobots often depend on cameras mounted around a room. Those cameras track reflective markers, a computer calculates the vehicle's position, and a radio or cable sends control information back. The arrangement can produce precise flight, but it cannot follow a tiny robot through crops, damaged buildings or industrial equipment.
The new vehicle carried three kinds of measurements. An inertial measurement unit estimated orientation, an optical-flow sensor tracked lateral motion, and a time-of-flight sensor measured height. A microcontroller combined them through simplified estimators and ran the low-level flight controller. The team found that updating position quickly with a less elaborate calculation worked better than waiting for slower, more precise estimates.
The system maintained position within a few centimetres while hovering and following trajectories. In the headline demonstration, an operator issued high-level position targets with a joystick. The onboard system handled stabilization and movement, guided the robot around plants and landed it on a sunflower without external motion tracking.
How the robot carried the electronics
Adding sensors creates a basic problem at insect scale: every milligram reduces the lift available for flight. The researchers increased the radius and length of the soft dielectric-elastomer actuators by 50%, raising net lift by 46% while preserving the flapping performance. The resulting lift-to-weight ratio was 1.9, enough to carry the electronics package.
The electronics consumed about 6% of the measured power. Most energy still drove the artificial muscles and wings. That split suggests navigation hardware is not the main energy burden, but a battery and its power electronics would add substantial mass that the present platform did not carry.
What the experiment establishes
The peer-reviewed paper establishes sensory and computational autonomy at insect scale. It does not establish mission autonomy: the robot did not choose its own route or identify a flower, and a human operator supplied the trajectory targets. The 30-second test also does not demonstrate endurance, operation in wind or reliable navigation through unstructured vegetation.
Before we overstate the result
The robot remained physically tethered to an offboard power source. Its sunflower landing was a controlled demonstration, not autonomous pollination. The team must integrate a battery, power electronics, obstacle avoidance and object recognition without losing lift or stability. Outdoor air movement and repeated collisions will add challenges absent from the reported test.
The next engineering threshold
The researchers are working on an onboard camera for object recognition and obstacle avoidance, alongside a battery and power electronics for untethered flight. Success would move the platform from self-stabilizing flight toward independent field operation.
The result complements NewTqnia's coverage of a 0.45-gram soft robot with four movement modes, which also remained wired, and a neuromorphic feedback controller that demonstrated local closed-loop control. Here, the advance is a different layer: carrying enough perception and computation to fly without a room-scale tracking system.
Verified topics and entities
Sources and citations3 sources
External references used to support the reporting in this article.
Published by
NewTqnia Robotics Desk
An institutional editorial team within NewTqnia