Robotic varmints


There’s the wiggly robot developed by a team from Ecole Polytechnique Fédérale de Lausanne.

It’s based on the salamander’s spinal column and, directed by a lab top, it swims, crawls, changes speed and directions­, giving scientists information about locomotion.

The Waalbot, developed by a team from the NanoRobotics Lab, Carnegie Mellon University, is like a synthetic gecko with dry adhesives on the bottom of its “feet.”  This robot walks up walls.


And then there’s RoboFly, a project that Axel Borst, Ph.D., director of Max Planck Institute of Neurobiology, introduced to his audience last month during his talk, “From the Cockpit of a Fly: How Visual Information is Processed.”

RoboFly, as well as the other robots, will eventually go where humans won’t (or can’t). But, really, that’s just a byproduct, because their real contribution is to help scientists understand biological principles.


Here’s a little background on Max Planck Institute, which received almost $190 million in state and local money in return for a promise to hire 135 employees. It has operated at its temporary Jupiter campus at Florida Atlantic University since 2008, while waiting for its 100,000 square-foot facility to be finished sometime in 2012, and, currently has a staff of 45.

A neuroscience company headquarted in Munich, its 80 institutes focus on a variety of fields. Borst, from Munich, specializes in neurobiology. Since he’s interested in how motion information from changing retinal images is computed and guides the fly in flight, his research is centered on two types of flies: the big fly and the fruit fly.  In addition, Borst serves as an advisor here in Jupiter (see the sidebar on the one-whiskered mouse).

Dr. Borst

How the fly sees and moves helps us understand how the brain works and how visual motion is processed in the brain, Borst said.

“As neuroscientists, we’d like to understand how the brain works. That would answer philosophical questions: How do we perceive the world? What makes our personality? How should we teach children? How can we help cure diseases of the brain?”

We do know that the brain’s volume is 1.3 litres and it contains a trillion neurons, he said, and we can assign function to different areas of the brain and we know that behind the visual cortex is where vision is processed.

Through MRIs, scientists have learned a lot in the last 20 years, he added. “A scanner subject is given a task and his or her activity is recorded. But an MRI’s resolution is one cubic millimeter, so we can’t assign the function of different parts of the brain.

“The connection between single neurons and what we see on the MRI is what we are missing. In between is the circuitry, and we don’t understand that.

“In computers, for example, we understand the chips but what we don’t have is the circuit diagram that makes up the function of the chip.”

And since, Max Planck scientists want to understand the brain at that level, they studied the neurons and their connections so that they could figure out the “circuit diagram.”

While the brain volume is 1.3 litre, the fruit fly brain is much smaller – .0000001 litre. Human brains have trillions of neurons while a fruit fly has 100,000, but the volume per neuron is the same, he said.

“The fly brain is just as complex, but has fewer neurons.  This is why I picked the fly.

“Each of the facets of its eye can be regarded as a single pixel. Each facet looks at a neighboring point and they form an image like we do with a single lens.”

When the fly is buzzing around, it’s hard to catch. That’s because it’s relying on optic flow information, Borst explained.

“Moving forward to the tree, imagine that the world is expanding away from where you are heading. The images are shifting as you move forward.  You can describe this image by a vector field where you indicate the motion of each pixel.  Image pixels that you are heading to seem to be moving slowly, and at your sides, moving fast, like you are going down a lane. That’s optic flow, and it depends on how you move, and when you rotate, it moves in the opposite direction.

“Flies use that to navigate through the environment. We do that as well. If I tell you to walk straight with your eyes closed, you couldn’t.  But if you look at an image of a rollercoaster, you just see it and you can experience it, without being on it.

“The brain area where neurons respond to optic flow, it’s like those 3D movies. In flies, we call it the cockpit of the eye.”

Researchers study the optic flow by filling neurons with dyes and using electrodes to make intracellular recordings, looking at connectivity and the properties of the lobula plate cells and the descending neurons.

“Just like the human, neighboring points always are processed by neighboring points, and, in the fly, the lobula plate creates the map,” he said.

Local motion vectors are calculated from local changes in retinal brightness. From the resulting optic flow, course control parameters are extracted.

“Following this motion is computed by a mechanism that does sequence detection.  You get directional information: vertical, horizontal and rotation,” he said.

“We copied the circuit, and made a computer simulation.”

Now back to his team’s creation, RoboFly, the flying robot whose flight trajectory is controlled by visual stimuli.

So, asked Borst, “Why not build an artificial eye? And we started the RoboFly project.

“We decided to buy a commercially available platform (it looks like a gyrocopter) and we mounted a camera with two wide-angle lenses, and then we followed that by parallel chips that would implement the locomotion detecting followed by the lobula plate network. We still fly it with remote control, but we plan to close the loop and unleash it to see what it does.”

RoboFly is an opportunity to test the function of the circuit and to see if it works the way he think it works as well as to give insight as to why the neurons are wired up the way they are, he said. And the technical application for autonomous robots is to send them where people don’t want to go. “So if you have the robot, an autonomous regulation mechanism that avoids collisions and stabilizes even with a gust of wind, it can send me the images and fly back.”

Written for palm2jupiter