From Our Robots, Ourselves: Robotics and the Myths of Autonomy by David A. Mindell. Out now from Viking.
The two mobile robots Spirit and Opportunity were launched from Earth in 2003 and arrived on opposite sides of Mars in 2004. A suite of cameras, instruments, and tools allows them to traverse the landscape for several kilometers, map the area, and drill and analyze rocks.
Though originally designed to operate for only 90 Martian days (known as “sols”), both rovers operated for many times longer. Spirit got stuck in the soil in 2009 and lost contact in 2010, while Opportunity continues to work a decade after its planned demise. As of 2014 Opportunity had driven farther (40.25 km) than the Apollo 17 lunar rover (35.7 km) and farther than the Lunokhod 2 vehicle (39 km), setting a record for off‑Earth planetary driving.
For all these years the rovers have been directed by NASA’s Jet Propulsion Laboratory in Pasadena, California. There, engineers and scientists sit in windowless rooms, issuing commands for the rovers, reviewing the data, and generally exploring Mars (after the first few months, the “nominal” missions, many of the scientists returned to their home institutions to participate through the Internet).
The rovers are dependent on power for their solar panels, so most operations need to happen during the Martian day, which is about 40 minutes longer than an Earth day. Sometimes a Martian sol corresponds to the Earth day—hence a normal work day for the crews in Pasadena—but other times it is nearly opposite, leading to strange work hours. Some participants began wearing multiple watches to remind them of both Mars time and Earth time. In the long run, the strains on human performance of these unusual schedules, or “planetary jet lag,” actually began to pace the rovers’ scientific output.
Nonetheless, in the first decade of this century, a small number of people learned how to drive to work in the morning and go to work on another planet, much as Predator drone crews, not far away at Creech Air Force Base in Indian Springs, Nevada, were driving to work and going to war on another continent.
It was an unusual kind of work. Geologists who may have chosen their careers because they liked to be outdoors, now found themselves in air‑conditioned rooms, looking at screens and going to meetings while still living away from their families. Their work required more collaboration than traditional field geology, coordinating people and machines over time‑delayed limited‑bandwidth links between two planetary environments.
Bill Clancey is a computer and cognitive scientist who had studied scientists’ use of robots while in isolated environments in the Arctic. When he turned his attention to studying workers at JPL to determine how they learned to work on Mars, he became interested in the scientists’ experience of presence on the distant planet.
Public and press accounts, including those from NASA itself, often refer to the MERs (Mars Exploration Rovers) as “robotic explorers.” But clearly they are not. The robots do not explore on their own, they do not make judgments, and they do not do any science. They are more like remotely operated undersea vehicles than robotic explorers—except with 20‑minute time delays between commands and responses.
On average, Mars is 140 million miles away from Earth, which means it takes about 12 minutes, at the speed of light, for a command from Earth to reach the rovers, and about 12 minutes for any results of that action to be perceived on Earth (actual travel time varies between three and 22 minutes). In practice, this means that engineers on the teams issue commands to the rovers and see the results about once per day.
Observers of these missions often assume that such time delay destroys any possibility for feeling present in the Martian landscape. Clancey’s research found exactly the opposite: the delays work into a daily cycle of interactions, “enabling a feeling of synergistic operation, indeed of being there on the planet.” With the Jason robot scientists learned to turn the control van into a real‑time seminar on the ocean floor. On Mars, this daily cycle enabled deep immersion in the data being sent back, actually enhancing the sense of presence.
Clancey argues the rovers are mechanisms that people “acted through”—extensions of human eyes and hands of people on earth. The rovers are more like programmable, mobile laboratories than scientists, physical more than cognitive surrogates. He writes about the scientists’ experience of “becoming the rover.” In language analogous to that of Predator pilots, scientists talk about “projecting yourself into the rover,” and find themselves turning their heads to look behind the rovers, straining their necks to look around rocks, as though they are physically there. “It’s been some kind of weird, man‑machine bond,” one scientist says. “It has morphed into us and we’ve morphed into it.” Another reports, “My body is always the rover.”
Spirit and Opportunity do not work as autonomous beings, but rather as physical surrogates for the scientists’ bodies and senses. Cognitively, the work remains in Pasadena, shifted in space (by millions of miles) and in time (by the daily cycle). Scientists feel they are working on Mars because their perceptions, their teamwork, the interplanetary system, and the rovers make a kind of cognitive sense. The team on the ground sees things in the world, considers the data and imagery, makes decisions, sends commands to the rovers, and sees the results of their actions. That this whole cycle takes a day instead of the milliseconds that it would take if one were hammering on a rock in an earthly desert is, in the long run, irrelevant.
Chief scientist, Steve Squyres has often commented on the slowness of working with rovers: “It took four years to do a week’s worth of fieldwork! It has unfolded in excruciatingly slow motion.” The sentiment seems odd because it suggests that the major reason to send people to Mars would be speed, which few suggest as a rationale. It will cost hundreds of billions of dollars to get humans to Mars for a few months, whereas the rovers have already enabled work to take place on Mars for more than 10 years for about the cost of a single space shuttle mission.
Scientists recognize a “fundamental fallacy” in the “Geologists could do in a minute what the MERs do in a day” sentiment. The latency time actually favors analysis of the data, thoughtful consideration, and scientific deliberation among the group before the next move. Teams of scientists on the ground can talk decisions through before pursuing the next step—new configurations of work in space and time.
One source of the rovers’ designation as “robot geologists” was the sense that they act autonomously. But the richer idea of immersive presence in a remote environment mediated by 20‑minute time delays allows us to situate the rovers’ autonomy. It makes sense that Opportunity should be able to execute local commands during the long delays between earthly commands—and indeed it conducts numerous local feedback loops and housekeeping jobs, controlling the instruments and keeping the rover healthy, with no human intervention.
But in practice, the rover’s autonomy serves as a resource for the human engineers who command it. For example, the rover can autonomously plan a route around a series of rocks or obstacles using imagery it gathers from its camera, using a program called AutoNav. But to do that it stops every 10 seconds to look at the terrain for 20 seconds. Thus autonomy is costly in time—the rover can drive three times faster when the human planners give it the route in advance—enabling the rover to arrive more quickly requires many hours of human analysis and planning on the ground.
In Clancey’s words, autonomy here is “a relation between people, technology, and a task environment.” One of MER’s robotics engineers was “surprised” that the robot he helped build, when it got into the field, acted more like a “partner” than a free‑acting agent, more like a human collaborator than a technical bot.
Some point out that human presence on Mars would be a more efficient way to do the work. But why the need for efficiency, doing more work in less time? Well, the reply goes, because field time is expensive and difficult to get, you always want to get the most data in the shortest time you can. But with MER, the time between sols was valuably used by the science team to organize their thoughts, achieve consensus, and plan their next moves.
Remember the subject matter: the geology of Mars. This is an environment that hasn’t changed for millions or perhaps hundreds of millions of years. There’s plenty of time to study it.
Sure, one can think of cases where the real‑time dynamics of the phenomena are quicker, and the scientists need to poke and prod in real time. Mud and lava flows perhaps, or the behavior of critters at deep‑sea vents. In planetary exploration, the phenomena under study are slow; the difference between a two‑week human mission (cost: $100 billion) and a 10‑year robotic mission (cost $1 billion) has no relevance for rocks.
Dan Lester, an astronomer at the University of Texas, argues that we need to rethink our traditional concepts of exploration. “When Congress starts using the phrase ‘human presence’ to authorize a $17B agency,” Lester writes, “the phrase takes on some importance.” Even though the human scientists and their Mars rovers are clearly conducting exploration, he points out, NASA still uses the term “exploration” to refer only to human spaceflight. In doing so, NASA suggests a fallacy: that presence through a time delay is not real presence.
Everything in Bill Clancey’s studies of the Mars rovers teams, everything in his richly empirical and systematic data, contradicts this assumption. When the thing you’re studying hasn’t changed in millions of years, why is 20 minutes too long to wait?
My goal here is not to argue for or against human spaceflight, the justifications for which have always been, and will continue to be, primarily about engineering demonstrations, national prestige, and international competition more than any cognitive or motor task advantages. Rather, spaceflight offers a dramatic and salient example of the relationships among space, time, task complexity, robotics, and human experience. In low earth orbit, with relatively low latencies, telerobotic systems can accomplish a great deal through direct manipulation. On the moon, with only slightly longer delays, teleoperation offers great potential not yet explored by NASA. Mars, with its much longer delays, requires distributing human action and agency across time, through both work practices and technologies like autonomy, and creating new ways of working. None of this precludes the experience of presence in the Martian landscape, and in fact each enables collaborative presence, new ways of doing science, and new ways of exploring our world and our solar system.
From Our Robots, Ourselves by David A. Mindell, published on October 13, 2015 by Viking, an imprint of Penguin Publishing Group, a division of Penguin Random House LLC. Copyright by David A. Mindell, 2015.
Update, Oct. 23, 2015: The headline of this article was updated to better reflect its content.