Robots with Legs
Interview with David Wettergreen: Part I
David Wettergreen is an associate research professor at the Carnegie Mellon University Robotics Institute. Wettergreen works in the institute’s Field Robotics Center, designing and testing robots for deployment under harsh, inhospitable conditions. Many of the projects he has worked on were funded by NASA, as part of its process of developing robotic spacecraft to explore nearby worlds. In this first part of a four-part interview, conducted by Astrobiology Magazine’s field research editor, Henry Bortman, Wettergreen discusses his early work with walking robots.
Astrobiology Magazine: What is the focus of the Field Robotics Center?
David Wettergreen: The charter is quite broad. The original projects that Red Whittaker (the center’s director) started in the 1985 time frame were fielding a series of three vehicles to enter into Three Mile Island right after the accident there. One was a vehicle put together quickly, a radiation-hardened platform with cameras on it, to go in and take a look, and there was a second vehicle that had coring tools. The question was, How far had the radiation penetrated the concrete walls? And then there was a device that was a real workhorse, built for demolition, and also to wash it down.
In 1989 we won our first contract with NASA to develop a Mars rover concept system. At that point the Mars reference mission [NASA’s design requirement] was to go 1000 kilometers, to be able to climb a 30-degree slope, and to be able to surmount a one meter obstacle. So we created a walking robot with an unusually efficient “circulating gait.” The robot, called Ambler, was 4 meters (about 13 feet) tall so that it could carry all of the computing and sensing needed to be autonomous while stepping over boulders.
AM: What happened with that?
DW: We made it walk. It had a laser scanner on it, one of the first scanning laser range finders. We showed that we could autonomously map terrain, select footfalls and precisely control the motion of this large walking machine. It had six legs, mounted on two vertical stacks, three legs to a stack, that could circulate fully around. The two stacks were connected above by a bridge. Ambler walks the way an egg beater beats, with alternating circulating blades, or legs. It picks up a foot, swings it from behind through the two stacks and places it out in front; and then picks up a foot from the other side and swings it through. All of the legs keep cycling around, which turns out to be incredibly energy-efficient due to the long strides. An insect, which also has six legs, walks by picking up its front leg, then moving its middle leg to that spot and moving its rear leg to where the middle leg was, taking many short steps. We were trying to be more efficient than insects.
AM: But NASA never sent a walking robot to Mars? How come?
DW: Shortly after that the Mars reference mission changed, from 1000 kilometers, steep slopes, big boulders, to “Land and go a couple hundred meters and deploy an instrument.” The Sojourner rover made a whole lot more sense for that scenario. But that is the function of the Field Robotics Center: to start with a specific application and look out into the future, to come up with new concepts and new ideas about how to achieve the goal. Technology has advanced quite a bit since then, but if you turned around today and said you wanted to go 1000 kilometers on Mars and climb huge slopes and be able to go over boulders, there are still good reasons to look at using a large walking machine to do that.
AM: Did you build any other walking robots?
DW: Yes, after Ambler we built a robot called Dante. That project came about in some ways because NASA is interested in using Antarctica as an analog, to learn about how to explore interesting scientific locations here on Earth so that we can extrapolate what we learn to the other planets. There was interest in a couple of things in Antarctica at the time – this was 1990. One of them was Mount Erebus, which is the only active volcano in Antarctica. It’s interesting because it’s in a cold place, and because it’s one of only three volcanoes in the world that has an active magma lake, where there is fresh magma cycling up to the surface and descending back down. In most volcanoes, magma moves closer and closer to the surface, sometimes collecting in chambers during its one-way trip up, but with Erebus it’s actually cycling up and down. When you look at it, it looks like oatmeal boiling. And it throws up lava bombs. So there is great interest in getting samples of gases from the surface of the magma lake, and potentially scooping a sample of the magma.
AM: How hot is it?
DW: It’s about 1800 degrees, it’s pretty hot. To get samples, you have to use like a dipper on a long ceramic stick. Several times folks had tried to get down close to this lake. The crater rim is about 500-by-600 meters (about one-third of a mile) across, and it’s about 150 meters (about 500 feet) down to the floor of the outer crater and then another hundred meters (330 feet) down to the lake itself. Years before our project, mountaineers had gotten into the outer crater and over the rim of the inner crater, but the lava bombs flying out were a little bit too dangerous. Somebody got close and then their rope was hit by a lava bomb – so everyone got out, for good. It’s clearly not a place you want to send people. But there were scientific questions about the chemistry of fresh magma, understanding this volcano in a very cold environment. So we built another walking robot, Dante, to descend into that crater.
AM: Was Dante’s design similar to that of Ambler?
DW: No, it was quite different. Dante had eight legs and a low- center-of-mass for stability on slopes. The slopes there are as steep as 80 degrees, nearly vertical. Dante had legs but it was more of a rappelling machine that hung from a tensioned tether.
AM: Rappelling? I get this image of the robot jumping off the rim and bounding down the crater wall.
DW: No. It’s not like a gung-ho Marine zipping down the wall. If you talk to cavers, they’ll tell you that when you rappel you don’t want to bounce much, because as soon as you lose contact with the wall you have no control. Staying in contact is a better way to walk down the wall. It’s more like an on-the-wall ballet.
AM: Did Dante ever get deployed?
DM: It was deployed in Antarctica in December 1992, but we had a failure in a primary component in the tether and for logistics reasons we couldn’t get a repair out to where we were. We demonstrated operation of the system but we didn’t get very far into the crater, so we didn’t get the scientific samples that we wanted.
A couple of years later, for the second phase of the Dante project, we reworked a fair bit of the robot. A lot, actually – Dante doesn’t exist anymore because it was torn apart to create Dante II. John Bares was leading the effort and we wanted the robot to be completely bomb-proof.
We were planning to take Dante II to Mount Spurr, which is a volcano in Alaska. Mount Spurr is interesting because it’s about 80 miles from Anchorage and, in prevailing conditions, it is upwind. When it erupts, it sends up huge ash clouds that throw ash over Anchorage. In the 1990s Anchorage was the busiest cargo airport in the U.S. on the main route to Asia. When you flew into Anchorage airport, there were a dozen or more 747 cargo planes parked on the tarmac from FedEx and DHL and other companies. And when ash clouds blow into the engines of 747s it’s really bad. In 1989 a passenger jet flew through an ash cloud from the Redoubt volcano nearby and it ruined all four engines. They landed safely but the engines had to go in for $80 million of service. And then in 1992 Anchorage was blanketed with ash from Mount Spurr.
In 1993, when the Dante II project began, Mount Spurr was becoming increasingly seismically active and the U.S. Geologic Survey (USGS) didn’t know whether it was preparing for a major eruption or not. Previously the USGS had volcanologists who would go down into the crater and measure gases. But they could see the seismicity was up and gases were coming out. So that became another target for a robot, to descend into the crater. Mount Spurr doesn’t have a magma lake, but there are fumaroles and toxic gases, and because of the seismicity, there was almost a constant crack and rumble of boulders rolling down off the crater walls. So we dropped Dante II on the rim, attached it to an anchor and satellite dish, and then everybody left. We watched from Anchorage while it descended to the bottom of the crater and sampled gases from the fumaroles and measured temperatures. By looking at different gas ratios, scientists could figure out what state of activity the volcano was in, and they determined that it was in a milder state of activity than they thought – it wasn’t ready to blow up. And it still hasn’t. That project was successful technically as well as scientifically.
AM: It sounds like there’s a theme here of robots going places where people can’t go, or can’t go safely.
DW: That is exactly correct. The overarching goal of the Field Robotics Center is to develop robotic systems and deploy them to the field. But my personal interest is in exploration, creating robots to go to the places where people can’t, or where it’s too dangerous or too risky, or too expensive. Places where there are interesting scientific questions that we want to answer, and where a robotic solution is the best one.