MILITARY TECHNOLOGY

More Than Meets the Eyes

Advanced Robots Have More to Learn From Insects Than From People

Picture a housefly in your kitchen. You want to hit him with a fly swatter. Not so easy a job. It has been buzzing around and finally lands on the edge of the kitchen stove. As you quietly walk over to get the swatter, it sees you and takes off again. Now swatter in hand, you wait. He buzzes around the kitchen and eventually lights on the kitchen sink. But you move too quickly and it takes off again. Drat! You wait. Luckily he comes back to the sink. You take aim, let fly with the swatter…and miss. Double drat! You wait some more. Luckily he comes back to the sink once more. You take aim, let fly with the swatter, and now have a squashed bug! You’re so proud of yourself. But think about it. You have a huge, complex marvel of a brain the size of a small cantaloupe and the former fly has, well almost no brains at all. Why was it so hard? What gives such little insects those kinds of abilities?

Insects pack an amazing set of capabilities into a very small package. Understanding and mimicking those abilities—like the fly’s sensitive vision—can allow researchers to shrink the size of autonomous robots to proportions like those of household pests. The trick may lie in skipping all the complex equipment necessary to imitate real vision, and building sensors that handle the bare minimum of navigational data, as one example. To understand the process of building devices like this, it worth considering what insects are capable of and how they surpass even our most well-developed existing technologies.

Forget the self-preservation instincts of that fly. It, like you, has to find dinner. It has to find a mate and a suitable housing tract for the little loved ones. A bee flies out of a hole in a tree in a random pattern looking for food for the hive. A kilometer away, in an unplanned direction, it scores. It has found your neighbor’s blooming hydrangea bushes. Now it flies directly back to the hole in the tree and magically dances for its hive-mates to tell them where they can find dinner.

That’s all very impressive, but humans are pretty bright, too. Some of us are currently making robot-like, semi-autonomous systems. These moving devices can make decisions and perform complex tasks, like negotiate and avoid a set of obstacles on complex terrain. A perfect example is the development of Martian Rover-type vehicles—though with these devices, humans have final say over the machine’s decisions.

So how do these things work? First, there’s the basic vehicle or “platform” as we call it in Department of Defense lingo. If the platform is the size of an automobile, then engineers have lots of room to work with and house the necessary components, which include a motor, storage for fuel or some other energy source, and a mechanical apparatus for propulsion. Turn here, fly there, or whatever. So far, so good.

All this we have built for years, and very well I might add: cars, planes, and boats. But what about equipment for gathering information about surroundings? This starts off simple too. In the dim past, we put cameras on the vehicle and remote operators used the sensory input to navigate. More high-tech solutions put a small radar or ultrasound device on the vehicle to determine distances and display the results to a human operator. Again, so far, so good.

It is impractical to just keep shrinking the systems we have used successfully for years.

But we want a real robot. That’s what we’re shooting for. So we hook up all the sensor information to a big computer and write a program for the computer to extract the information it needs to navigate from the images provided by the sensors. Systems like these are a marvel of modern computational capability.

Now comes the real question. What do we do when the vehicle, or platform, needs to get smaller? And not just a little smaller, a lot smaller—insect size. Perhaps half an inch long or smaller. How do we make operational platforms this small? Do you just keep shrinking things down?

We could perform rather expensive experiments to try and answer those questions. Or, if the experiment has already been done—and it has—we could look to the available data for the answers. Actually, nature did the experiment hundreds of millions of years ago, and I will paraphrase the results this way: If we have enough room on board a creature, or platform, to perform complex processing, then we have the luxury of using “regular,” eyes to provide an image of the outside world. The on-board processor, the brain, has the computational power to determine the key survival/decision making issues from those images.

But what happens when the creature/platform is too small to carry a complex processor on board? We could just skip the image part all together and get the information needed directly from the outside world. This is what some, but not all, insects do.

Let’s try a simple thought experiment. Take your old boss, the one you did not like, along with a honeybee, and imagine them in a large featureless plain in say, Nebraska, with an overcast sky. Explain to each of them that if they travel east of north by 37 degrees, they will find adequate food and water. The bee is in good shape, but your mean old boss is in trouble because he has no good images to work with and is lost. The bee, on the other hand, notices that the ultraviolet light from the sun is polarized by the atmosphere in a manner consistent with its position in the sky, immediately orients its flight path, even with cloud cover, to 37 degrees east of north, and takes off into the wild blue yonder. The crucial point here is that the bee brain is capable of processing data for certain tasks without the need for visual images.

It is impractical to just keep shrinking the systems we have used successfully for years. Building a one-inch platform that can fly, navigate, and store energy is difficult. Adding sensors, processing, and communications capabilities so that is can send all the information it acquires back to home base is just too hard. Forget it. The insect model is the only practical one. What we need is to better define that model and its capabilities. Then we can better define the possible.

Recent advances in optical design allow for crude models of the insect hardware: real compound eyes coupled to complex sensors. These are “instrumentation sensors,” rather than imaging sensors, though they are capable of crude imaging when necessary. They can be made small and can be replicated cheaply. But to fully exploit this technology, we need a better understanding of the myriad of tasks performed by land, sea, and airborne creatures that are useful to our needs, before we can reproduce them in hardware.

Shown schematically is the fiber bundle interior of a compound eye manufactured by a process called “laser ablation,” which is similar to the method used to correct peoples’ vision. On the right is a completed eye with three zones of short and long focal length lenses. The light captured by each lens is directed by its fiber to the sensor array plane.^[1]

We spend much of our research resources on learning how higher-order animals think and how they process information. Human brains are a marvel of complexity. But is there a primordial component within our own processing that predates imaging? A part tied to survival of the individual that is faster and tied directly to data? When we study the insect thought process, we might learn a little more about our own.

Remember that fly? It was not so easy to swat—but you were proud of the fact that you could. Perhaps we have a bit more to learn than meets the eye.

Dr. Jerry Franck is a Senior Scientist at the Night Vision & Electronic Sensors Directorate (NVESD), Ft. Belvoir, VA. He has authored a patent on behalf of the Army on this topic titled, “Method of Manufacture for a Compound Eye.” He has joint responsibilities as head of the International Programs and as a research physicist working in the area of optical design including such topics as “High Power Coherent Laser Combination,” and “High Power Compact Short Pulse Laser.” He received his Ph.D. at Optical Sciences from University of Arizona’s Optical Science Center, Tucson, AZ and his Masters Degree from the University of North Texas. He is interested in paleontology and is an amateur fossil collector.

Notes

[1] U.S. Army Patent Application, “Method of Manufacture for a Compound Eye,” J. B. Franck, Filled 21 April 2005, Ref #: NVL 3304.

Tags: nanotechnology

Comments on this article

You can skip to the end and leave a response. Pinging is currently not allowed.