Flying insects and robots

TED just released an excellent talk by Michael Dickinson from the University of Washington about how flies fly.

For years now, roboticists and biologists have been working together to understand how natural and man-made systems fly. We actually organized a conference on that topic in 2007 called Flying Insects and Robots, where Dickinson was one of the keynote speakers. The questions that most often arise are “how do you build a machine that flies” and “how do you control the behavior of these flyers”?

To tackle the first question, Dickinson’s lab built the Dynamically-Scaled Flapping Robot, or Robofly. The video below shows its assembly, but you can see it in action in his TED talk. The robotic model had a 60 cm wingspan, could flap around 5 times a second, and was immersed in 2 tons of mineral oil. By measuring instantaneous forces and flow patterns, his lab was able to explore aerodynamics of flapping flight.

To answer the second question, his lab uses high-tech fly arenas (fly-o-rama, fly-o-vision and rock-n-roll arena) that control the visual surrounding of the fly and record behavioral and neuronal activity. Insight gathered from such experiments, for example on how flies use optic flow to navigate an environment, have inspired a large number of robotic systems. Optic flow can be understood as the speed at which an image moves on the surface of your eye. If you’re in a car, looking at a distant mountain, the image of the mountain will move very little on your retina. If you are about to slam into a wall, the image of the wall will expand very rapidly. Large optic flow can tell you that an object is close. Optic flow is interesting for robots because it only requires very simple sensors (see Centeye sensor below) and processing.

The lab where I worked before at EPFL had lots of examples of how simple flying robots could use optic flow for 3D obstacle avoidance. The video below shows a flying wing that uses optic flow sensors found in your typical computer mouse to avoid the ground and obstacles.

Finally, going back to biology, one of Dickinson’s recent papers explores the reaction of real flies to moving objects, in this case a fly-sized robot (flyatar). Results showed that, similar to interactions between pairs of flies, walking female flies freeze in response to objects that move from back-to-front, wherease they ignore objects that move from front-to-back.

Here is a small video of the setup showing a flyatar tracking a fly, or being controlled by a joystick:

You can check Michael Dickinson’s lab website for more fly related material.

Sabine Hauert is President of Robohub and Associate Professor at the Bristol Robotics Laboratory

Sabine Hauert is President of Robohub and Associate Professor at the Bristol Robotics Laboratory