Back

The Robobee project

I played a significant role in the development of the Robobee at Harvard University from 2010 to 2015 as my PhD thesis. The Robobee is, in essence, a robotic version of a real honeybee, in both size, form, and function. The project, to develop a swarm of small, autonomous flying robots, was started in 2009 by my PhD advisor Prof. Robert J. Wood, under a grant from the National Science Foundation (link to announcement). This project goal was inspired by the news of rapidly declining honeybee populations around the world and the ecologically disastrous consequences if they disappeared (link). A swarm of robotic bees could perform mass-dexterous manipulation of plant pollen for agriculture, removing our dependence on honeybees for our food supply and relieving bees of the industrial abuse.

This was a very exciting project, with lots of media coverage, including over 1 million Youtube hits. And lots of very interesting photo opportunities, as seen below. See an abridged list of publications for more details: (link)

I developed the mechanical design, actuation scheme, and custom manufacturing techniques for construction of the insect-scale flying vehicle. I also investigated system dynamics modeling, analysis, and experiment design for the research project. Along with flight controller development by colleague Pakpong Chirarattananon, my work resulted in the first hovering flight demonstration of the world's smallest flying vehicle.

The Robobee (published version from May 2013) has a 3.5 centimeter wing span, a flapping frequency of 120 Hz, and weighs 80 milligrams with a maximum payload capacity of 40 mg. The prototype only consists of the propulsion system and body structure (see below), and requires offboard sensing, computation, and power sent through a hair-thin wire tether, in order to achieve stable hovering flight.

Building a fully autonomous robotic system requires many different components integrated onto a single system. Because our robot is so small—with limited vehicle volume (2 toothpick diameters), limited carrying payload, and limited power—everything has to be tailored for this application, or built from scratch. There is a very limited number of off-the-shelf components on the robot. The major subsystems of the robot include:

My PhD dissertation encompassed the propulsion system and robot body structure, or vehicle chassis. A brief overview of my contributions follows.

To address the challenges of designing a robotic flying insect, we study the wing kinematics and aerodynamics of real insect flapping wings, to mimic their form and function for our robot. At the scale of a few centimeters, aerodynamics enter a laminar flow regime where the fluid viscosity becomes much more significant (Reynolds numbers on the order of 1000) and invisid airflow assumptions that govern much of macroscale, fixed-wing air vehicle design principles do not apply. Between flapping wings and revolving wings, flapping wings have a distinct advantage over revolving helicopter blades: flapping wing flight is inherently unstable and requires active sensing and control of the vehicle's force output. This dynamic instability, while challenging for flight control, allows a flapping-wing vehicle to be more maneuverable. The flapping wings act as both thrust generators and control surfaces.

A two-wing design was chosen for mechanism simplicity. Flying insects such as bees and flies flap their wings over 150 times a second, with wing stroke amplitudes over 90 degrees. The wings must go through large motions, at high velocities with high forces. To drive this motion, we need a flight muscle, or actuator, that is very power dense. From our survey of technologies, piezoelectric actuators offer the force and bandwidth required. We fabricated our own actuators and optimize them for this application. The wings—stiff and lightweight—are polyester film with carbon fiber frames. Interfacing the wing to the actuator is a transmission mechanism. This transmission mechanism needs to be efficient (no friction, no backlash), precise, and last millions of cycles. It needs to amplify the output motion of the flight muscles, and also remain very lightweight and compact. To address all of these specifications, we construct it from polymer flexure hinge mechanisms, using custom-developed manufacturing methods.

Every component must weigh as little as possible to maximize the vehicle’s payload capacity. Carbon fiber is the dominant structural material, precisely machined with a custom-built laser micromachining system capable of 20 micrometer feature sizes. A principal challenge for the propulsion system’s mechanical design lies in fitting all components into a very small body volume.

Below is an early iteration of the Robobee design (circa 2011), with a single actuator driving two wings simultaneously. Without asymmetric control of the wings, the robot was incapable of flight control.

To achieve flight control using only flapping wings, the robot must be able to flap its wings asymmetrically. I proposed an aggressively minimalistic design with two mirrored halves—two actuators, one for each wing—to enabled independent control of each wing while minimizing mass of the structure and mechanisms.

An issue that often goes unnoticed is: how is everything put together? The many separate components are assembled together by hand under microscope, like a mechanical watch. Unfortunately, this tends to result in inconsistent results, especially without years of practice. I innovated on fixture and introduced design-for-assembly concepts to greatly improve our yield rate of robots. Our custom developed “pop-up manufacturing” methodology also played an important role and has taken on a life of its own (link). This design effort was particularly critical because the two-actuator mirrored design exacerbated the need for vehicle symmetry and construction precision. The improved construction consistency of my robot design has enabled the work of dozens of other experiments on the Robobee and has becoming a versatile experimental research platform.

Beyond the mechanical design and manufacturing methods, I studied system design, modeling, and analysis. Taking into account flapping wing aerodynamics, system dynamics, and efficiency considerations, I also developed and built a larger scale version of the Robobee specifically designed to support the additional payload requirements for a fully autonomous vehicle. (link)

The project remains ongoing. Interesting extensions to the research results include controlled landing, perching, and even swimming (link). The integration of onboard power and electronics to achieve a fully wireless and autonomous system remains imminent, as of January 2016.