Our team has developed a novel bio-inspired butterfly drone that leverages the unique low-frequency, high-amplitude flapping wing mechanism of real butterflies. Unlike conventional micro aerial vehicles suffering from high noise, short endurance, and poor stealth, this butterfly drone achieves stable omnidirectional flight with a wingspan of only 295 mm and a total mass of 17.9 g. In this paper, we present the complete structural design, mechanical transmission system, flight controller integration, and iterative center-of-gravity optimization process. Extensive CFD simulations and experimental verifications confirm the effectiveness of our approach.
1. Introduction and Biological Inspiration
Butterflies are among the most efficient flapping-wing flyers in nature. Their flapping frequency is typically around 10 Hz, with a stroke amplitude approaching 180°, and the wing-body coupling generates unsteady aerodynamic forces and leading-edge vortices that enable long-distance migration and high maneuverability. These characteristics are highly desirable for small-scale drones used in agricultural monitoring, environmental rescue, and covert reconnaissance. However, replicating butterfly flight in a micro drone presents significant challenges: the unsteady aerodynamics are complex, the wing-body interaction is hard to model, and the tight constraints on weight and power require innovative design choices. Our butterfly drone attempts to overcome these hurdles by employing a single-motor drivetrain, a lightweight carbon-fiber skeleton, a flexible membrane wing, and an integrated PCB flight controller. The following sections detail our design philosophy, mechanical implementation, and validation results.
2. Overall Design Strategy
The primary goal was to achieve stable, controllable flapping flight with minimal noise and high stealth. We began by selecting a single-motor configuration to reduce parasitic mass and simplify power distribution. The motor is coupled to a gear train and a crank-slider mechanism that converts continuous rotation into reciprocating wing motion. The entire airframe is built from 0.5 mm carbon-fiber rods, chosen for their excellent specific strength and stiffness. Wing surfaces use P3N1 kite fabric, which offers a good balance of flexibility, tear resistance, and lightweight. The flight controller board integrates wireless communication, power management, and sensor interfaces on a compact multi-layer PCB. Center-of-gravity (CoG) tuning was performed iteratively using both numerical simulations and flight tests. Table 1 summarizes the key design parameters.
| Parameter | Value | Unit |
|---|---|---|
| Wingspan | 295 | mm |
| Total mass | 17.9 | g |
| Flapping frequency | 10 | Hz |
| Stroke amplitude | ~180 | deg |
| Motor type | Coreless DC | – |
| Gear ratio (total) | 1:16 | – |
| Wing membrane material | P3N1 kite fabric | – |
| Frame material | Carbon-fiber rod (0.5 mm dia.) | – |
| Battery capacity | 180 | mAh |
| Flight controller | Multi-layer PCB with MCU & RF module | – |
3. Mechanical Transmission System
The power train is the heart of the butterfly drone. We use a single coreless DC motor driving a two-stage gear reduction. The first gear pair (pinion on motor shaft, gear 1) reduces speed and increases torque. The second stage consists of a symmetric pair of gears (gear 3 and gear 4) that split the power to left and right wings simultaneously, ensuring synchronized flapping. Each gear has an eccentric pin that connects to a carbon-fiber crank-rod, which in turn drives the wing root. The kinematic relationship between motor rotation and wing angle can be expressed as:
\[
\theta(t) = \theta_0 + A \sin(2\pi f t)
\]
where \(\theta(t)\) is the instantaneous wing angle, \(\theta_0\) the mean angle, \(A\) the amplitude, and \(f\) the flapping frequency. The crank-slider mechanism introduces a slight asymmetry in upstroke and downstroke, which we tuned to match the butterfly’s natural wing beat. The gear reduction ratio is:
\[
i = \frac{N_2}{N_1} \times \frac{N_4}{N_3} = 16
\]
with \(N_1=8\), \(N_2=32\), \(N_3=10\), \(N_4=40\). This ratio yields the desired 10 Hz flapping at a motor speed of ~9600 rpm. Table 2 lists the gear parameters.
| Gear | Teeth | Module (mm) | Material |
|---|---|---|---|
| Motor pinion | 8 | 0.3 | Steel |
| Gear 1 | 32 | 0.3 | Acetal |
| Gear 3 (left) | 10 | 0.3 | Acetal |
| Gear 4 (right) | 10 | 0.3 | Acetal |
The symmetric gear arrangement ensures that both wings flap in antiphase (as in real butterflies), preventing yaw moments. The crank-rod length is 8 mm and eccentric offset is 3 mm, giving a wing stroke of approximately 170°.
4. Wing Skeleton and Membrane
The wing skeleton mimics the radial venation pattern of a butterfly wing. We use 0.5 mm carbon-fiber rods arranged in a symmetric fan shape. The main spar runs from the wing root to the leading edge, with cross ribs providing stiffness in the chordwise direction. This design distributes aerodynamic loads evenly while allowing the membrane to deform passively. The wing area is about 120 cm² per side. Figure below shows the skeletal layout (a photo of the actual prototype is inserted later).
The membrane is made of P3N1 kite fabric, a lightweight polyester material coated with a thin polyurethane layer. We evaluated several candidates: PE film (too brittle), standard ripstop nylon (too heavy), and P3N1 (optimal). Table 3 compares material properties.
| Material | Thickness (mm) | Areal density (g/m²) | Tensile strength (MPa) | Tear resistance |
|---|---|---|---|---|
| PE film | 0.03 | 28 | 15 | Poor |
| Ripstop nylon | 0.05 | 45 | 60 | Excellent |
| P3N1 kite fabric | 0.04 | 32 | 40 | Good |
We chose P3N1 because it provides the best trade-off between lightweight, flexibility, and durability. The membrane is attached to the carbon skeleton using cyanoacrylate adhesive and reinforced with thin Kevlar thread at stress points.
5. Flight Controller and Electronics
The butterfly drone uses a custom 4-layer PCB integrating an STM32 microcontroller, an nRF24L01+ wireless module, a 6-axis IMU (MPU6050), and a low-dropout regulator for the 3.3V rail. The PCB dimensions are 25×18 mm with a total mass of 1.2 g. Power is supplied by a 1S 180 mAh LiPo battery (5.6 g). The IMU provides real-time attitude feedback, and the wireless link supports command transmission and telemetry (battery voltage, pitch/roll angles) at up to 2 Mbps. The firmware implements a simple PD controller for roll stabilization:
\[
u = K_p e + K_d \dot{e}
\]
where \(e\) is the roll angle error and \(u\) is the differential motor speed offset (though only one motor is used, we apply asymmetric damping via wing timing modulation). The controller runs at 100 Hz. Battery management uses a fuel gauge integrated with the ADC.
6. Center-of-Gravity Optimization and Flight Testing
Initial flight tests revealed severe instability: the butterfly drone exhibited periodic yaw oscillations and could not maintain a straight flight path. The flight distance was 40% shorter than the design target. We suspected the CoG was misaligned with the aerodynamic center of the wings. Using computational fluid dynamics (CFD) simulations in ANSYS Fluent, we analyzed the pressure distribution over the wing surface at different CoG positions.
The CFD model assumed a flapping frequency of 10 Hz and amplitude 170°, with dynamic meshing to capture the wing motion. We observed that when the CoG was located 5 mm behind the wing pivot line, the pressure contour showed a concentrated negative pressure zone near the wingtip, causing an unbalanced pitching moment and energy loss. After shifting the battery forward by 3 mm and the flight controller board by 2 mm, the pressure distribution became uniform. Figure below shows the normalized pressure contours (the actual photo is inserted above).
We applied an iterative tuning procedure: for each CoG location, we computed the aerodynamic force and moment using the formula:
\[
\overline{F_z} = \frac{1}{T} \int_0^T \oint_S p \, dS_z \, dt
\]
where \(\overline{F_z}\) is the average lift force, \(p\) the pressure, and \(S_z\) the vertical projection of the wing area. The optimal CoG was found when the pitching moment coefficient \(C_m\) was below 0.01. Table 4 summarizes the iterative steps.
| Iteration | Battery X offset (mm) | PCB X offset (mm) | Average lift (N) | Pitching moment (N·mm) | Flight performance |
|---|---|---|---|---|---|
| 1 | 0 (reference) | 0 | 0.168 | 12.4 | Unstable, heavy yaw |
| 2 | +2 | 0 | 0.172 | 8.2 | Moderate oscillation |
| 3 | +3 | +1 | 0.176 | 3.1 | Slight drift |
| 4 (final) | +3 | +2 | 0.178 | 0.8 | Stable straight flight |
After the final iteration, the butterfly drone achieved stable flight over the full designed distance. The measured flapping frequency was 10.2 Hz, and the peak lift reached 0.18 N, which is 1.02 times the weight (0.176 N), confirming positive net lift. The endurance was measured at 4.5 minutes in hover and 6 minutes in forward flight.
7. CFD Simulation and Aerodynamic Analysis
We performed unsteady CFD simulations to characterize the vortex structures and pressure distribution around the butterfly drone wings. The simulation used a 3D incompressible Navier-Stokes solver with a sliding mesh. The Reynolds number based on chord length was approximately 12,000. Key results are summarized in Table 5.
| Parameter | Value | Unit |
|---|---|---|
| Average lift coefficient \(C_L\) | 1.45 | – |
| Average drag coefficient \(C_D\) | 0.48 | – |
| Lift-to-drag ratio | 3.02 | – |
| Peak negative pressure (wingtip) | -120 | Pa |
| Leading-edge vortex (LEV) strength | 0.045 | m²/s |
The pressure contours after CoG tuning showed uniform distribution across the wing surface, with no localized tip stall. The leading-edge vortex remained attached during most of the downstroke, enhancing lift. The wake structure revealed a classical reverse von Kármán vortex street, indicating efficient thrust generation.
8. Conclusion and Future Work
We have successfully designed, built, and tested a bio-inspired butterfly drone with a wingspan of 295 mm and total weight of 17.9 g. The single-motor drivetrain, carbon-fiber skeleton, P3N1 membrane, and compact flight controller enabled stable flapping flight after systematic CoG tuning. The iterative optimization process, guided by CFD simulations, eliminated attitude instabilities and extended flight distance to the full design target. Our butterfly drone demonstrates the feasibility of high-performance flapping-wing micro air vehicles for applications requiring low noise, high stealth, and maneuverability.
Future work will focus on increasing endurance through higher-density batteries and more efficient motor drivers, adding obstacle avoidance using lightweight ultrasonic sensors, and exploring active wing morphing for improved agility. The aerodynamic model will be refined with fluid-structure interaction to capture membrane deformation effects. With these improvements, the butterfly drone can be deployed in real-world missions such as agricultural pest monitoring, search-and-rescue in confined spaces, and covert surveillance.
The butterfly drone presented here provides a solid engineering foundation for next-generation bio-inspired flight systems, unlocking the potential of nature’s most elegant flyer for practical aerial robotics.