A Word With the Inventor of the Battlefield Snakebot and the Wall-Scaling Snailbot

Israeli roboticist Amir Shapiro looks to the animal kingdom to design robots that can go where humans can't

The Israel Defense Forces are preparing to deploy a camouflage-wearing, camera-toting robot snake. The spybot, which slithers through cracks and caves using principles of motion derived from those of actual snakes, is just one of roboticist Amir Shapiro's clever designs based on animal physiology. We visited Dr. Shapiro's lab at Ben Gurion University of the Negev to get a closer look.

The inventor's guiding principle is one familiar to engineers everywhere: "KISS: keep it simple, stupid." But the applications of the robot prototypes he designs are anything but routine: besides the snakebots, which can carry explosives for military use or slither into collapsed buildings with a camera to search for survivors, there are tunnel-mapping robots that travel in pairs to correct each other's errors, and robots that use magnetic wheels for inspecting ships below the waterline, or adhesive treads for scaling vertical walls like a snail.

In contrast to versatile robot prototypes like Boston Dynamics' "Big Dog," your robots are each designed to solve a specific mobility problem using simple solutions. Do you have a "low-tech" philosophy?

Not really -- in fact, some of the ways we design and manufacture our components, such as 3-D printing, are very high-tech. But we do work with a bottom-up approach. We are in the stage that each robot has a specific task, and we design something for that task. This is the art of science, to take a complex task and break it into small pieces that are easier to handle, then combine them together. Simple is good in general. The greatest inventions are the simplest ones -- like the wheel, for instance.

What problem were you trying to solve with the snakelike robots?

Robots that move like snakes are not new; what I added is the idea of advancing the snake by creating "rolling contact" with the environment. Rolling contact is just what a wheel does: making contact with the environment in a continuous way. In the snake robot, rolling contact is maintained by a traveling wave through the links on the body. I call it a "deformable wheel," because each link rolls on the ground, and when the contact gets to the end of the link, the next link comes and continues the rolling.

Rolling contact is useful for two reasons: one, for measuring how far the robot advances -- you can obtain odometry information just like you can from a wheel. The traveling wave lets us measure the angle of each link in the robot with respect to the environment, so we can estimate how far it crawls. In the future, we can put tilt sensors and accelerometers on each link to get even more accurate measurements. The second reason is for climbing -- you want contact to be continuous in order to maintain contact force with the environment, so the robot can hold itself up.

The "2-D" snake can only move forwards and backwards, because it only uses one traveling wave in its links, but it can also climb between two rigid surfaces. In the "3-D" snake, two perpendicular waves -- one horizontal, one vertical -- travel through the links, and the superimposition creates a screwlike motion. It's more versatile, and can be steered left or right by changing the speed and phase difference between the waves. It should be able to climb as well, but we haven't tested that yet.

What is the principle behind the tunnel-mapping robots, which work in pairs?

Basically it's error correction. There is a unit [in the IDF] that finds tunnels between Gaza and Israel, and they asked us to design a unit that could be driven into these tunnels to map them safely. Obviously there's no GPS down there, so you need to rely on odometry to do the mapping. But with one robot, if there is any slippage during rotating or turning, then you introduce error into the angle of the robot, and after a certain distance you get a very big error in localization.

With two robots, you can gain additional information about the relative configuration of the pair, in addition to the odometry. The arm that connects them has six passive joints that record the relative positional information between the two, and this redundant data allows for error correction when slippage occurs. The arm can also transfer force between the two robots: one can push or pull the other to help it overcome obstacles, which is particularly useful in confined spaces like tunnels.

Your two other prototypes were designed to attach to specific surfaces while moving. What was your approach for each problem?

The first was brought to me by two students who are both officers in the navy, who wished they had a way to assess damage to the hulls of ships just below the surface without sending a diver out. We needed a robot that could attach to metal structures, so placing magnets on the robot's wheels was the simplest and easiest solution. The problem was that there are usually obstacles that it has to travel over, like rivets and seams. Therefore, we mounted all the magnets on springs, so they could adjust to the terrain.

The second problem was inspired by an incident about twenty years ago when a soldier was kidnapped and held on the second story of a building. Having a robot that could quietly scale the wall outside and survey the interior would be very useful. Suction is often not appropriate, because the wall surface is rough or unsealed, so we looked at the snail, which can climb on almost any surface simply by secreting an adhesive. We discovered that a hot glue gun mounted over each tread of a robot worked quite well for imitating this, and it was strong enough to hold much more than the robot's weight.

What are your plans for future research efforts?

I'd like to study more dynamic mobility processes. All of these prototypes rely on what we called quasi-static motion: the robot is stable at all times. And I think the future would be in investigating processes that are more dynamic -- that can use inertia, jumping, or even falling. After all, when we're walking, we actually fall forward. So I'd like to investigate these applications for dynamic locomotion.