Harvard Robobee

A fantastic Friday morning story this one. Lots of very cool stuff coming out of the Harvard Robobee project, the seemingly impossible seems to be happening. Aviation at a size that could not have been dreamt about 100 years ago.

Printed using methods inspired by pop up books and with vision based stabilization from a tiny camera. Will we really release robots to the wild to pollinate our food supply?

The collaborators envision that the Nature-inspired research could lead to a greater understanding of how to artificially mimic the collective behavior and “intelligence” of a bee colony; foster novel methods for designing and building an electronic surrogate nervous system able to deftly sense and adapt to changing environments; and advance work on the construction of small-scale flying mechanical devices

Am I the only person though to be worried though about the prospect of robotic bees pollinating plants? What was wrong with plain old bees?

I can imagine though, such a lightweight device being lifted by thermals or storm clouds and bouncing around in the sky. Gathering data whilst having a free ride from mother nature.

The ability for such a small device to perch and stare will not be lost on the military. The swarm and hive intelligence that they hope to give the devices will be of interest as well.


The basics of Robobee in their own words

INSPIRED by the biology of a bee and the insect’s hive behavior …

we aim to push advances in miniature robotics and the design of compact high-energy power sources; spur innovations in ultra-low-power computing and electronic “smart” sensors; and refine coordination algorithms to manage multiple, independent machines.

Practical Applications

Coordinated agile robotic insects can be used for a variety of purposes including:

  • autonomously pollinating a field of crops;
  • search and rescue (e.g., in the aftermath of a natural disaster);
  • hazardous environment exploration;
  • military surveillance;
  • high resolution weather and climate mapping; and
  • traffic monitoring.

These are the ubiquitous applications typically invoked in the development of autonomous robots. However, in mimicking the physical and behavioral robustness of insect groups by coordinating large numbers of small, agile robots, we will be able to accomplish such tasks faster, more reliably, and more efficiently.

Centeye are adding vision to the project as Geoffry Barrows explains over at DIYdrones

As part of Centeye’s participation in the Harvard University Robobee project, we are trying to see just how small we can make a vision system that can control a small flying vehicle. For the Robobee project our weight budget will be on the order of 25 milligrams. The vision system for our previous helicopter hovering system weighed about 3 to 5 grams (two orders of magnitude more!) so we have a ways to go!


We recently showed that we can control the yaw and height (heave) of a helicopter using just a single sensor. This is an improvement over the eight-sensor version used previously. The above video gives an overview of the helicopter (a hacked eFlite Blade mCX2) and the vision system, along with two sample flights in my living room. Basically a human pilot (Travis Young in this video) is able to fly the helicopter around with standard control sticks (left stick = yaw and heave, right stick = swash plate servos) and, upon letting go of the sticks, the helicopter with the vision system holds yaw and heave. Note that there was no sensing in this helicopter other than vision- there was no IMU or gyro, and all sensing/image processing was performed on board the helicopter. (The laptop is for setup and diagnostics only.)

The picture below shows the vision sensor itself- the image sensor and the optics weigh about 0.2g total. Image processing was

Centeye Sensor

performed on another board with an Atmel AVR32 processor- that was overkill and an 8-bit device could have been used.

A bit more about optics: In 2009 we developed a technique for “printing” optics on a thin plastic sheet, using the same photoplot process used to make masks for, say, making printed circuit boards. We can print up thousands of optics on a standard letter size sheet of plastic for about $50. The simplest version is a simple pinhole, which can be cut out of the plastic and glued directly onto an image sensor chip- pretty much any clear adhesive should work.The picture below shows a close-up of a piece of printed optics next to an image sensor (the one below is a different sensor, the 125 milligram TinyTam we demonstrated last year).