The Atacama Desert. which takes up the northern third of the long, skinny country of Chile, is known as the driest place on Earth. On average, it gets less than 1.5 centimeters of rain a year. California's Death Valley gets more than three times as much and Mongolia's Gobi Desert about six times as much. Some spots in the Atacama haven't seen rain in centuries.

To find the robot team, head south from the industrial shore town of Antofagasta on Chile's main north-south highway, Ruta 5. The two-lane road winds through an austere landscape. As far as the eye can see, the land is nearly unbroken by any vegetation or landmark, except for the occasional hand-built roadside memorial, usually beside a bad dip in the road or a "curva peligrosa," designating the spot where some hapless driver met his Maker.

After about 3 hours, around 24.98049 degrees south, 69.90336 degrees west, turn left at the entrance to the Guanaco Mine, and take the dirt washboard road another 42 kilometers to the mine gate. If there's daylight, multicolored piles of slag—toxic by-products of the gold excavating that goes on here—will greet you in the distance. Just beyond the gate sit the researchers' dormitories, a set of low-slung, corrugated metal buildings. You have arrived.

The mine is the sixth and final site that the team is exploring during its three-year project, which began in 2003. About 1400 km north of Santiago, Chile's capital, the mine lies at an altitude of 3000 meters, at the base of the Domeyko Mountains; further to the east are the snow-capped Andes.

In addition to the field team assembled here in Chile, a group of geologists, biologists, and others are gathered in Pittsburgh. They are the science team, and it's their job to parse the data that Zoë collects and then send back a set of instructions to launch the next day's mission. The idea is to simulate, as much as possible, an actual mission on Mars. So at each site, the robot "lands"—that is, it's disgorged from the back of a moving van—takes a reading of its surroundings, and then uploads photos and telemetry data via satellite to Pittsburgh. The team in Pittsburgh pores over the data and then discusses (or, more often than not, argues over) what kind of investigations and maneuvers the robot should do next. The instructions are subsequently sent back, again by satellite, to the robot.

Zoë is solar powered, so it can operate only during the day. The science team receives its data in the early evening, and it spends a good part of the night refining the plan for the next day. The scientists' knowledge of the site is limited to what they can glean from the data sent back by the robot; the field team is permitted to tell them only as much as they need to know to plan their operations. Occasionally, the science team misses an obvious chance to gather data—for example, they could instruct the navigation cameras to collect periodic images, but they don't. The engineers in the field can only stand by and watch. "The rover could give a lot more data or images than the scientists actually request," software engineer Dominic Jonak says. "Sometimes it's as if their eyes were closed."

At breakfast on the first day after landing at the Guanaco Mine—which in the researchers' lingo becomes "Sol 1 at Site F"—the field team hears the plan from the science team. Wettergreen reports that the latter couldn't decide whether to start by inspecting the landing site for signs of life or by sending the robot to a distant point to look around. "As usual, they decided to split the difference," he says. And so Zoë will begin with 10 sequences of fluorescence imaging, traveling several meters in between sequences, and then make a 2-km traverse with a stop for photos, winding up the day with a 1-km traverse.

It's an ambitious plan. Chris Williams, a mechanical engineer at the Robotics Institute and the robot's chief wrangler, heads out early to boot up Zoë, which had been unloaded the day before and left out in the desert overnight. When the rest of the team arrives at 9 a.m., it's immediately obvious that doing anything will be difficult: the wind has picked up to a blustery 70 kilometers per hour (43 miles per hour), with gusts of 90 km/h. It's a punishing wind, the kind that sends unsecured headgear cartwheeling out of reach, turns normal conversations into shouting matches, and makes standing upright a test of will. (Mars gets windy, too, but the effects are far less noticeable, because its atmosphere is less than 1 percent as dense as Earth's.) Wettergreen is worried that a hard gust might catch on the rover's solar panels and launch it. So he decides to wait.

And wait. Done in by the wind, most team members soon retreat to their trucks, emerging at intervals when they feel restless. Occasional updates come over the two-way radio:

"Wind has dropped down to a measly 45.5 miles per hour."

"Glad to hear we're below hurricane force."

"No, wait, it just kicked up to 54.4...Now it's gusting to 53.2."

"Miles per hour or kilometers per hour?"

"I'm afraid that's miles per hour."

"Ouch."

Spending time in the desert, in all its featurelessness, induces a form of sensory deprivation. It's so devoid of obvious life that the sighting of a fly or a beetle or a small green plant is a revelation. For the purposes of the project, though, such macroscopic sightings don't really count. After all, a rover on Mars will never encounter so much as a clump of sagebrush. Amid this paucity of stimuli, something like lunch can take on near-mystical import, even if it's only Fanta soda, apples, and ham-and-cheese sandwiches—a slight repackaging of the ham, cheese, and bread from breakfast.

It's 2 p.m. before the wind finally calms down enough to allow for some robot action. Zoë spends the first few minutes driving comically in circles, the result of some confusion as it attempts to home in on its local coordinates.

After that false start, though, the robot begins dutifully running through the first of its 10 fluorescence-imaging cycles. Tucked behind the robot's removable fiberglass panels, the imager descends from the robot's belly [see photo, "Moving Parts"], and two thin arms emerge, mantislike, to spray water, acid, and dyes on the soil and rocks below. The flashlamps begin to pulse, and the CCD camera clicks away. It takes about 20 minutes for the rover to complete a full imaging sequence; after it's done, Zoë rolls on a few meters and begins ministering to another patch of rocks.

Clearly, there won't be time to complete the day's science plan. Three hours later, with the sun heading quickly toward the horizon, Sol 1 at Site F has ended.