⚡ electrical engineering

Controlled Flight of an Insect-Scale Flapping-Wing Robot via Integrated Onboard Sensing and Computation

Researchers have developed a 1.29-gram flapping-wing robot capable of autonomous, centimeter-scale trajectory tracking and obstacle avoidance using only onboard sensing and computation, marking a significant step toward deploying insect-scale robots in real-world environments.

Original authors: Yi-Hsuan Hsiao, Quang Phuc Kieu, Zhongtao Guan, Suhan Kim, Jiaze Cai, Owen Matteson, Jonathan P. How, Elizabeth Farrell Helbling, YuFeng Chen

Published 2026-02-10

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yi-Hsuan Hsiao, Quang Phuc Kieu, Zhongtao Guan, Suhan Kim, Jiaze Cai, Owen Matteson, Jonathan P. How, Elizabeth Farrell Helbling, YuFeng Chen

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Tiny Pilot: How a Robot the Weight of a Paperclip Learned to Fly Solo

Imagine you are trying to fly a tiny drone through a dense forest. Usually, to keep that drone from crashing, you’d need a massive computer nearby, a team of engineers, and a high-tech "eye in the sky" (like a motion-capture camera system) watching every single move. Without that "big brother" computer telling it where it is, the drone would just tumble blindly.

For a long time, that has been the problem with insect-scale robots—robots so small they weigh about as much as a single paperclip. They are too tiny to carry a heavy computer, so they usually have to be "tethered" to a giant brain sitting on a desk.

This paper describes a breakthrough: scientists have built a 1.29-gram robot that carries its own "brain" and "senses" inside its tiny body, allowing it to fly, dodge obstacles, and even land on a flower all by itself.

The Three Big Challenges (The "Tiny Pilot" Problem)

To understand why this is hard, think about the three things a pilot needs:

The Senses (The Eyes and Inner Ear): You need to know if you are tilting, how high you are, and if you are moving left or right.
The Brain (The Processor): You need to take all that sensory info and decide, "Hey, I'm about to hit that leaf! Move left!"
The Muscle (The Wings): You need to flap your wings fast enough to stay up, but not so hard that you lose control.

In a tiny robot, these three things fight each other. If you add a bigger "brain," the robot becomes too heavy to fly. If you add better "eyes," the robot uses up all its battery. It’s like trying to pack a smartphone, a flashlight, and a GPS into a single grain of rice.

How They Did It: The "Smart Suitcase" Approach

Instead of just throwing parts together, the researchers used a "holistic design." This means they designed the robot and the electronics at the exact same time, making sure they fit together like a perfectly tailored suit.

The Sensory Suite: They packed in a tiny motion sensor (to feel tilt), a distance sensor (to see how far the ground is), and an "optical flow" sensor (which works like a tiny camera to see movement across the ground).
The Lightweight Brain: They used a microscopic computer chip (an MCU) that is incredibly efficient. It’s so smart that it can do all the math required to stay stable while using almost no power.
The Super-Muscles: They used "dielectric elastomer actuators"—essentially artificial muscles that flap at 330 times per second.

The "Grand Finale": Landing on a Sunflower

To prove this wasn't just a lab trick, the team took the robot out of the controlled lab environment and into the real world.

Imagine a tiny, winged superhero navigating a jungle of plants. Using only its own tiny onboard brain, the robot:

Took off from the ground.
Dodged obstacles (like leaves and stems) without being told to by a human.
Found a target: A sunflower.
Nailed the landing: It descended and perched itself right on the flower.

Why Does This Matter? (The "So What?")

This isn't just about making cool toys. This technology opens doors to a future where tiny, autonomous robots can go where humans and even large drones cannot.

Search and Rescue: Imagine a swarm of these tiny robots flying into the rubble of an earthquake to find survivors in tiny crevices.
Precision Agriculture: Imagine "robotic bees" that can fly from flower to flower, pollinating crops with extreme precision to help grow more food.
Environmental Monitoring: Tiny scouts that can fly through dense jungles to monitor plant health or detect forest fires.

In short: We have moved from "remote-controlled" tiny robots to "thinking" tiny robots. The tiny pilots have finally arrived.

Technical Summary: Controlled Flight of an Insect-Scale Flapping-Wing Robot via Integrated Onboard Sensing and Computation

Problem Statement

A fundamental gap exists in the field of insect-scale micro-aerial vehicles (IMAVs). While biological insects navigate complex environments using highly efficient, onboard sensory-motor loops, most existing IMAVs rely on offboard sensing (e.g., motion capture systems) and computation to maintain stability. This reliance restricts IMAVs to controlled laboratory settings, preventing their deployment in real-world applications such as search-and-rescue, precision agriculture, or pollination. The primary challenge is the "trilemma" of extreme payload limitations, high power consumption, and rapid, oscillatory body dynamics that make onboard state estimation and control exceptionally difficult.

Methodology

The researchers employed a task-driven, holistic design approach, co-designing the robot's physical structure and its avionics to meet specific flight requirements (e.g., centimeter-scale accuracy for landing on flowers).

Hardware Design:
- Actuation: The robot uses four independent flapping-wing modules powered by Dielectric Elastomer Actuators (DEAs). The researchers scaled up the DEA dimensions to increase the lift-to-weight ratio to 1.9.
- Avionics Suite: A 244-mg flight package was integrated onto three custom flexible PCBs. It includes an IMU (LSM6DSV80X), a Time-of-Flight (ToF) sensor (TMF8806), an optical flow camera (PAA3905E1-Q), and a dual-core Microcontroller Unit (MCU) (CY8C6245FNI-S3D41T).
- Structural Mitigation: A 0.4-mm fiberglass substrate was used for the IMU to mitigate mechanical vibrations and thermal noise from the flapping wings.
Estimation & Control Architecture:
- Cascaded Estimator: To minimize computational overhead, the team avoided complex Extended Kalman Filters (EKF) that require heavy Jacobian calculations. Instead, they implemented a three-stage pipeline:
  - Attitude/Angular Velocity: A Mahony nonlinear complementary filter.
  - Altitude: A Kalman filter integrating ToF data, attitude, and thrust commands.
  - Lateral Position/Velocity: A Kalman filter integrating optical flow, attitude, and altitude.
- Hierarchical Control: A low-level high-bandwidth controller (480 Hz) handles stability, while a high-level controller (operated via joystick in experiments) manages trajectory commands.

Key Contributions

First Fully Autonomous IMAV: The first insect-scale robot to demonstrate flight relying exclusively on onboard sensors and an MCU, rather than external motion capture.
Integrated Co-Design Methodology: Demonstrating that scaling actuators (DEAs) and avionics jointly is more effective than attempting to retrofit sensors onto existing platforms.
Computationally Efficient Estimation: A cascaded estimation framework that achieves high accuracy while consuming minimal MCU resources (utilizing only 25% of a single core).

Results

State Estimation Accuracy: The onboard estimator achieved errors of approximately $3.05^\circ$ in attitude and $2.5\text{ cm}$ in position.
Flight Performance:
- Hovering: Achieved centimeter-scale accuracy (RMS lateral error of $3.24\text{ cm}$ at $5\text{ cm}$ altitude).
- Trajectory Tracking: Successfully executed "infinity" and circular patterns with lateral RMS errors between $2.19\text{ cm}$ and $6.09\text{ cm}$ .
- Complex Tasks: In a 30-second flight outside a motion capture arena, the robot successfully navigated around obstacles and performed a precision landing on a sunflower.
Power Efficiency: The entire sensing and computation suite consumed only $120\text{ mW}$ , which is just 6% of the power used by the flight actuators.

Significance

This work represents a major milestone in microrobotics, moving the field from "laboratory curiosities" that require external infrastructure toward truly autonomous agents. By proving that centimeter-scale precision is possible with a sub-gram payload and minimal power, the researchers have opened the door to:

Onboard Planning: Using the remaining 75% of MCU capacity for autonomous obstacle avoidance and path planning.
Power Autonomy: Integrating small batteries or solar cells for long-duration missions.
Real-World Deployment: Enabling micro-robots to perform delicate biological tasks, such as assisted pollination, in unstructured natural environments.