CHAPTER THIRTEEN

Fallback Positions

THE MINING CAMP IS ABOUT THREE MILES DOWN THE harrowing road Finch had catapaulted us over the previous afternoon. Breakfast takes place at 9:00 A.M. after two crews of miners have eaten, the first at the end of the night shift and the second beginning the daylight rotation. The men who work the salt mine labor in three eight-hour shifts, digging and loading salt into trailers and hauling the coarse, white cargo to the port at Iquique where it is shipped in gigantic freighters. The reserve in this mine is enough to supply the world’s demand for salt indefinitely.

Breakfast is usually quiet and congenial with a smattering of jokes and a brief discussion of the plan for the day. The food is tasty, bursting with carbs and fat: Delicious homemade bread with jam, greasy sausages, and slices of cheese are standard fare. The scene reminds me of the mess hall of my summer camp when I was a kid, for it is warm inside, compared to the chill outside, and the cacophony of the background, with music and laughter in the kitchen, provides a brief but welcome contrast to the isolation of the base camp, seemingly far from civilization.

Arriving at the mining camp, some folks carry a towel or a personal hygiene kit and duck into the bathrooms immediately before or after eating. Ample rolls of toilet tissue are stored in the trucks. The rolls of tissues are like trophies, removed from the truck and stacked at the end of the mess hall table. Periodically we will slide out of our seats, walk around the table, snatch the toilet tissue and disappear, and then, after a while, reappear, returning the roll to its position of prominence. David Wettergreen seems to judge when the crew is ready to return to base camp by watching the toilet tissue disappear and reappear on the table. When it looks like no one else will be snatching one of the rolls, Wettergreen stands up.

The team is congenial most of the time, with only a few minor squabbles and personality conflicts. Good behavior is normally the rule on field trips, but there have been glaring exceptions. In Antarctica with Nomad, “there were two guys who got into a nasty ongoing conflict over what ‘autonomy’ means,” Mike Wagner tells me. “The first guy insisted that Nomad autonomously found meteorites based on his definition of autonomy and the other guy disagreed with the first guy’s definition of autonomy. They went at it until we went home.”

Although everyone is congenial now in 2004, I can sense a subtle, mounting wave of tension—a collective and pervasive nervousness gradually infiltrating the air. Zoë is designed to function as an integrated unit, with one system reliant on another system. And most everyone on the robotics team is responsible for a part of the system that makes Zoë effective. For Wagner, navigation and vision and Tompkins’s Tempest, the planner. Anything pertaining to the operation of the mechanical or electrical functioning of the robot falls on Teza’s and Heys’s shoulders. Everyone knows that these few remaining days of shakedown to get Zoë functioning as advertised are key to the upcoming science operations the following Monday morning.

Thus, everyone is anxious to check out their particular system on Zoë and eliminate lingering problems so that the OPS runs without a hitch. Or, at the very least, without a hitch for which they are directly responsible. This is especially critical now, since Zoë had not been ready on time for adequate testing in Pittsburgh. No one is comfortable with how far his project has progressed, nor confident that the systems will perform up to par, most especially the youngest and least experienced member of the field team, Dan Villa, responsible for the position estimator.

Simply put, the position estimator (PE) tells the robot and all of its supporting systems where it is in the world. This information is essential. Neither Tempest nor the navigator can plan or map with accuracy without a starting point. And neither system can reevalute and update, as designed, without precise information concerning how far and how fast Zoë has traveled.

The PE gathers information from encoders that count how often each wheel turns; from a sun sensor that records the position and angle to the sun and from which heading is inferred; from a fiber-optic gyro, or FOG, that measures the heading of the vehicle; and from a potentiometer, or POT, that measures wheel angles. All of this information and much more—every morsel of data supplied by Zoë’s myriad of sensors, which includes acceleration, distance, direction—are fed into a filter that munches it all together and provides an estimate of where the vehicle is. The PE updates and communicates position data ten times per second.

Prior to this point, no one seemed to pay much attention to the PE in the planning meetings. Hyperion’s PE had functioned efficiently on Devon Island. Matt Deans, a Robotics Institute grad now working on the LITA project from his new home at NASA Ames, had updated the software after the first LITA OPS in 2003, making it ready for Zoë. The job of debugging the new version of the software and integrating it into Zoë fell to grad student Dan Villa, a tall, lanky, Seattle native. Since Villa was the least experienced member of the LITA staff, I assumed that the PE was a low-level program that required some attention, but not a great deal of concentration. I was wrong. No one could have predicted that the PE was going to change the entire scope and breadth of the OPS.

For Villa, the LITA opportunity had emerged out of the blue. He had come to Carnegie Mellon’s one-year master’s degree program almost as a lark. Robotics was interesting, but he was not driven or obsessed by it, as were some of his classmates. He’d be in and out quickly. But then David Wettergreen offered him a research assistantship—free tuition, a stipend, and another year of study. So why not? Villa had never been entrusted with anything as challenging and stress-producing as a NASA-sponsored project, and he was feeling rather inadequate. After joining the LITA team, he told me, “I looked around and asked myself, ‘What am I doing here? I feel lost, like I am a fake!’”

Wettergreen was confident Villa could do the work, primarily because there was not a lot of work to do. The PE had not been a problem on Hyperion. As with the vision system, the PE had worked well in simulation and in the brief testing period on Flagstaff Hill. I had observed part of that testing period in the early spring when Wagner and Wettergreen were attempting to connect Hyperion with Vijay Singh at NASA. I remember meeting Villa that day because I appreciated the uncomplicated definition of the PE he provided: “The thing that tells the robot where it is.” There was no reason to assume that the PE would not be in perfect operating condition. But, quite simply, it was not functioning adequately now that it was needed.

With Mike Wagner struggling to bring clarity to the vision system and with the PE completely unreliable, most of the men sitting in the mess hall that morning share a silent deep-seated trepidation: That the science operations will be a disaster—and all of their efforts individually and collectively over the past two years since the LITA project was conceived, could be for naught.

Perhaps no one is more anxious than Paul Tompkins, whose precious Tempest will be ineffective without a fully functioning PE. And since he will receive his PhD and leave Carnegie Mellon in a few months to work at NASA Ames, this may be his last opportunity to give his undivided attention to Tempest and Zoë.

Both Tompkins and Wagner understand the pressure on Villa and the embarrassment Villa is experiencing. In the Arctic in 2001, Tempest was such a disaster that it polarized the entire camp. In protest, one rebellious contingent staged a permanent work stoppage. Those choosing to continue with the Tempest experiment carried the ball. “Nobody was happy,” said Tompkins. “It was a dark time.”

To their credit, no one on the LITA team publicly voiced any complaint or criticism toward Villa. Everyone understood that robotics is frequently frustrating, a gauntlet of trial and error and unyielding disappointment, as Tompkins had experienced with Hyperion and Wagner was enduring now. But Villa could not help feeling ashamed. “I felt horrible, like everyone was thinking that I had ruined their experience.”

But this morning, after breakfast and back at the base camp, three short days before the OPS would begin, even their always calm leader and anchoring spirit has the jitters. Wagner and Heys watch as Wettergreen struggles to find a place on Zoë for a white plastic slab—called a “white reference”—which resembles an oversized cutting board or serving platter. “It is 98 percent white, and believe it or not, it cost $1,500.”

Zoë is equipped with a VNIR—a visible near-infrared spectrometer, which measures energy in order to assess minerology. The white reference determines how much sunlight or energy from the sun is there to begin with before the spectrometer readings occur. During last year’s OPS with Hyperion, the spectrometer was carried in a backpackable unit, separate from the rover. Now it is attached to Zoë, but Heys has not established a place for the reference. To calibrate existing sunlight a person holds up the reference and points it at the camera.

“I’d like to dispense with it and get the $1,500 back,” says Wettergreen.

“Let’s just see where it will fit,” says Heys. “We’ll use Velcro to anchor it down.”

“That’s our problem,” Wettergreen replies, “we don’t have enough Velcro on this robot.” In last-minute situations, roboticists will inevitably employ what they sometimes refer to as the “Velcro solution.”

“OK,” Wettergreen says, opening and closing drawers and muttering to himself, “so where’s the damn Velcro?” The area around the three OPS tents is in as much disarray as is Zoë’s corner of the High Bay.

“I’ve got my master’s degree in Velcro engineering,” says Wagner, overhearing the conversation. “I can find Velcro anywhere and anytime.” He immediately produces the roll of Velcro, concealed under a plastic container directly behind Wettergreen.

While Wettergreen and Heys are engaged in Velcro engineering, Alan Waggoner is staring down at Zoë, as Jim Teza and Mike Wagner lift Zoë’s lid in order to debug one of Zoë’s computers, which suddenly doesn’t want to boot up. “Zoë is making a funny noise,” Waggoner says.

“I don’t hear anything,” says Heys, now finished with the Velcro, “but then, too, I keep imagining I see Zoë moving around in the middle of the night.”

“I think that’s called a dream,” says Wagner.

“I am not dreaming that this weather station is dead,” says Finch, pointing down at a white box with a bunch of dials, resembling a portable apartment refrigerator. “I don’t know what happened, but all of the data [are] lost.” The science team had requested a weather station to do temperature and climate experiments. “I guess we can do without it. We have no choice,” he adds.

“Just another thing going wrong,” says Heys quietly. “We are getting down to the wire, and I am getting more and more worried.”

Jim Teza has gone into the supply tent to find a monitor to hook up to the computer to try to find out why it won’t boot up, and Wettergreen, who has followed, says he wants him to go into town with Finch to pick up some supplies.

“But I want to be doing some of my own work.” Teza has developed new lithium batteries that will double the power on Zoë. They cost $25,000, a significant investment, and he had been hoping to test them on Zoë today. “This is not a good time for that,” says Wettergreen.

“There never seems to be a good time,” Teza answers. “I can’t get my work done.”

“It can’t be helped,” Wettergreen replies quietly.

Teza says no more. Teza is dark and tall, a ponytailed, hippie type about Tompkins’s age, who hardly ever speaks. And when he does, he is so soft-spoken his colleagues are frequently asking him to repeat what he says. Teza has a 1996 degree in electrical engineering from the University of Pittsburgh, and a passion for literature (mostly, but not exclusively, science fiction) rather than robotics. “But I was always interested in computers, only I wanted them to do something in the physical realm instead of staying so stationary and static. It was a disappointment when I realized the limitations of robots. I was naïve.” Now, he says, robots are catching up to his intellectual expectations—slowly. Teza works at Carnegie Mellon on a project-by-project basis, as do most engineers in this software-driven science hub—not particularly conducive to stability. Teza tried programming to enhance his capabilities and found it “tedious and exhausting.” Zoë is Teza’s fourth robot and this is his second foray into the Atacama.

He produces the Samsung flat-screen monitor to connect to one of Zoë’s four computers, along with a keyboard to use to debug Zoë. The computer sounds like it wants to boot up, but the monitor is dark. Zoë’s six computers were purchased at the start of the project early in 2003, while the monitor was purchased right before they packed up for the Atacama—a light, flat screen selected for clarity and portability. For a while, there’s hope:

“It will be all right,” says Wagner.

“Give it a little time,” says Teza. “The computer is slow.”

Now we were all standing around Zoë and waiting. “This is not going to happen,” says Wagner, now looking over the manual that comes with the monitor.

“What’s wrong with Zoë’s computer?” asks Waggoner.

“I think there’s nothing wrong with the computer,” says Wagner, “that a little tinkering won’t fix. But it is a few years old, and the monitor is too sophisticated to display the data. Too new for the job. We didn’t consider that when we bought it.”

“Thank God we are in Chile and not New York,” says Waggoner. “Here we can find equipment that is very low-tech.”

“Now I am really worried,” says Heys.

“Oh boy,” jokes Waggoner, “Zoë is having a nervous breakdown.”

“That makes two of us,” Heys says.

 

LATER THAT DAY, Wettergreen and I are driving to the airport to pick up Vijay Singh, the NASA programmer who wrote the executive. I ask how he thinks things are going. “We had a clear set of objectives to start with, but much less of an idea of how everything was going to work once we got here. By experience I know that things will go wrong. The real test is how good we are at improvising and dealing with our difficulties and how to recover from them. We are always talking about our ‘fallback’ position.” Wettergreen is referring to how the roboticists adjust to changing priorities and the inevitable decreasing expectations precipitated by technological uncertainty.

“How do you think everyone is holding up under the pressure?”

“These are all good people, and I can count on them to make individually good decisions. But what I need them to do is to make collectively good decisions. Everyone has to do their work and, at the same time, pitch in and help others. We are all interconnected,” he says.

“Are you nervous about what is going to happen on Monday when the OPS begin?”

“I feel responsible. NASA has poured a lot of money into this project and expects us to get results, and I don’t want to disappoint them.”

He suddenly downshifts into second gear, stomps the throttle to the floor, and whips the Toyota around a turn. I fly up in the air, bounce against the door, hold my breath, close my eyes, and give thanks for seat belts and roll bars. So even the cool, calm, and collected David Wettergreen unleashes a wicked, crazed frustration when behind the wheel of a truck. Wettergreen’s aggressive driving makes Finch seem timid. Here again the roboticist’s syndrome—explosive wildness as a balancing point and escape valve from the stationary intensity of the keyboard and display. Or, in this case, an anxious and ecstatic moment of relief from the stress and fear that all of his plans may go awry at the critical moment.

Wettergreen doesn’t slow down when we pass the mine. In fact, he cranks the throttle. “This is not going to turn out to be a perfect field trip, I know that.” The throttle is nearly to the floor, the desert whizzing by in a burnt red blur. “That is what you learn early in robotics. Some projects are not going to work perfectly, and you are definitely going to have to adapt and work around the systems that are not functioning. These field experiments are lessons in prioritizing, making choices. You learn to make the most with what can be accomplished.”

We are nearing the airport, and Wettergreen is gradually slowing down, thank God. Failure in itself is only a setback, but not necessarily irreversible, he observes. “A negative experiment can sometimes be a good thing, as long as you are aware of your mistakes and figure out how to make them more positive the next time.”

“That seems to be the lesson of robotics,” I say. “It’s the ‘never give up’ scenario.”

Robotics, he answers, can’t be compared to disciplines like engineering or surgery. “I mean, there’s no plan or manual. Nothing is structured or expected. Most of the things you try don’t work right away. It requires constant diligence to make a system work, and sometimes, even after all of the time and effort, blood, sweat, and tears, they don’t work at all. That’s the nature of the beast.”

 

AT THE AIRPORT, I decide to go into town with Finch and Jim Teza. Wettergreen asks me to copy a shopping list he has drawn up in his notebook—at least I had thought it to be a notebook until I got a better look at it. Black and spiral-bound, it is actually an artist’s watercolor journal. It is made of fine, high-quality, unlined pages. I flip through it quickly. There are lists and notes everywhere in very neat script or clear block letters. Wettergreen often inserts a square or cube in front of each point to bullet the items on his lists. “I like to draw lots of pictures,” he says, “especially when I am in meetings or sitting around and thinking.” I turn the pages as he is speaking. Once, paging through his notebook, I spotted a series of delicate watercolor landscapes. German farmland, painted on a recent trip. “My mother was an art teacher,” he says. “My brother is a potter.”

Many of the computer geeks here maintain diaries or hand-written logs alongside their computer records and code writing. It makes me feel good—and still relevant—that pen and paper are respected tools of the trade and are not going to be squeezed into cyberspace, at least for the moment. And they all have books they are reading—novels mostly—except for Teza who is more eclectic. On the way into Iquique, Teza discusses John McPhee and Diane Ackerman, authors I admire. He peppers me with questions about the lives of writers as we wend our way along the winding highway that follows the shoreline into town.

Teza is probably the most distant and reticent member of the group—a soft-spoken loner who rarely complains. Despite their disagreement earlier in the day, Wettergreen is sensitive to the individual personalities of each member of the group—perhaps Teza especially. A few days later, during dinner at the mining camp mess hall, Teza was missing. It had been a long day. Everyone was tired and hungry; it was nearly 9:30 P.M., but Wettergreen got up from the table without eating, jumped back into the truck, crashing through the rutted road in the pitch black to find out what had happened to Teza. They returned together a half-hour later. Teza had fallen asleep in his tent and wasn’t aware when everyone mustered in the dark for dinner. “Teza is so quiet,” someone remarked, “you don’t know when he isn’t there.”

The security guards at the supermarket in Iquique are wearing bullet-proof vests. Finch calls this “the intimidation factor,” the Latin American macho experience. Finch’s mother is an English teacher and his father is a physician. This is his second year working with the LITA team. After last year’s OPS, Wettergreen arranged for Finch and a companion to work at the Robotics Institute for a month. Finch was graduating college this year and hoping to study robotics at the RI, beginning fall 2005.

I am aware of the country’s history of violence and political unrest, but firearms and protective vests seem so unusual in relation to what I have observed in Chile so far. The people I meet seem gentle. Most of the people in service positions—from supermarket employees to airline reservationists—wear neatly pressed unforms, with a quiet and beaming essence of pride. The stores are much cleaner and better organized than in parts of the United States, and even the poverty-stricken sections have a dignity and neatness to them.

In the supermarket, we heap two dozen five-gallon jugs of water and 200 rolls of toilet paper into three shopping carts, but no one seems to bat an eye at checkout. Stocking up for the desert is evidently a normal procedure; there are many scientific expeditions in the Atacama. We find a duty-free store in a modern indoor shopping mall and purchase an inexpensive monitor for $80. Next, we locate a store similar to Home Depot, where we buy bolts, cargo straps, and lumber. As we hop from place to place, Jim Teza wonders aloud why we are getting all of this material, especially the lumber, and why we have been sent on an errand without being told the objective. We find out later that the materials are for Stuart Heys in response to Wettergreen’s request that he build a roof rack on one of the trucks to transport Zoë to the OPS site. We have also rented an SUV we call “the soccer-mom van” to use as our base camp and wireless network hub in the field.

The fact that Finch and Teza are unaware of the reasons for the materials and that they are somewhat annoyed by their ignorance is indicative of the difficulty of maintaining a stable group dynamic here. Working together on the same project for a long period of time and now living together, a certain intimacy develops in which everyone assumes that he is totally aware of what everyone else is thinking and doing. But that is often not the case. When I ask Alan Waggoner and his assistant Dave Pane if Zoë’s inability to see or know where she is disturbs them or their work, they both look up at me in surprise. “That’s for the roboticists to deal with,” Pane says. “We’ve got our own stuff to worry about.”